           Me la Uacintyn Loglytuan *    Number 1 -  July 1986

*(Washington loglan-worker - the old metaphor behind Loglantan (Loglan
language talk) would result in Loglylentaa(n) which seems too long. Of
course log is free as an affix and could be assigned to generically
reflect all of the L-prims related to Loglan (culture, language, and
people), resulting in Loglentaa(n). In any case, there are as of yet, no
speakers of the language, and I want people who are working purposefully,
which is what turka means).


WHO - From Bob LeChevalier(rjl). Washington DC area Loglan worker, and
currently coordinating Loglan dictionary update work. Intended originally
for DC area loglanists, the audience has expanded to include any
loglanist who is sufficiently local that they may be interested in
special interest group (SIG) activities, and those who have evidenced by
past contribution that they are capable and possibly willing to work to
complete the current language effort. It is my hope to inspire other
local areas with a high concentration of loglanists to form their own
SIG; I have therefore included all known loglanists in the U.K., Boston
area, and Raleigh NC area, as areas where I know that there is interest
in establishing such SIGs. I've called the majority of you, and the
interest is certainly there.

When SIG membership is clear from your responses, I'll include an address
list. Please let me know phone numbers, and whether you wish them
released. I do a lot of business by phone, and I expect others do as
well. But I also have a personal standard to not release phone numbers
without permission. If you have volunteered your number in Lognet, I will
assume it is releaseable.

WHAT - This newsletter was originally intended to bring together all of
the loglanists in the DC area. There has been a dearth of information
about what is going on in the language, and there are a lot of 'coiled
springs' out there looking for a language to work with. I found that
there was desire for information about other projects beside the
dictionary, and a significant interest in volunteering to help in those
efforts. In attempting to determine the status of those efforts for this
newsletter, I have committed to sending this to a much larger audience
than originally intended. As such, at least for the nonce, the newsletter
is available for anyone who wishes to be active in the Loglan community.

I am putting this out unofficially. The newsletter, and the DC SIG that I
am trying to organize, are not part of the Loglan Institute. We overlap
Members of the Institute, ex-subscribers of TL, and miscellaneous others
who have expressed interest in the language. As such, all statements made
in this are the opinions of the author, and are not in any way to be
construed as official Institute policy. As such, the author of all
material will be identified. At the moment, I'm paying all costs - which
will be some 20 to 30 dollars this issue (not to mention the phone
bills); I will accept contributions, but there is nothing required except
interest at this time.

On the other hand, because the Institute is a vital part of the Loglan
effort, and Jim Brown(jcb) is the inventor of the language, all material
will be made available to him with an open invitation for him to respond
to any and all topics raised in these pages. In addition, all material is
being given to Ed Prentice, or to anyone else who serves as Lognet
editor, with permission to use any or all material herein. Since neither
I, nor anyone else except possibly jcb, can be considered a master of
either the language or the ongoing work, I expressly invite him or anyone
else to offer any corrections of fact stated in this publication, subject
only to an attempt to constrain against 'political dispute' (see below),
which is not the purpose of this effort.

Anyone may submit material. Send it to me at the address below. Please
distinguish private correspondence from SIG submissions. Preferred
material is:
     a. self identification, especially if you have never been identified
in Lognet. Among other things, I suggest giving your background in Loglan
related subjects, your Loglan-related interests, areas you have or would
like to work in. How much time you are willing to spend on the language,
and whether you especially are interested in working face-to-face, by
mail, network, or phone with others. Computer information will help
others who are designing Loglan software.
     b. anything related to ongoing projects covered by this newsletter.
     c. news about available Loglan-related materials
     d. reactions to language proposals, GPA materials, etc., of general
interest
     e. questions and answers of general interest
     f. technical material of general interest. This may be published as
Appendices, so people can read the news without being bogged down in
details. I also may eventually include such material for selected
interested people only, to reduce costs. In such case, I'll list the
material in the newsletter, and let people write to ask for it. If we
start publishing word lists or lengthy translations, this may become
necessary.

Since I'm a slow typist, I'll obviously give priority to computerized
submissions (CP/M or MS-DOS), or short hand/type written pieces. Longer
stuff not on computer format will get in when I (or some other volunteer
typist) has time.

WHERE - The DC SIG is primarily intended for those who live in the DC
area. jcb has contacted at least one Delaware resident and suggested he
contact me. As such, I am opening the SIG to any who express interest
without geographical constraint. My address, for correspondence related
to this newsletter or the SIG or dictionary work is:

                           Robert LeChevalier
                             2904 Beau Lane
                            Fairfax, VA 22031
                       phone: 703-385-0273 (home)
                              703-847-4465 (work)

WHEN - The DC SIG will tentatively hold an organizational meeting on
Saturday, 26 July, so we can all meet each other. DC area people will be
contacted with specific times. Others who would attend, please call or
write and I'll let you know. Guest space is available at my place for the
weekend, and I am located on the DC Metro so that I can be easily reached
without a rental car from Amtrak Union Station or Washington National
airport.

There are also plans to hold a Logfest at my house in the late August/
early September time-frame. I am suggesting any of the weekends including
16 August, 7 September, or possibly Labor Day weekend - 30 August. (I
have plans for the 23rd.) The time-frame would take advantage of anyone
who may pass thru on the way to the World Con SF convention in Atlanta,
since I know several Loglanists who usually go to such things. Later
tends to be better - DC is noted for August heat - but I'm flexible. All
Loglanists are invited, and we'll be trying to have versions of the
various unpublished software available for demo, at least on PC-
compatibles, by then. If you are interested, and would like to vote on a
date, let me know. Try to express your interest prior to the SIG meeting
on the 26th.

WHY - Loglan is achieving stability again as a language. jcb is writing a
summary of the language (NB3), and then is planning on dropping out for
three years or so. I'd like to talk him into an extra 6 months after NB3
to help consolidate one or more SIGs like this one to carry on active
Loglan work without requiring his active involvement. To do this, we must
convince him of the dedication and interest of the community that would
make further such effort worthwhile. I believe the existing community has
more potential to carry on active Loglan work than will a half-time
graduate assistant (HAGA), although I agree that there is more than
enough work in Gainesville with the Institute to justify such a person.

Most Loglanists I have talked to have cited the following as reasons for
their current lack of involvement. The SIG is specifically intended to
counter those reasons - by any and all means feasible.

     a. The death of The Loglanist(TL), and Lognet(LN), have left the
community with little current news as to what is going on. This is
coupled with the fragmentation of the community into 'members', 'TLers',
and various other interested people including those of the above who
lapsed or resigned.

     b. A dispute that occurred in the 1983-85 time-frame which has been
labelled as 'politics'. There were some personality conflicts among
active workers, as well as some significant technical disputes, which
culminated in an apparent struggle for control over the Institute. (I was
not involved, and intentionally so. I consider myself strongly loyal to
the Institute, and to jcb's efforts as father of the language, and saw no
benefit in the dispute to either the membership or the language.) It is
my intention to support no debates over Institute policies or decisions,
and to maintain as apolitical a stance as possible, given that most
Loglanists are highly intelligent, creative, and individualistic.

     c. A large portion of the Loglan community are not skilled in either
or both of languages or linguistics, and the technical level of the
material in TL was far above that of L1 in those areas. The vocabulary is
arcane and somewhat jargonized, and many L1 readers have not been able to
follow it.

     d. The work has in general not been divided up so as to allow a
person with little time to feel that da's contribution can be worth the
time commitment, even if small. The primary low-effort activities, taste-
tests and word-making, have been slow, multi-person efforts with little
response given to give da a strong feeling of participation.

 To get the various tasks on track, accurate direction must be given, and
the workers must be kept informed of progress. This newsletter will
attempt to do so for all projects which SIG participants are involved in.
It is my belief that 2 hours a week on average by any of you will enable
a major contribution. If the whole group of about 30-40 that I expect
will continue to receive this can overall average 3-4 hours a week, our
total effort will be the equivalent of 2 full time people or 4 HAGA's.

     e. While there have been calls for volunteers for many activities,
there has been too little information for people to judge whether they
can meet the commitments of volunteering. There has also been a lack of
clear direction as to the procedures of the various activities. In the
case of the dictionary, luckily, jcb has written me a 17-page Manual of
Dictionary Reformatting, and I've had a weekend in Gainesville to gain
further details, coupled with 9 years of TL and LN commentary suggesting
changes and means for dictionary rewriting. Most of the other tasks, on
the other hand, have relied on close direction, interaction, and feedback
with and from jcb - who is after all only one man. This has been
heightened by the haphazard entry of Loglan into the computer age, with
inherent compatibility problems, and a general tendency away from
correspondence as a means of communication. Those efforts being
coordinated through the SIG will obviously require less of jcb's time.

I'm considering building SIG membership on a couple of computer networks.
Compuserve and PeopleLink (American Home Network) have been suggested.
The former has wider coverage. The latter is cheaper. Any others, or
votes for the above? Such networks, and an emphasis on quick
redistribution of input data will hopefully keep you better informed and
interested

     f. Many people want to learn the language. Books, and computer
programs will help, and more current tapes would also be useful. But most
people want someone to speak to or write to. This is the main purpose of
a SIG - to form a subcommunity of people to work together and to
interact. Those SIG members who are geographically close to each other
should meet. Its a lot easier to learn to speak a language if there is
someone to listen and to give feedback.

     g. There has been no accurate information as to the specifics of the
language available roughly since about 1979-80, when the original Loglan
books (grammar-L1, and dictionary- L4/5) were sufficiently outdated to no
longer serve as a basis for active work. As such, many Loglanists have
not felt capable to participate due to a lack of current knowledge.

There have been several updates and partial summaries since 1979, but
many have not reached the entire community. The following summarizes
these, which are frequently referred to in Loglan discussions. Many are
still available from the Institute, and if you ordered or expected any
and did not receive them (except for MacTeach and NB3) , please contact
the Institute. Your balance may be too small, or your order may have been
lost. I'm suggesting to the Institute that any new orders for old
material be billed at a higher rate - say 10 cents per page - to cover
inflation and to conform to the new policy of supporting the Institute
with product sales, rather than just paying costs.

TL 4/2 was intended as a supplement to update L1, but it was written at a
much more technical level, and generally stated the issues more than
resolved them. Notebook 1 (NB1) documented the machine validated grammar
(MacGram), as it currently existed. Notebook 2 (NB2) documented changes
to the morphology, primarily in complex-making, as a result of a research
effort known as the Great Morphological Revolution (GMR). Both NB1 and
NB2 were apparently obsolete within a month after each appeared, but too
little of the problems in each and their resolution appeared in LN, the
only active publication at that time. The last 2 issues of TL were TL6/1
and TL7/1, which updated the morphology and grammar, respectively. On the
whole, each is reasonably accurate, though there are significant changes
to each, and TL7/1 does not contain the complete grammar as did NB1, in
the detail needed to actively work with the language. There have been
abortive or sluggish attempts to update other material, including the
Primer, L1, and the L4/L5 dictionary, but none has completed. jcb is
currently working on a current state-of-the-language document (NB3),
which may resolve this problem, and Glen Haydon and jcb have worked off-
and-on to complete an automated flashcard program called MacTeach. Other
efforts, such as my dictionary work are now rapidly accelerating in light
of the potential for NB3 to bring everyone back up to date, and the
response to my telephone calls to start this SIG have indicated an
underlying enthusiasm that should allow many of the rest of the ongoing
efforts to be completed within a few months after NB3 becomes generally
available, thus allowing the Loglanist masses to help out again.

2. The Status of the Loglan Project. The following is as complete and
accurate as I'm able to determine. I'll keep everyone up to date with any
new information I acquire.

Going Public Again (GPA) - Our goal. This has evolved over the years as
participation has risen and fallen. The following is a summary of
projects that people have included in the effort in previous
descriptions, and a short status. I'll then go into detail on those which
are still (to my knowledge) active.

a. L4/L5 Dictionary Update - This has cycled between being a
reformatting, and a true update. Time and participation will determine
whether the update will consist only of correcting typos, updating
existent word changes, and adding approved new words, or a more major
revision that might include redefinitions and even new changes (such as
argument structure) to old words. The time-frame is also negotiable, but
I think a good product can be out in a year with preliminary drafts
available in 6 to 8 months for workers.

b. L1 Update - considered by many to be the most essential element, this
is on hold and may not be done given jcb's other priorities. However:

c. NB3 - this is to be jcb's statement of the language as it is. 40 odd
pages are done, and if he does the whole language as he has covered the
portion so far, the L1 update may be superfluous. NB3 is apparently what
used to be called L6, but seems to be going beyond at least what I
expected. Probably out by the end of the year.

d. MacTeach - this is a series of 8 or so flashcard programs for
computers, along with cassette tapes to allow the user to hear what da is
working on. Glen Haydon and jcb have worked off and on for a few years,
and have settled what is wanted. I have an early version of the first
program, and it seems useful. SIG members will be helping finish the
programs, which may be in FORTH as currently, or rewritten in other
languages. Machine transportability, and program size are problems,
especially for CP/M machines. A problem exists in timing responses, since
various machines have different clock speeds, or means of determining
time. The first program will probably be out within a month after jcb's
return. Depending on our approach, the rest could take from a couple of
months to a year to complete.

e. LYCES/MacGram - The computer verification of the language's
unambiguity. Scott Layson has updated this to run on an MS-DOS machine.
With the larger memory, the whole language will again fit. jcb apparently
has solutions to three old problems to be confirmed and the work will be
done.

f. Loglan Interactive Parser (LIP) - This is a subset of LYCES that
verifies whether a given Loglan statement is grammatical, and how it
should be grammatically interpreted. This is the test for an English-to-
Loglan translation, or for an original Loglan composition, and a possible
basis for a Loglan-to-English translator. Until we have significant
experience writing/speaking the language, this tool will be vital. I've
been trying to reach Scott Layson to determine the status; we have people
willing to work on the update. One known change besides updating the
grammar, will be the capability to parse larger chunks of text.

g. Trial N Grammar - Related to LIP is a simple definition of the current
grammar. NB1 had the last published version. TL7/2, which never came out,
was to have the final grammar. I'm trying to get the current grammar so
that people have a good tool to supplement the TL7/1 Teaching Corpus. In
any case, NB3 should include this with the last of the LYCES-verified
changes if I can't get it out sooner.

jcb has asked for volunteers to write a short description of the grammar.
Such a description is needed for the dictionary, as described in my
Lognet submission. With no current formal definition of the grammar, I've
heard reluctance to try to use the language. (TL7/1 does not say what has
been changed since NB1) Any volunteers, or is someone already working on
it?

h. SA Translation Project - as a test of the language, jcb has translated
portions of Scientific American articles from 8 languages into Loglan.
This has added many new words, some new letters to the alphabet, and in
general has proven the language. jcb intends to include excerpted
translations in NB3.

i. Other translations - there is little in the way of other translation
in progress. A couple of linguists have agreed that the SA effort should
be confirmed with some Bible (and perhaps Shakespeare) translations.
Literary use of Loglan will be verified, and The Bible, regardless of
one's religious beliefs, is the most linguistically studied text in
history. The current scientific bent of translations will also weight
down the dictionary with specialized scientific terminology.

j. Primer - a written grammar of the language. MacTeach is only useful to
those with computers. This would be an expansion and completion of the
original primer that is now found in TL2. Chuck Barton was working on
this, and may be interested in completing it when details of the current
grammar are available to him.

k. Other computer aids - Nora Tansky had a working flashcard program a
few years ago, Which should be compared with the current MacTeach. Anita
Lees also was developing such a program, but I have no status on it. Nora
also had a random Loglan sentence generator and a Loglan-to-pidgin-
English translator. These might be combined with LIP and the MacTeach
framework to complete the MacTeach set, or may be developed separately.
Nora also has programmed an appendage to her machine to 'speak' Loglan.
She says it has a few problems, and the device doesn't fully support the
Loglan phoneme set. Perhaps newer devices are more powerful, and a tool
can therefore be devised that will speak text while it is being worked
on, say in conjunction with MacTeach or LIP.

l. Eaton Interface/Shakedown Cruise - originally a test of GMR; this has
become a cornerstone of GPA as the primary organized Loglan work being
worked on outside of Gainesville the past few years. jcb wants good
coverage of the most used concepts in the natural languages, and Eaton
has been the standard reference in this field. Kieran Carroll has
patiently coordinated this effort, but volunteers have drifted away and
progress is slow. With his new job and an impending marriage, Kieran's
effort faces continued slow progress unless a new infusion rescues it.

m. Complexing - my proposed solution to the Eaton project, which has
stemmed from a need to verify the complex-making algorithm. I'll describe
this below, but it will closely tie the Eaton work to the dictionary, and
generate lots of words. I'm hoping to involve all of you in this.

n. Universals - Chuck Barton has told me that there are new theories of
linguistic universals that may be supplanting Eaton as the measure of
language commonality and coverage. Someone needs to look into this
further.

o. Yngve - jcb has requested people to look into this researcher's work.
I don't know what it is about, but it may have to do with the adequacy of
the Corpus as a test of the grammar.

p. Publications - Loglan has seen little general print over the years
since the June 1960 SA article. There have been other articles published
which I'll list below. With GPA, we need to get more national attention.
Birrell Walsh has offered to write a piece for the publishers of the
Whole Earth catalog, who are interested.

q. Funding - jcb is going to seek funding for his graduate assistant to
support the Gainesville Institute work. This will give some stability to
the Institute and its finances.

Apparently, though, a lot of linguistics work is being funded now through
the Defense Department. There are applications for Loglan in artificial
intelligence(AI) and in international communications systems, which could
bring DOD funding. But there is the obvious need to prevent possible
classification of any funded research, and the DOD is not the most
popular of institutions. But some people I've talked in the DOD
environment say that the time has come for a Loglan-like language to be
used by computers in AI-aided communication and translation, and the DOD
may develop one of its own within a few years. The side-effects of such
funding could ensure Loglan's acceptance, especially for those interested
in its use as an international language or in AI, as well as giving the
Institute a steady income to fulfill the language's potential. A
controversial topic - opinions are welcome.

If I've forgotten your project, let me know, and I'll include it next
issue.

3. More on L1 update and NB3

As I said above, NB3 could completely replace L1 if jcb continues writing
at the level, and with the completeness and clarity he seems to be using
so far. But if he is not going to cover the whole language at this level
in NB3, an update to L1 should be a GPA priority. I'm hoping that a
massive show of support may result in both.

There was talk at one time that instead of rewriting L1, that jcb would
paste up a copy into a notebook and annotate it with updates and
corrections. I should think that this would be comparatively easy after
NB3 is done, especially if someone else volunteers to paste up the master
for him. After he marks it up, we can set up a team to get it typed into
a machine. He could choose to just outline the changes he wants and
cross-reference to NB3, and the team could flesh out the text, subject to
his approval when it's done. This minimizes jcb's time spent on this high
priority project that would interfere with more technical tasks that
require greater time investment.

I personally would like the revised L1, or NB3 to mention all little
words(LWs), if only in an Appendix, with an example or two of the usage
of each. That and an index would be the most significant improvements to
L1 besides correcting for changes in the language.

4. More on the Dictionary work

I am including in the Appendices a copy of my lengthy submission to LN
last year which includes several notes and suggested plans for the
dictionary update. I've updated these to reflect the current status as
much as possible, but I tried to retain my original flavor, so I could
not cover everything current about the dictionary. The following are
items about the status, and where you can help.

a. New Format - I have converted the raw L4/L5 keypunched card image data
to a delimited field format. The format and a sample of the text can be
found in the LN submission Appendix 2. I believe my conversion is better
than 95% accurate, which will minimize the need for manual massaging. For
practical reasons, I may have to do that massaging myself, or use
volunteers with directly compatible computers who can give short
turnaround. The data is voluminous! As noted in the LN text, the comma-
delimited format(see Appendix 1) given in LN for wordmakers is useful for
their short-form lists. But the longer Universal format is needed to
support the dictionary work, since much of the data - especially English
Translation(E-trans) data doesn't fit in the comma delimited format. I am
asking for volunteers to write programs to convert between the two
formats, and from Universal format to dictionary text. (The Dictionary
Reformatting Manual is needed for the latter).

b. Remade Primitives/Etymologies - I have replaced all of the old
primitives with the new primitives. This list is substantially identical
with that in TL6/1 and NB2. I have also updated the primitive
etymologies. These are the same as in NB2. Three of the etymologies had
to be changed. Appendix 3 lists corrections to the NB2 primitives,
affixes, and etymologies, so as to bring everyone current. I have a
complete set of primitive etymologies on disk, if any linguists want to
check the data vs. their own dictionaries/knowledge. I wouldn't doubt
that there are a few typos in there, as well as possible phonological
errors - and I have no knowledge of most of the languages in question. I
will try to get a list of the dictionaries jcb uses for his etymology
work, since standardization of our references is highly desirable.

c. Word Types - Dictionary data has been sorted by word type to allow
groups of people to work on specialized sections of the text. With the
primitive changes and the upcoming remaking of the complexes, the
dictionary can no longer be maintained in sorted order. However the
Universal format was designed to enable most sort programs to be able to
produce a sorted listing easily, although it will be time-consuming. The
current text takes over 6 Meg on my disk. I expect it to double (or more)
before it is done.

d. Priority Efforts - My commitments to Jim means that the highest
proportion of time will be spent on complex regeneration. He has also
approved of my coordinating an analysis of the little words. Some must be
remade because the primitives they were derived from changed. And all the
new little words must be written up in Universal format, preferably with
many examples for the E-trans section. A third effort is the
reconstruction of the Element Names and primitives. The final rules on
borrowings made the NB2 list obsolete, and the new alphabet makes it
possible to make the names more phonologically correct. With help from
jcb and Rebecca Bach (a new recruit), I have a new list which I'll try to
have ready for kibbitzing within the month. Any kibbitzers out there.

e. Old New Words - A most important task to be done with the dictionary
is to comb the back issues of TL and LN for little words, new primitives,
metaphors, etc. that need to be added to the dictionary. I believe pc,
Chuck Barton, and Colin Fine have notes on these topics, but I don't know
how complete they are. And they must be converted to Universal format.
I'm hoping to get one of them to volunteer to collect all of the cards
and notes from everyone and preferably to write (or better word process)
them in Universal format. We can then generate word lists for others to
review in whatever format is found most useful. All proposals should be
so included, but the collector will inherently have primary kibbitzing
rights as to which should become 'official'. The final decisions of
course, are up to the Academy.

f. New New Words - A similar task is to collect all of the new words that
jcb has collected from the translation work, Eaton work, and words of
opportunity, and put them in Universal format to be merged with the
others. Some kibbitzing has been done on the Eaton words. But the
conversion is still necessary to get them into the dictionary.

g. Complex-Making Algorithm - There is now a formal algorithm for complex
making. It is in 3 parts, and I'll try to include all of them in the
Appendices. Part 1 is Kieran Carroll's algorithm for choosing the best
metaphor for an English word. Parts 2 and 3 are the integrated generation
of the Loglan complexes from the metaphor affixes.  First a set of all
legal complexes for a metaphor must be made (Part 2), and then the best
must be selected (Part 3). There are actually two proposed Part 3
algorithms so far, and discussion indicates that a new one will
eventually be necessary due to the 'untastiness' of certain consonant
pairs, or interactions between affixes.

A computer program must be written to automate Parts 2 and 3. I've had a
couple of volunteers, but the project is important enough that I want
several versions. I will try to distribute the best to all who do word-
making. Word-makers need a user-friendly interface. The dictionary effort
needs a version which can take the various Universal format inputs, and
either print the best choice, all choices, and/or substitute a selected
choice back into the Universal format dictionary where appropriate. A
sophisticated user interface is needed here.

h.  Complexing - As part of evaluating various Part 3 algorithms, I came
up with an idea which may solve Eaton interface problems while helping
verify the algorithms. While in Gainesville, the two versions were
developed by generating random test case complexes using convenient
affixes. jcb agreed that algorithms should be tested like the old taste
tests - by generating complexes, and letting human evaluators verify the
computer choices.

The logical extension of this is to take the set of primitives, and
generate 'metaphor pairs' to test the algorithm. This primitive set
exists separately from the dictionary (as part of the MacTeach
development effort), in a form most suitable for a computer program to
play metaphor generator. A variety of the complex building program could
then take the output of that program and generate taste tests for any
given Part 3 algorithm.

I then realized that the intermediate product is even more useful. We
have the capability, by computer, to generate all of the n-part metaphors
possible with the set of primitives. There are, of course, millions. But
if a systematic approach to selecting which n-combinations were most
productive of useful metaphors, we could easily generate lots of new
words. The affix set provides this approach - the most useful words in
metaphors have the largest affix sets.

A computer program which generates the 2-part metaphors possible with the
affix set would generate tens of thousands of possible metaphor words.
Human reviewers can then go down lists, and determine which, if any,
bring to mind useful English (or other languages) translations/concepts.
Da would then write these down. The results of all reviews would be
sorted by E-trans, giving cases of English words with multiple metaphors,
of which the best can be chosen.

Since a similar data set for argument structures of primitives is needed
for MacTeach, a similar program could apply quantifier and conversion
little word affixes to primitives to generate all of the other kind of 2-
part complex metaphors.

An interesting thing about this approach is that it exactly models the
process by which a Loglan listener will recognize a new metaphor. Hearing
a complex, da will decomposes it into the metaphorical roots, and try to
determine the concept that is intended. And because we are attempting to
model the process, experienced list workers should be able to analyze
prospective metaphors at several per minute. We should literally be able
to produce thousands of words if a significant percentage of metaphors
are useful. (I suspect they won't - if even 10% are useful, the 900 odd
primitives and thus 800000+ 2-part metaphors would produce a staggering
80000 new words). As such, 1000 metaphor-blocks should be able to be
analyzed in an hour or two, at most. With a discriminator to reduce the
volume of useless algorithm produced by the computer, the effort, while
time consuming, could be completed within a few months, if everyone gets
involved.

The other interesting thing about this approach is that it will produce a
uniquely Loglan set of common concepts. With Zipfean assumptions, the 2-
part metaphors should represent the most common and useful words in
Loglan. We will end up with a complete set. I have no doubt that, when
applied against the Eaton list, this set will cover most Eaton concepts.
The remaining can be analyzed for patterns that might suggest additional
primitives and affix requirements, but the job will be much smaller, and
it may be superfluous since the word set will most likely exceed the
goals of the original Eaton project in coverage.

We obviously need some programs as described above, and we need someone
to generate the raw argument list for the primitives. The latter will be
useful for MacTeach in any case.

This idea was developed after jcb left for Europe. I would like to have
useful (and impressive) results for him when he returns. A coordinator
other than myself may be required eventually, since this work is
voluminous enough to conflict with the dictionary effort. But it will be
fun, perhaps more fun than the Eaton effort, which some have told me was
often slow, tedious, and unproductive. And working on this will help us
all learn the language and contribute, without having to have
dictionaries, grammars, or a knowledge of computers or linguistics. I
intend to involve everyone in this if it looks like it will work, and
there are at least 90 Loglanists out there still active according to jcb.
Watch out world!

5. Changes to the Language

I won't try to describe all of these. NB3 does so. But since the last
reporting in LN, some things of note have occurred.

Hyphens - My proposal for the use of y and iy as hyphens, described in
the last LN, was modified and merged with the TL6/1 comments on the same
subject, and accepted into the language. The details of hyphenation are
in the complex-making algorithm - see Appendix 4. In general, when
hyphenation is required, use y (schwa) after all affixes except
CVV/Cvv's, use r in the latter, unless the first letter in the next affix
is r, in which case use n. The latter preserves the mandatory consonant
pair in all cases, and is easier to say. The n is necessary to preserve
permissible medials. If a buffered dialect (used when a speaker cannot
manage specific consonant pairs), then use y as the buffer, and iy (yuh),
as the hyphen.

Tosmabru - The *Tosmabru test, as reported in TL6/1 was erroneously
stated, and the example of a word that should pass should instead have
been detected to decompose. The complex-making algorithm in Appendix 4
has the correct test.

Alphabet - jcb has devised a scheme for representing Linnean binomials
(the formal names for plants and animals). Partly from this and the SA
translations, he determined a need for additional coverage of phonemes in
the alphabet to replicate names and binomials. Thus:

q         is the 'th' in 'theta', the greek letter
x         is the hard ch in 'Bach'
w         is the u umlaut in the German name for Munich

Thus the latter is now la Mwnxen, Bach is la Bax and Khruschev is now la
Xrustcev.

These letters are never found in 'true' Loglan words, except in the names
for themselves as use in acronyms. They specifically cannot be used in
either predicates, borrowings, or little words.

Cases - jcb has devised an approach to representing cases in Loglan. All
of the possible arguments have been classed into 13 types.  The details
have not yet been described, and the LWs have not been assigned. LN48.9
described the basis for the case system, which would expand the lexeme
set consisting of the pua family. A set of 13 words, similar to sau would
be used as 'case-tags', in the same way that pua words number arguments.
Sau for example could be redefined to cover 'sources, authors, or
starters' since it now means 'from source/origin'. He suggests dii for
'destinations or receivers', and mou for 'moveables or transmissibles'.
The labels would make learning arguments easier, and eliminate problems
caused by the sometimes arbitrary argument order. Perhaps they could even
be used to solve the problem of how to relate an argument not normally
part of the predicate definition, by providing a clue as to how that
argument relates to the predicate. The problem of reordering arguments
for style and emphasis would obviously be made easier.

The negatives include the possible standardization of these words in all
utterances, which would be redundant and inelegant. They are, in effect,
like prepositions in English, which are not always necessary, but
required. The English 'I go store' is clear to the listener, but not
grammatical. (These would be optional, of course, in Loglan; if the
speaker wishes to use either the conversion operators nu, fu, ju, or the
numerical argument pua series labels, they will cover all usages.)

I am personally undecided on cases. There might be a tendency for people
to retain more of their old cultural thought patterns when using Loglan,
which could reduce Whorfian effects. And they add more semantic content
to LWs, which as lexemes have been almost entirely grammatical. And they
seem to eliminate the simple elegance of the predicate as a universal
word type with all of the rest of the utterance implicitly related to it.
On the other hand, cultures such as German which have strong case
structures will find the language easier to use. English is relatively
case-free, and we don't have a good feel for how other cultures will find
the language. And the change is potentially major in its usage
implications - too little time remains before GPA to thoroughly study
those implications.

Commas - the written form of the language uses written commas in multiple
ways. They represent pauses for emphasis, and close-commas embedded in
vowel strings can be used to alter the default left-pairing of vowels, as
described in TL6/1.

Borrowings - for those without TL6/1, borrowings are no longer required
to be 1 mod 3 in length. The algorithm for defining the borrowing space
is non-trivial. See my comments in the LN submission Appendix, as an
alternate or supplementary proposal to the current system.

Optional Vowel Disyllables - the iV, and uV vowel pairs can all be
pronounced monosyllabically or disyllabically, depending on
circumstances. The circumstances have been stated narrowly (only after
vocalic consonants), and I personally prefer less restriction, since I've
found some possible complexes almost un-mouthable monosyllabically, while
others are easy. Try mreduo, glopuo, and polpuo as examples that I find
difficult to say without tongue-twisting, or splitting the disyllable.
(Incidentally, similar tongue twisting effects are why I believe some
changes will be be needed to the Part 3 complex-building algorithm. Too
many vocalics or monosyllables, and certain of the permissible medials
seem very untasty compared to hyphenated forms.

6. Material I Have and Policy

jcb has given or sold me copies of nearly everything the Institute has
put out over the years, including a complete set of TLs and LNs, and
several things such as the Dictionary Reformatting Manual, and advance
copies of NB3 text and MacTeach. He has given me no clear direction (and
the Institute has never published to my knowledge) about any restrictions
on the reproduction of this material for Institute business, such as the
dictionary work, or his requested assistance in reviewing NB3 and
completing MacTeach. Since the Institute sells much of this material, I
do not feel at liberty to haphazardly reproduce it. Yet some things, like
the new primitive/affix list, are needed by anyone who wishes to work
actively on the language.

Since I wear 2 hats, editing this unofficial newsletter, and serving the
Institute as a task foreman, I'm going to have to keep my hats separate.
Thus, as editor, I will not be printing or reproducing any Institute-
proprietary material, until jcb gets back from Europe and says it's OK.
As foreman, however, It should be OK for me to make copies for those who
need them to work on tasks that I am coordinating. I will list such
material that I have that seems useful in such contexts, and in some
cases, I will send it to you directly. If something I list seems like it
will be useful in your Institute-related work, please request it and I'll
oblige. This policy may, of course, be amended when jcb returns, if he so
directs.

7. Bibliography

Things of Interest to Loglan-workers

There is an article on Loglan in Dr. Dobbs' Journal, by Dave Cortesi, in
the September 1982 issue. I do not yet have a copy.

jcb and Bill Greenhood wrote an article in Cultural Futures Research, a
quarterly publication of a consortium including the American
Anthropological Association. It takes up the entire issue of the Winter
1983/84 edition, and is entitled Paternity, Jokes, and Song: A Possible
Evolutionary Scenario for the Origin of Mind and Language. The article is
intriguing, is partially derived from what has been learned in developing
Loglan. Since the Institute did not copyright it, I can't make copies.
The Institute does have a set of reprints available at the issue price of
$7.50. If you can't find it in your local university library, and don't
want to wait for jcb's return, I have an address for the journal editor.
A related article appeared in TL6/1. jcb is presenting material on the
subject at Oxford in England this August, for any of you in the U.K who
he didn't tell. I don't know the exact dates.

jcb has been working on a paper with Scott Layson and several others on
the machine grammar for the Communications of the ACM. I don't know if it
has been published yet.

I have A Manual for Dictionary Reformatting, 17 pages, for those who work
with me on the dictionary.

I have a complete list of primitive etymologies post-NB2, for linguists
who would like to verify them. (Any volunteers?)

The following are references to LN articles I think are particularly
useful to active workers. If you were not a member and so did not get
them, or have lost your copy, I have a complete set up to the last (#52).

47.6 - jcb's request for a short summary of the current grammar.
48.9 - details of the 3 post-TL7 grammar changes. I believe these are the
ones as yet un-YACCed. One relates to a new use of ga to mark untensed
predicates. The 2nd is the case system referred to above, and the 3rd
relates to a way to collapse jia and jio into jo.
50.10 - the current policy on when and how to borrow scientific words.
50.15 - more on cases
51.7 and 52.30 - jeff taylor's question and jcb's answer on terminating
descriptions.
52.11 - jcb makes clear that old hands who may have disappeared during
rough times are welcome to come back and work on the language. (James
Carter, pc, Bob McIvor, and Chuck Barton were specifically mentioned as
having volunteered to help on Eaton).
52.26 - jeff taylor and jcb, a letter in 'New Loglan', with jcb's
analysis and translation. The detailed analysis is a very helpful
analysis of grammatical problems. If there is any problem, it is that the
number of comments is so large, that it causes me to doubt I'll ever
master the grammar. But those who are more familiar with it may reassure
the rest of us by producing more such material and translations.

Most everyone has TL6/1, which is the latest summary of GMR, and TL7/1,
which has the Teaching Corpus and is devoted to grammar. If you don't,
the last Institute price was $3.25 for TL6/1 and $2.95 for TL7/1. It's
hard to work without them.

Incidentally, while there are stacks of old dictionaries, there are
effectively no remaining copies of either L1 or the Supplement (TL4/3).

Notebook #1 ($10.00 at last word) has the last complete corpus, and the
formal grammar. TL7/1 had the updated Teaching Corpus, but this is
apparently a subset of the full Corpus, is missing all the background and
many of those little words I never have really mastered. Some have found
TL7/1 to be very useful; its inductive approach models natural language
learning methods. But inductive methods do not give confidence that ba
correctly understands the nitty gritty rules behind ba's learnings. I
hope NB3's summary of the grammar includes the formal description,
definitions of terms, a complete list of little words, and explanatory
material like NB1, as well as the excellent teaching style of TL7/1.

Notebook #2 (also $10.00) describes the process used in remaking
primitives, devising affixes, and otherwise completing GMR. There are
several useful tables and lists, but most are outdated. The updated list
of primitives and affixes in the Appendix to this newsletter will be a
better reference, especially when coupled with the Complex-Making
algorithm.

Notebook #3 (to be published this fall - estimated $50.00 at 15
cents/page) will be the most useful recent work on the language, and may
supersede L1. I have the first 40-odd pages for review, and may ask for
help in areas such as the phonology.

MacTeach (current deposit $50.00 - price will be a multiple of production
and mailing costs to include small royalties to the authors) The first of
these programs should be out by fall, and is an updated teaching/oriented
flashcard manager. About 7 or 8 others, covering the whole language in
ascending difficulty, are expected - though I have some ideas on how to
extend the series using LIP to include translation training.
As currently advertised, this includes permanent free update rights, but
I think the Institute is risking something in making such a promise
without requiring a shipping and handling fee for such updates. As
planned, each ordered list in the series will come with a cassette tape
so that you can hear the words/sentences as you work on them when you
work in sequence order.

LIP (last listed at $60.00) This also allowed unlimited update rights,
and since the current version needs update, will test that policy. (I
think we should all be willing to contribute to costs given the
Institute's financial state). jcb asked Scott for a version that could
digest larger portions of text, as well as include new grammar changes.
This may require larger machines than those that run CP/M. LYCES was
rewritten to use the larger memory of jcb's new Z-100 under MS-DOS.

9. What Can You Do

Nearly everyone I've talked to has been interested in working. Some have
been concerned about time commitments; some want to meet with other
Loglanists, while others would prefer to work alone. Some feel relatively
familiar with the new changes, others are completely lost and will be
until a new L1 or similar lay-oriented text appears. Some are computer-
proficient, others are 'users', and others don't have (or want to have) a
computer. Some have remained active over the years; others have been
recently laying low; others have been TLers, or even non-subscribers who
do not know what has been changing in the last 5 years.

                  YOU ALL HAVE THE CAPABILITY TO HELP!

We have tasks, like complexing, that don't take a lot of time or
knowledge of Loglan, nor do they require a computer. But lots of you are
needed. The original Eaton work, while slower in apparent progress, is
equally undemanding of time.

We have materials necessary to learn what is needed to do some of the
more technical tasks.

If you want to help on any of the computer-related efforts, there are
more than enough programs for everyone to write. Let me know.

And if you are uncertain whether you can do a particular job in the time
requested, just say so. Neither the Institute nor myself are slave-
drivers. There will be coordinators for each of the tasks, and we'll
repartition the work as necessary to keep everyone busy and get the
language done.

And them everyone can get on with the fun of learning and using our new
language.


List of Appendices/Attachments

1. Comma-Delimited Format
2. rjl submission to LN
3. Affix and etymology changes since NB2
4. Complex-Making Algorithm, Parts 2 and 3
5. Current List of Prims and Affixes
6. Complex-Making Algorithm. Part 1 (if received from Kieran Carroll
prior to publication.
7. Manual of Dictionary Reformatting - This 17 page document is not going
out to everyone, but only to those who have volunteered for dictionary
work. Since it was written by jcb, I am not yet releasing it generally.
Anyone who thinks that they need it may have a copy. See #7 for my policy
on giving out copies.

                   Appendix 1 - Comma-Delimited Format

When to use which formats:

1. Universal format is preferred for direct dictionary work, since it
contains everything necessary for building dictionary entries.

2. Comma-delimited format is preferred by jcb and Kieran Carroll for
Eaton work, and is OK for dictionary work if you want to save time - we
can fill in the other data later.

3. For complexing, it is sufficient to just return word-lists with the
English word (or phrase) suggested by the metaphor written alongside.
Cross out metaphors with no obvious meanings, so we know you looked at
them. If a list of taste-test trial words for a metaphor is given, circle
the best, and/or put a number by them to indicate an order of preference.
Simple!

All fields separated by commas. If you need a comma in the text, use a
semi-colon instead. Don't use semi-colons (they'll be changed to commas).
If you do not have data for a given field, just put the commas together
with nothing in between. If you don't know what to put in a field, put a
question mark.

Field     Contents

1         5 digit Eaton number (3 digit page and 2 digit number
               indicating order on the page)
2         English keyword - 1 word to help identify the E-trans (and
               sort)
3         Loglan word/phrase
4         Etymology - if any. Use the format in the old dictionaries for
               primitives. If a complex, give the complete metaphor.
5         A full English translation of the word
6         Any Affixes for primitives, otherwise null
7         Derivative Complexes, if any (give the metaphor, we'll build
               the word)
8         Morphological Type - possible values are the same as for Line
               Type 1 in Universal format
9         Author's Initials
10        Year the word is invented
11        Eaton Rank in 1000's.

If submitted on computer, add notes with an ** in the first column of the
first line, and at the end, so any computer processors don't try to
figure it out. Notes could include references for etymologies, related E-
trans, such as listed in the lines below each main entry in L4/L5, etc.
Anything that will aid the conversion to Universal and then Dictionary
format, or will explain your logic to any kibbitzers.
    Appendix 3 - Changes to the NB2 Primitive/Affix lists (pg 37-42)
                         Changes to Etymologies

(These can also be used to mark up the list in TL6/1, though some are
already there.)

Old       New       Affixes        English keyword

blabo     bulbi     bui bul        bulb
brani     brona                    brown
bulju               buj            boil
carta     curtu                    shirt
cidjo     cibra                    bridge
citre               cie            thread
dampu     pudja                    thumb
dertu               deu der        dirt
detri     detra     dea det        daughter
durzo               dru dur duo    do
dutci               dut            doubt
flofu               -              float
folma               flo            full
forma               fom foa        form
gotri               got            industry
gotso     godzi     god goz goi    go
gusta     tasgu                    disgusted
kampe     kambi                    compare
klira     kalra                    collar
krena     kurti                    curtain
madji               maj            magician
marke     marte     mae            market
matca     metca                    match
matci               mai mac        machine
metro               meo            meter
pento     penta     pea pet        point
petri               -              distribute
pidra     hompi     hom hoi        drink
pirle               pie            parallel
rorno     horno                    horn
selba     helba     hel hea        help
slano               sla            slow
tatro               tat            theater
virsa               vir            poetry
virta               vit            ad

godzi  3v x goes to y from z. 2v x goes to y (from here).
godzi  -b 2/2e go 2/4c dzou 2/4R idti

marte  2n x is a market of town/district y, a place for trading
marte  -b 4/4e mart 4/5g markt 3/4f marche 4/6j maketto 3/7s
mercado 2/5h bazar 2/5R bazar

sluko  2n x is a/the lock of/on/in y. 1n lock
sluko  -b 2/3c suo 2/3e lock 2/5G schloss



[1993 note - the following is obsolete, but is the version as written
by JCB and Lojbab in May-June 1986.]

                        CPX - Making Algorithm

     The following is the algorithm for generating Loglan complexes
that is being implemented as part of the dictionary reformatting
project.

Given an n-term metaphor P1 P2 ... Pn and the (unflagged)* instruction
to find the best** reduced word:

1) Look up or generate all of the affixes (3- and 4- letter forms) of
P1, P2, ... Pn-1, forming the sets {A1}, {A2}, ... {An-1}.

2) Look up or generate all of the affixes (3- and 5- letter forms) of
the final term Pn, forming set {An}. Eliminate any CVC affixes from
this set.

3) Form all of the combinations, without hyphens, a1I a2J ... anK,
where a1I  {A1}, a2J  {A2}, etc. (There will be N{A1} x N{A2} x ...
x N{An} such combinations. Form them all.)

4) Install hyphens where necessary in any combination. Specifically:

     a) If n>2, put r or n after a1I if it is a Cvv or CVV. If n=2,
     then put r or n after a1I if it is a Cvv or CVV, and if a2J is
     not CCV (e.g. saicli, which requires no hyphen). Use r as the
     hyphen unless the first consonant of the next affix is also r, in
     which case, use n as the hyphen.

     b) Put y at any proscribed C/C joint (e.g. mekykiu). See Section
     10, page 29, of TL6.

     c) Put y at any proscribed C/CC joint (e.g. menydjo). See Section
     11, page 29, of TL6.

     d) Put y after any 4-letter affix form (e.g. mrenysai).

     e) Test all CVC ... CVC + CVCCV forms for "Tosmabru failure", as
     follows. Examine all the C/C joints between the CVC affixes, and
     between the last CVC and the CVCCV term. If the first one or more
     of those C/C joints are "bridged" by permissible initials, listed
     in Section 13, page 30, of TL6, then the trial word will break
     up. But if the first C/C joint is unbridged, i.e., is
     impermissible as an initial CC, the trial word will not break up.
     It has passed the Tosmabru Test. Only the first joint in a trial
     word needs to be unbridged in order to ensure resolvability.
     (Note that this definition of the Tosmabru Test is a change from
     the version in Section 24, page 35, of TL6. ?Gusnilbo'tci can
     resolve to gu snilbo'tci, and the Tosmabru Test must therefore
     detect this failure.) Install y at the first bridged joint if the
     Tosmabru Test fails (e.g. tosymabru).

5) Evaluate all combinations and select the best**.

* Unflagged: It will be possible to specify certain flags to constrain
the algorithm in its evaluation and selection. The currently
identified flags are:

     NR - Non-reduced. The non-reduced (4- or 5- letter affix) form
     will be automatically selected as the best, overriding the
     evaluation.

     RR - Right-reduced. Only the right-most term Pn may reduced using
     a short affix form. All other terms will use the non-reduced
     form.

     LR - Left-reduced. The opposite of RR. All terms except the
     right-most term may be reduced using a short affix form, but Pn
     will use the non-reduced 5-letter form.

     LL - Leftmost-reduced. Only the left-most term P1 may be reduced
     using a short affix form. All other terms will be left unreduced.

     MO - Manual override. When previously processed through this
     algorithm, human review of the evaluation caused the manual
     selection of another form instead of the best scoring form.
     Determine the best form and report it, but indicate the manual
     override condition to allow human re-review. If available, also
     report the 'old' form that was determined by human review. (This
     is intended to cover two cases - those where some constraint on
     the word form is needed that is not covered by the other flags,
     such as 'middle term reduced' or, more likely, a form such as
     long-short-long-short, which reflects the compounding of 2 2-term
     metaphors. The second case is when usage or some non-algorithmic
     determination of 'tastiness' dictates that some form is
     preferable to the algorithmically selected form.) In case of
     manual override, the dictionary reformatting processing will not
     automatically change the entry, thus possibly leaving an improper
     word - if affixes have been changed - until the manual review
     takes place.

** Best: Options will exist in the implementation of the algorithm to
report out any of the following:

     -all wordforms and their scores. If flags constrain the
     algorithm, wordforms excluded by the flags may be optionally
     either excluded or listed separately below the forms permitted by
     the constraints. This option will be used until an evaluation
     algorithm has stabilized.

     -the 5 best wordforms and their scores, again with optional
     separate listing of the 5 best non-constrained wordforms and
     their scores. This option will be used to reduce output for human
     review when a degree of confidence has been established in an
     evaluation algorithm, but while it is still considered desirable
     to allow human 'kibbitzers' verify the evaluation.

     -the one best scoring wordform, and its score. (for dictionary
     reformatting). Optionally, the best non-constrained form and its
     score will also be reported. In case of manual override, the
     previous form and its score, and an indication of manual override
     will be reported as well, and no change will be made to the
     dictionary.

                        Evaluation Algorithms

     Two types of evaluation algorithms are currently being
implemented. Results from each will be produced and compared and
evaluated by the community of word-makers. After tuning, one algorithm
will eventually be used to actual reformat the dictionary and to
compose words from metaphor lists submitted by word makers.

Algorithm 1 (JCB) - This algorithm form is similar to that of the
Taste Test 5 evaluation. (See pp. 20-26 of NB2, especially Table 3.)
Values are attached to individual affix forms, to hyphens, and to
certain specified interactions between affixes. The score is the sum
of these values.
     The advantages include a short mathematically determinate result
where the effect of each component can be clearly seen. A disadvantage
is that the setting of values for each component or interaction is
difficult, given that it is desirable to achieve a particular order of
wordform scores as results of the evaluation. The effect of a small
change in any value can be the remaking of many words. It is thus
difficult to accurately reflect the results of taste tests in this
algorithm form.
     The current set of component/interaction values being implemented
is:

                    AFFIX FORMS

          Non-Finals               Finals

     Cvv- (sai-)    7         -Cvv (-sai)         7.6
     CVV- (veo-)    5.25      -CVV (-veo)         4.4
     CCV- (gra-)    4         -CCV (-gra)         3
     CVC- (men-)    5

     CCVC- (mren-)  -20.5     -CVCCV (-sadji)     -9
     CVCC- (matm-)  -25.5     -CCVCV (-brudi)     -9.5

                    INTERACTIONS

                    y hyphen       -6
                    r/n hyphen     -14
                    C/CC juncture  -1.3

Resolve ties by using the wordform with the minimum number of
consonants.

                    EXAMPLES

men + sai                     mensai
 5  + 7.6                          = 12.6
gra + sai                     grasai
 4  + 7.6                          = 11.6
sai + gra                     saigra
 7  +  3                           = 10
men + gra                     mengra
 5  +  4 - 1.3                     = 7.7
men + y + nai                 menynai
 5  - 6 + 7.6                      = 6.6
sai + r + sai                 sairsai
 7  -14 + 7.6                      = 0.6
gra + sadji                   grasadji            (Left-Reduced)
 4     - 9                         = -5.0
sai + r + sadji               sairsadji           (Left-Reduced)
 7  -14    - 9                     = -16.0
mren + y + sai                mrenysai            (Right-Reduced)
-20.5 - 6 + 7.6                    = -18.9
mren + y + sadji              mrenysadji          (Unreduced)
-20.5 - 6   - 9                    = -35.5

gra + gra + sai               gragrasai
 4  +  4  + 7.6                    = 15.6
men + gra + veo               mengraveo
 5  +  4  + 4.4 - 1.3              = 12.1
sai + r + gra + veo           sairgraveo
 7  - 14 + 4  + 4.4                = 1.4
gra + mren + y + sai          gramrenysai         (only possible by
                                                       Manual
Override)
 4  - 20.5 - 6 + 7.6               = -14.9
gra + mren + y + sadji        gramrenysadji       (Leftmost-Reduced)
 4  - 20.5 - 6    - 9              = -31.5
gra + gra + sadji             gragrasadji         (Left-Reduced)
 4  +  4     - 9                   = -1.0
mren + y + mren + y + sai     mrenymrenysai       (Right-Reduced)
-20.5 - 6 - 20.5 - 6 + 7.6         = -45.4
mren + y + mren + y + sadji   mrenymrenysadji     (Unreduced)
-20.5 - 6 - 20.5 - 6   - 9         = -62

Algorithm 2 (RJL) - This algorithm is the opposite of Algorithm 1.
Algorithm 1 attempts to define the factors governing tastiness of
individual components, and builds a mathematical model composed of
those factors, and then modifies the results minimally by the effects
of interactions between components. This algorithm uses the tastiness
of 2 adjacent interacting components as a gestalt, and does not
directly reveal specific reasons or factors for the tastiness.

     The algorithm assumes the following types of interacting
components are found in complexes:

     Initial/Final (all 2-part complexes)
     Initial/Medial (the beginning of a 3-or-more part complex)
     Medial/Medial (The middle of a 4-or-more part complex)
     Medial/Final (the end of a 3-or-more part complex)

     All legal forms of each interaction are enumerated (there are 37
forms where the 2nd component is medial, and 30 where the 2nd
component is final) and evaluated via taste test. The ordered list is
then scored descending from 37/30 for the tastiest or most desirable
form to 1 for the least desirable. A given wordform is then evaluated
by looking up the component interactions, and summing the scores. For
a 2-complex, no summing is necessary, making this algorithm very
straightforward in reflecting tastiness patterns. For more than 2
terms, there is a question whether middle terms might be improperly
over-weighted over initial and final terms. (e.g., in a 3-term
complex, the sum of the scores for Initial/Medial and Medial/Final
would appear to give double weight to the single Medial term). Only
review of algorithm results will determine if this bias exists.

     It is possible to correct for such biases by weighting the
scores, and by adding in factors similar to those in algorithm 1 to
reflect preferences in initial and final terms. It is also possible to
adjust scores for interactions not reflected in the separate component
scores, such as the interaction between specific consonants or vowels
appearing in adjacent components (e.g. -saicai- might be down-weighted
because of the proximity of s and c causing the Loglan equivalent of
the English tongue twister 'She sells seashells ...').

     The advantages of this approach include the direct reflection of
the choices made by a speaker or auditor in the algorithm. As such,
tuning the algorithm to reflect changed perceptions of the relative
tastiness of words is relatively simple. The disadvantages include the
hiding of the reasons behind the ordering used, and thus the
possibility that hidden factors in the specific words comprising the
taste tests might bias the scores in the algorithm. For example, the
original auditor detected a personal bias against test words
containing mre, presumably because the CC pair is so unfamiliar to an
English-native speaker. Such biases can be corrected when identified,
but this algorithm can be most effective only when many independent
reviewers have confirmed the tastiness ordering used by examining its
results.

     The specific initial algorithm and values are as follows:

     1. Determine the interacting components in the wordform to be
               evaluated.
     2. Obtain the values for each such component.
     3. Apply any appropriate weighting factors.
     4. Sum the values to get a total score.
     5. Apply any corrections for being used initially initial and
               final components.
     6. apply any corrections for special vowel or consonant patterns.
     7. Automatically evaluate all unhyphenated wordforms as better
               than any hyphenated wordforms regardless of score. (or
               apply a negative correction for hyphens as in Algorithm
               1. The correction is different for different numbers of
               terms.)
     8. Resolve ties in favor of the wordform with fewer consonants.

Values for interacting components may be found on the following table:

Interaction                                2-term CPX        3-term
CPX
Pattern                         1st/Final  1st/Medial    Medial/Medial
     Medial/Final

CCV + Cvv                       30         37            30 29
CVC + Cvv                       29         36            28 21
Cvv + CCV                       28         illegal       27 28
Cvv + CVC                       illegal    illegal       26 illegal

CCV + CVV                       27         31            35 22
CVC + CVV                       26         30            32 20
CVV + CCV                       25         illegal       37 27
CVV + CVC                       illegal    illegal       34 illegal

CCV + CVC                       illegal    35            25 illegal
CVC + CVC                       illegal    34            24 illegal
CCV + CCV                       24         33            36 23
CVC + CCV                       23         32            29 26

Cvv + Cvv                       illegal    illegal       22 30
Cvv + CVV                       illegal    illegal       33 24
CVV + Cvv                       illegal    illegal       31 25
CVV + CVV                       illegal    illegal       23 19

CVC + y + Cvv                              22            29 21   18
CVC + y + CVV                              21            28 20   17
CVC + y + CCV                              20            27 19   16
CVC + y + CVC                              illegal       18 18   illeg
al

Cvv + r/n + Cvv                            19            26 illegal
illegal
Cvv + r/n + CVV                            18            25 illegal
illegal
Cvv + r/n + CCV                            illegal       24 illegal
illegal
CVV + r/n + Cvv                            17            23 illegal
illegal
CVV + r/n + CVV                            16            22 illegal
illegal
CVV + r/n + CCV                            illegal       21 illegal
illegal
Cvv + r/n + CVC                            illegal       20 illegal
illegal
CVV + r/n + CVC                            illegal       19 illegal
illegal

Cvv + CVCCV/CCVCV                          illegal       illegal illeg
al   15
CVV + CVCCV/CCVCV                          illegal       illegal illeg
al   14

CCV + CVCCV                                15            illegal illeg
al   13
CCV + CCVCV                                14            illegal illeg
al   12
CVC + CVCCV                                13            illegal illeg
al   11
CVC + CCVCV                                12            illegal illeg
al   10

CCV + CVCC                                 illegal       17 17   illeg
al
CCV + CCVC                                 illegal       16 16   illeg
al
CVC + CVCC                                 illegal       15 15   illeg
al
CVC + CCVC                                 illegal       14 14   illeg
al

CVC + y + CVCC                             illegal       13 13   illeg
al
CVC + y + CCVC                             illegal       12 12   illeg
al

CVC + y + CVCCV                            11            illegal illeg
al   9
CVC + y + CCVCV                            10            illegal illeg
al   8

Cvv + r/n + CVCCV/
            CCVCV    9          illegal    illegal       illegal
CVV + r/n + CVCCV/
            CCVCV    8          illegal    illegal       illegal

Cvv + r/n + CVCC/
            CCVC     illegal    11         11            illegal
CVV + r/n + CVCC/
            CCVC     illegal    10         10            illegal

CCVC + y + Cvv                             7             9  9    7
CCVC + y + CVV                             6             8  8    6
CCVC + y + CCV                             5             7  7    5
CCVC + y + CVC                             illegal       6  6    illeg
al
CVCC + y + Cvv                             4             5  5    4
CVCC + y + CVV                             3             4  4    3
CVCC + y + CCV                             2             3  3    2
CVCC + y + CVC                             illegal       2  2    illeg
al

CCVC/CVCC + y +
     CVCCV/CCVCV                1          illegal       illegal 1
CCVC/CVCC + y +
     CVCC/CCVC                  illegal    1             1  illegal


Variants for 3-or-more Term Complexes

Variant 1 - Multiply values for Medial/Final by 37/30 before summing
for all 3-or-more term complexes. This adjusts for the unequal maximum
value for that lookup. Alternatively, the following table summarizes
the adjustment from the unmultiplied value above. This value may thus
be added to the non-variant algorithm score.

Raw Value Adjust    Raw Value Adjust    Raw Value Adjust

     1    .2             11   2.5            21   4.8
     2    .5             12   2.8            22   5.1
     3    .7             13   3.0            23   5.3
     4    .9             14   3.2            24   5.5
     5    1.2            15   3.5            25   5.8
     6    1.4            16   3.7            26   6.0
     7    1.6            17   3.9            27   6.2
     8    1.9            18   4.2            28   6.5
     9    2.1            19   4.4            29   6.7
     10   2.3            20   4.6            30   7.0


Variant 2 - Apply the Variant 1 correction. Then apply a correction
for the Initial and Final terms as follows. The values are determined
by averaging those values for a given affix-form for Initial position
in the Initial/Medial look-up, and for Final position in the
Medial/Final look-up.

Initial Affix-form       Values                   Adjustment

Cvv + r/n                26+25+24+20+11 /5        21.2
CVV + r/n                23+22+21+19+10 /5        19.0
CCV                      37+35+33+31+17+16 /6     28.2
CVC                      36+34+32+30+15+14 /6     26.8
CVC + y                  29+28+27+18+13+12 /6     21.2
CCVC + y                 9+8+7+6+1 /5             6.2
CVCC + y                 5+4+3+2+1 /5             3.0

Final Affix-form                                  x 37/30

Cvv                      29+21+30+25 /4           32.4
y + Cvv                  7+4 /2                   6.8
CVV                      22+20+24+19 /4           26.2
y + CVV                  6+3 /2                   5.6
CCV                      28+27+23+26 /4           32.1
y + CCV                  5+2 /2                   4.3
CVCCV/CCVCV              15+14+13+12+11+10 /6     15.4
y + CVCCV/CCVCV          9+8+1 /3                 7.4


   EXAMPLES (listed in the same order as Algorithm 1 for Comparison

                                   Basic     Variations
                                   Score     1         2

men + sai                     mensai
                                   = 29
gra + sai                     grasai
                                   = 30
sai + gra                     saigra
                                   = 28
men + gra                     mengra
                                   = 23
men + y + nai                 menynai
                                   = 22(hyph)
sai + r + sai                 sairsai
                                   = 19(hyph)
gra + sadji                   grasadji            (Left-Reduced)
                                   = 15
sai + r + sadji               sairsadji           (Left-Reduced)
                                   = 9(hyph)
mren + y + sai                mrenysai            (Right-Reduced)
                                   = 7(hyph)
mren + y + sadji              mrenysadji          (Unreduced)
                                   = 1(hyph)

(Due to the hyphen rule, grasadji would be preferred over sairsai.)

gra + gra + sai               gragrasai
    33 + 29                                  +6.7      +28.2+32.4
                                   = 62      = 68.7    = 129.3
men + gra + veo               mengraveo
    32 + 22                                  +5.1      +26.8+26.2
                                   = 54      = 59.1    = 102.1
sai + r + gra + veo           sairgraveo
    24   +   22                              +5.1      +21.2+26.2
                                   = 46      = 51.1    = 98.5
gra + mren + y + sai          gramrenysai         (only possible by
                                                       Manual
Override)
    16   +   7                               +1.6      +28.2+6.8
                                   = 23      = 24.6    = 59.6
gra + mren + y + sadji        gramrenysadji       (Leftmost-Reduced)
    16   +   1                               +.2       +28.2+7.4
                                   = 17      = 17.2    = 52.8
gra + gra + sadji             gragrasadji         (Left-Reduced)
    33   +   11                              +2.5      +28.2+15.4
                                   = 44      = 46.5    = 90.1
mren + y + mren + y + sai     mrenymrenysai       (Right-Reduced)
    1    +        7                          +1.6      +6.2+6.8
                                   = 8       = 9.6     = 22.0
mren + y + mren + y + sadji   mrenymrenysadji     (Unreduced)
    1    +        1                          +.2       +6.2+7.4
                                   = 2       = 2.2     = 15.8


The following is material that I wrote last year annd submitted to John
Lees for Lognet. I sent some advance copies to others for comment, and
received enough such comments that I decided to rewrite it before
publication. Unfortunately, Lognet dies before I finished the rewrite. I
have revised the information to be current.


x.1 Net from Bob LeChevalier (RJL), Background

I was an inactive Loglanist starting in 1979, and a member for some 4 or
5 years now. I was JCB's only San Diego resident Loglanist during the
last few years that the institute was located there. I wasn't active
because my practical skills at languages and linguistics were negligible.
However, I was JCB's local sounding board during the early years of GMR.
I thus feel a sense of the history of the project even though I haven't
really contributed much time before the dictionary effort.

I am 32 and single. I have a B.S. in Astrophysics from Michigan State
(which I haven't used since joining the computer field). By trade I am a
computer systems requirements engineer working for Systems Development
Corporation. I am prone to occasional periods of heavy travel and lots of
work deadlines which have made my contributions to Loglan subject to a
lot of interruption.

I am the "dictionary update foreman" originally defined in LN29.22. jcb
has recently redefined this more narrowly to include only those aspects
relating to updating the old data to the new format, and changing that
data to match GMR and MacGram. Part of the reason for this reduced scope
was the lack of progress on the dictionary.

I personally want to serve the broader original dictionary update
function, and to build on work done by Colin Fine, pc, Chuck Barton, and
others in correcting flaws in the dictionary, improving definitions,
adding new words, etc. Jim has not granted me this, but I believe with
your help, we can convince him that the dictionary effort requires this
breadth, and that the team I'm trying to build can handle the effort on a
timely basis. His sabbatical this summer should give us the time we need
to so demonstrate. Of course, pending jcb's decision, this extra work I
am doing is unofficial, and will take lower priority to the tasks he has
specifically approved.

I am interested in the dictionary project since it is a long term effort.
As such, my irregular time availability isn't as noticeable. I am also
especially interested in the interface between Loglan and the 'real
world'. As GPA approaches, I haven't heard a lot of discussion about how
our esoteric research can be communicated to the lay public and to a
generally disinterested or even antagonistic academic community. Faced
with my own difficulties in learning Loglan, I'm sympathetic with those
to whom "affix", "morphology", "predicate", "intervocalic glottal stops",
"Eaton", and "penultimate" are a bunch of meaningless jargon (not to
mention our peculiar algebra of "CCV"s and "CVC"s, etc.). I'm afraid the
technical aspects of our effort will thus tend to turn off those without
linguistic background.

The dictionary will be a major resource to the new Loglanist. My
contributions will hopefully help render it more useful and intelligible
to the lay reader. The dictionary also happens to be an area where my
weakness as a linguist and as a linguisticist(?) will less impede my
contribution, while my particular variety of skills will be especially
useful. Since we have several Loglanists skilled in translation, and thus
familiar with lexicography, my reliance on them should prevent any
problems that might result from my technical weakness in the subject. On
the other hand, I bring skills from my professional life that are
necessary to this effort; specifically: organizational skill, an ability
to coordinate very complex tasks, and a reputation for enthusiasm that
encourages people to be active because they know that progress is being
made.

In volunteering, I have offered my computer as a repository for the new
Loglan, and my time in an attempt to make it intelligible to the non-
Loglanist. I am going to use some basic computer techniques to make the
job easier. I am NOT, in general, planning on doing a lot of programming.
The weaknesses of the last dictionary are principally those aspects that
are most difficult to devise general algorithms for. I spent several
hours trying to write a parser that could interpret 100% of the computer
files used to generate L4/5, and turn them into something easily
manipulable by word-rebuilding algorithms, human text massagers, and
formatters. My results show that it would take too much special case
processing (and programming time) to have such generalized programs at
this time.

Instead, I am automating only a few of the steps, including the final
formatting, and planning on a lot of hand massaging to fill in the gaps.


x.2 Net from RJL, Computer Resources

jcb has occasionally alluded to my computer problems, which luckily no
longer exist. I have a Zenith Z-151 PC-clone with a good supply of
software, including Microsoft WORD, DBASE III, and Turbo-Pascal. I also
have a C-compiler (CI's C86 - which I've never used) and the Zenith
equivalent of BASICA. I have an installed 22 Meg hard disk, of which the
partially reformatted dictionary already uses over 6 Meg.

I have a program called MEDIA-MASTER (by Intersecting Concepts) that
allows me to read most 5-1/4" 40-track disk formats (especially CP/M and
MS-DOS varieties). It is a cheaper version of the type that was mentioned
in Lognet last year. It costs $40 for the PC version and handles some 72
disk formats. I can probably read most CP/M or PC-related diskettes, and
a few others. It does not handle the Apple format, but we have Apples at
work (and also Burroughs B-20 Series workstations). We also have a Z-90
with a 8" high-density drive at work that is pretty much a clone of JCB's
system. I also have a modem (never used) to handle data exchange that
can't be done by diskette. I am not a member of any net, however, and
have no immediately plans to change that. Therefore, diskette is the
preferred means of data transfer unless you can afford long distance
bills. If you send diskettes to me, just let me know what machine/ format
you used.

x.3 Net from RJL, Ideas on The Dictionary Reformatting/Update Task

JCB spent many hours generating a concept for the new dictionary format.
As a result, he generated "A Manual for Dictionary Reformatting" for our
use. The printed dictionary formats that he would like to see have been
published in the March Bulletin (#5). I believe this format, or any
similar format, would require too many complex algorithms to be processed
easily by computer, especially since the dictionary text has embedded
commas.

An early part of the dictionary update task will have to be the
conversion of the bulk of the dictionary to a format that is as general
as possible, and yet human readable to allow for the extensive manual
work that must be done. That portion which cannot be converted
automatically must be performed by hand (or word-processor). Contributors
can assist by generating all dictionary inputs in either of 2 formats:

     a. the "dictionary style" format with all the commas given in
          LN47.5.
     b. my new Universal processing Format described below (x.5).

The latter is preferred, since I will be running the former format
through a parser to create the latter format. Remember, comma-delimiting
means that da can't use commas in da's definition, and that such
definitions cannot contain the complete data that is included in the
Universal format. The dictionary workers must work with both formats
because I don't want to cause a lot of work for those who have been
creating entries with all the commas.

My view of the dictionary task involves taking the reformatted dictionary
and massaging it in several ways:

     a. Substituting new primitives and their derivations into the old
          entries.
     b. Checking all the old metaphors for consistency with the current
          word-building philosophy.
     c. Building new complexes using affixes and the word-building
          algorithm.
     d. Incorporating new metaphors generated by Kieran Carroll's word-
          building team.
     e. Standardizing English translations and formats for maximum
          clarity.
     f. Adding complexes for metaphors suggested by the clarifying and
          standardization process.
     g. Adding selected acronyms, phrases, names, etc. that exemplify the
          use of those aspects of Loglan.
     h. Adding new grammar words and constructs that have changed since
          the last dictionary.
     i. Adding any approved borrowed words.
     j. Generating appendices of various sorts that aid the dictionary
          user.
     k. Generating and massaging a format suitable for typesetting and
          publishing, in both Loglan/English and English/Loglan
          directions.

JCB has assigned me some portions of the reformatting, and I am hoping to
demonstrate my capability to coordinate the rest, as well. If that sounds
like a lot of work, it will be; therefore, the dictionary effort can use
volunteers. I am especially interested in getting some of the members who
have worked with Kieran's team to perform double-duty in smoothing the
interface between our two efforts. The effort also needs someone
comfortable with Loglan grammar (especially the changes), so that it can
be properly documented in both dictionary entries and text. In the months
since JCB appointed me, I have not been contacted by anyone seeking to
contribute, which in turn diminished my enthusiasm for quite a while.


x.4 Net from RJL, The Plan

With the complex task described above, I have analyzed the dictionary
effort and tried to order it in such a way to maximize the use of
everyone's talents. I expect the dictionary to become the critical
obstacle to GPA, since it is one essential that an outsider can't do
without, and it will be the most time consuming of the efforts. JCB has
to coordinate many other tasks, so I want to try to minimize the
requirement for his time in the dictionary work.

I've submitted this plan to jcb, but feel that it should be subject to
all of your comments, as well, because you will be the workers who will
make it possible.

1. Obtain the raw files used in the old dictionary. (done 6/85)

2. Convert the raw files by computer program to the Universal format.
This format must be compatible with the word-builder's format, and must
be sufficient to allow easy human manipulation, sorting, and automated
formatting into both Loglan-English and English-Loglan dictionary
entries. (done 7/86)

3. Massage these files manually to eliminate the special case problems
that I couldn't easily program around.

4. Sort the old words by word type. (done 7/86) Divide lists among
volunteers who will make additions, deletions and corrections that are
word-type dependent. In addition to what is listed below, changes are
necessary to incorporate the material that has appeared over the years in
TL, Lognet, and the notebooks. pc, Chuck Barton, Colin Fine, and possibly
Bob McIvor should have collections of data in addition to that which jcb
has been accumulating.

     a. Names, letters, chemical words, etc. - redo to match the new
standard formats. Generate standardized definitions. Add fields found in
the Universal format but not in the raw data. Some may be remade to
utilize the modified alphabet.

     b. Primitives - automatically substitute remade primitives (done
11/85), and their etymologies (done 5/86). Add affix data, complete Eaton
references, and other fields found in the Universal format but not in the
raw data.
          1. Analyze English definitions. Make them semantically neutral.
          2. Where necessary for semantic neutrality, determine the
multiple English connotations and derive metaphors for them based on the
primitive. (See x.6). Coordinate these with Kieran's word-making team.
          3. Generate additional metaphors common in English phraseology,
but not necessarily in Eaton. In some cases, these will become classes of
borrowed words (See x.7). Coordinate these with Kieran's team.
          4. Analyze the argument structure. Determine if changes are
required. Add English translations to support all argument reorderings
(See x.8).

     c. Complexes
          1. Determine if the metaphor used is still optimal, given the
current word-building philosophy.
          2. Generate new metaphors as applicable, coordinating with
Kieran's team.
          3. Build new complexes according to the word-building
algorithm. (Algorithm written 6/86).
          4. Verify and split out multiple English meanings. Treat these
as b.2 and b.3 above.
          5. Determine an argument structure deducible from the primitive
components. Process arguments as in b.4 above.
          6. Generate standardized English definitions. Semantic
neutrality is less necessary than for primitives, but is a goal,
especially for those complexes included from Eaton. As such, additional
metaphors may be needed as in b.2 above, which will in turn be used to
generate 3 and more place complexes.

          7. (7/86) Partially as a test of the complex building
algorithm, I am now proposing that we use a computer to generate
primitive pairs (and perhaps some triplets) using some or all primitives.
The resulting metaphors may or may not mean anything; human reviewers
must so determine. This process will thus model the way a Loglan listener
will determine the meaning of an unknown metaphor, and thus make them
more learnable (at least for English speakers, since that is the common
denominator of Loglan workers - it seems to me that a Chinese speaker may
be unlikely to recognize many English metaphors, but I'll leave that for
the linguists to argue.) It is my belief that we can take care of most of
Eaton by this method. Remaining Eaton words may show patterns that will
suggest new primitives to complete the effort. In some cases multiple
metaphors may yield the same word. Both could be used, especially if
nuances in the metaphors cause connotation differences, or the kibbitzers
could select just the best. Meanwhile the complex-making algorithm can be
fine-tuned to give the tastiest words representing each metaphor.

     d. Grammar words should be treated as in a. However they will not be
done until step 9. below.

5. Enter all of the above changes into Universal format (manually,
unfortunately - but if people are able to use word processors and/or data
base programs to create either of the two acceptable formats, we can
minimize this step.) By this time the parser to reformat the comma
delimited format into Universal format must be written.

6. Add words generated by Kieran's team. These must be subjected to the
same procedure used in 4. above, so that this may be an iterative
process.

7. Add words generated by the Translation Project and other sources. JCB
has been collecting these up to this point. Also add words that were
submitted and published in Tl and LN, or that others have collected over
the years. Primitives and complexes should be treated as in 4. above.
Borrowings may require special screening and processing (See x.7).

8. Meanwhile, the language description currently found in the beginning
of the dictionary must be updated. This should be written by one person
to ensure consistency of style; someone who knows and has used both the
old grammar and the current one. Others can then review it for clarity.
Hopefully JCB has the rough text from the old dictionary in typist-
readable size. I don't expect that the text is on any computer. Putting
it on computer may require Institute resources; that's a lot of typing.

I believe that this update is more important than many of our other
activities, primarily because it is the shortest description of the
language. Once it is updated, we could conceivably GPA with just a
dictionary. The completed update should be distributed to all active
Loglanists as a short summary of the current language. In the new summary
should be the MacGram Normal Form description of the grammar. (Kieran
Carroll has suggested that a longer term goal should be to create this
language description in Loglan. Then, any Loglan-XXX dictionary would
have both the Loglan description of the language, and its translation in
language XXX. This is probably too much effort for this next edition.)

At this time, the language must stop changing until significantly after
GPA. In computerese, this is called a baseline. When a baseline occurs,
any later changes that are required are saved up, and then distributed
after a period of time with full documentation to all users. If no
baseline occurs, GPA will only put the language where it was in the early
1980's, when no one had complete current information about the language
except jcb, and the active portion of the community started declining
rapidly. On the other hand, the L1/L4/L5 publication established a
baseline which effectively stood until 1979-80. I believe that most
people felt that those books reflected the 'official' language, and that
no change (such as those in TL, which were considered only proposals)
became 'official' until the supplement was published. The new dictionary
and other publications must recognize their 'officialness', and
procedures to periodically update them must be put in place. But in the
meantime, proposals approved by the Academy should be accumulated for the
next 'release' (another computer systems term).

9. Add grammar words and constructs derived from the new language
description. Definitions should be consistent with the wording used in
the description from 8, since both will be in the same book.

10. Add any desired examples of acronyms, names, borrowed words, or other
special words in Universal format.

11. Hopefully, at this point, Kieran's wordmaking team will have
completed it's task. Otherwise the above should continue iteratively
until we can baseline the dictionary word set.

12. An ongoing parallel task will be to drive a set of appendices to the
dictionary. Some of these will stem from the language description,
including the very useful lists of little words found on the front and
back cover. Others will include:
     a. A description of how c-Prims are built, algorithmically, to serve
as a reference in case new ones must be derived. Preferred reference
standards for each source language should be given.
     b. A list of affixes.
     c. A description of the word building algorithm.
     d. A statement of the process for formal approval of new words.
     e. Perhaps an Eaton-ordered list of concepts and the associated
Loglan words.
     f. Useful lists such as colors, numbers, chemical names, plants,
animals, acronyms, musical instruments etc., and algorithms for borrowing
or deriving new words in each class that are not in the lists.

13. Merge and sort the baseline word set.

14. Generate a trial text. This should be partially formatted (single
column, both translation directions). Two teams will be needed for
review. One team will be primarily concerned with proofreading and
ensuring that the Loglan is correct. A second team, hopefully composed of
linguists who have done a lot of translation work (not necessarily in
Loglan) and are thus familiar with bilingual dictionaries. English and
grammar experts can also participate. The latter team will be examining
the text from the lay public and user point of view. Their review is
critical to GPA, and hopefully the reviewers will include some people who
can take a fresh look at the language.

15. By this time, we should have determined the publisher. The final
printed format should be coordinated with printing professionals for the
best 'look'. We also need to generate typesetting commands as part of the
output. Since I have Microsoft WORD, which supports most anything in the
way of fancy typesetting, this should not be too difficult.

16. Incorporate any last minute changes and corrections that have taken
place as a result of the reviews, as well as any changes since the
language was baselined that are absolutely essential.

17. Publish and GPA.

The following areas need immediate volunteers to do this plan (7/86):
a. metaphor builders/ complex reviewers
b. someone to review old TLs, Lognets, and Notebooks for previously
suggested changes to the dictionary, and new words. These must be written
in Universal format.
c. appendix generators
d. someone to rewrite the language summary for the dictionary
e. lots of text reviewers, kibbitzers, and hopefully people who can word
process.

x.5 Net from RJL, Universal Format

The ideal format for word submissions is a multi-line format with the
word and a line or record-type repeated on each line (to make sorting
easy). Each field on each line is separated by a backslash (\).  For each
word, include only the line types that are appropriate, but include all
fields. If computer generated, no line should be longer than 256
characters, and the character positions should be followed.

Line Type 1 - Main Word Line (required for each word)

Position
1:16      Loglan word, left-justified, space filled
17        line type '1,'
19        Word-type :    'v-LW  '
                         'vv-LW '
                         'cv-LW '
                         'cvv-LW'
                         'Cpd'
                         '2-Cpx ', '3-Cpx ', '4-Cpx ', etc.
                         'Phrase'
                         'Name  '
                         'Acron '
                         'I-prim', 'C-prim', 'N-prim', 'S-prim'
                         'Borrow'
          \
          Eaton frequency (n.n) in thousands and tenths
          \
          Creation year with apostrophe (e.g. '75 )
          \
          Eaton Number (page and sequence number on page)
          \
          Author's initials, if applicable
          \
          percent score ('nn%')
          \
          Eaton page and line number, as per comma-delimited format
          \
          affixes, if any, in an order (and with delimiters?) to be
          specified by the complex algorithm programmer.

Line Type 2 - Complex Etymology Lines - required for each complex.

Position
1:16      Loglan word
17        '2,'
19        Loglan primitives comprising metaphor, in order, separated by \
          \
          flags as defined in complex making algorithm

Line Type 3 - Derivation line(s) - required for those primitives with
language derivations. More than 1 line of this type is permitted if
necessary to fit all derivation data.

Position
1:16      Loglan word
17        line type '3,'
19        Derivation text

Line Type 4 - English Definition line - required for all words. It is
preferable to put only one definition/part of speech on a line if more
than one exists. The raw dictionary parser cannot do so and multiple
definitions from the old dictionary will have to be manually separated.

Position
1:16      Loglan word
17        line type '4,'
19        Loglan Part of speech, in parentheses (e.g. '(3n)',
'(2a)','(na)')
          \
          English definition (preferably under 80 characters)

Line Type 5 - English expansion line - at least one required for all
words. These are the raw lines for use in building the English-Loglan
side of the dictionary. The Eaton English key-word will be used from the
comma-delimited format.

Position
1:16      Loglan word
17        line type '5,'
19        English word or phrase
          \
          English Part of speech
          \
          English word definition/explanation
          \
          Loglan translation, including little words
          \
          Loglan word type. If blank, assumed to be that in word 1.

Line Type 6 - Derived Complexes - One line for every derived complex
using this Loglan word. If a 3 or more part complex uses metaphors based
on a shorter complex, the latter should have this line type, as well as
all of the component primitives. (e.g. rojmadsesmao would have type 6
lines for rodja, madzo, and sensi; however, it also might be listed under
sesmao (scientist), rojmao (farmer), and rojmadsensi (agronomy)). This
will aid workers building definition-refining metaphors by showing them
what already has been built. Metaphor makers who can generate Universal
format would assist greatly if they generate these for the components of
the metaphors. The comma-delineated format does not support this. Only 1
metaphor per line.

Position
1:16      Loglan word
17        line type '6,'
19        old complex
          \
          each Loglan word in the metaphor, separated by \
          \
          flags as per complex-making algorithm

A sample from the dictionary raw file, and the resulting Universal format
follows. Note that not all required lines are present, and that minor
parsing errors may exist. I haven't corrected these so that volunteers
can see what typically has to be done.

Raw data:

   -                                  -
cupli  2a is more coppery/copper-like than                cupriclika047
 a  coppery - more coppery than /copper-like/     cupli        2 7 cupri clika
   -                                  -
cupri  1n is a piece/particle/atom of copper; comb. forms: cup/cur.       s306
 a  copper - made of copper             cupri        1 6
 n  copper - a piece/atom of element 29      cupri /cu/ 1 6
 n  copper - the mass of all such atoms      lo cupri /cu/ 1 6
1 cupli = cupri clika
   -                                  -
cuprium, la  na copper /cu/, the 29th element; also la cup/lo cupri, q.v.    300
na  copper - its long international name:    la cuprium          0
   -                                  -
curdi  4v x insures y against hazard w for fee h. 1n insurer        44&5--3---106
curdi  -b 3/3e sure 2/4c i ding 2/6s asegur-ar                         1f
vt  insure - insure...against...for fee...   curdi        4 6
 n  insurer - one who -s           curdi        3 6
 n  insurance - a spec. act of -ing          po curdi     3 6
 n  insurance - mass term of acts    lo po curdi     3 6
 n  insured - an -ed person, conv. of insure nu curdi     3 6
 n  insurable - can be -ed              nu curdi     3 6
 n  risk - that which is insured against, qv fu curdi     3 6
   -                                  -
curdu  2v x waters y/pours/sprinkles water on y.          cutridurzo247
vt  water - put water on... /water-do/       curdu        2 7 cutri durzo

generates the Universal format:

cupli           1,2-Cpx\7.4\'75\  %\
cupli           2,cupri\clika\\
cupli           4,(2a)\is more coppery/copper-like than
cupli           5,coppery\(a)\- more coppery than /copper-like/\cupli\2 7
cupri           1,S-prim\3.0\'75\  %\
cupri           4,(1n)\is a piece/particle/atom of copper; comb. forms: cup/cur.
cupri           5,copper\(a)\- made of copper\cupri\1 6
cupri           5,copper\(n)\- a piece/atom of element 29\cupri\/cu
cupri           5,copper\(n)\- the mass of all such atoms\lo cupri\/cu
cupri           6,cupli \cupri\clika\\
cuprium, la     1,Name\3.0\'75\  %\
cuprium, la     4,(la)\na copper /cu/, the 29th element; also la cup/lo cupri, q.v.
cuprium, la     5,copper\(na)\- its long international name:\la cuprium\0
curdi           1,C-prim\1.0\'75\44%\
curdi           3,-b 3/3e sure 2/4c i ding 2/6s asegur-ar
curdi           4,(4v)\x insures y against hazard w for fee h. 1n insurer
curdi           5,insure\(vt)\- insure...against...for fee...\curdi\4 6
curdi           5,insurer\(n)\- one who -s\curdi\3 6
curdi           5,insurance\(n)\- a spec. act of -ing\po curdi\3 6
curdi           5,insurance\(n)\- mass term of acts\lo po curdi\3 6
curdi           5,insured\(n)\- an -ed person, conv. of insure\nu curdi\3 6
curdi           5,insurable\(n)\- can be -ed\nu curdi\3 6
curdi           5,risk\(n)\- that which is insured against,\qv fu curdi\3 6
curdu           1,2-Cpx\2.4\'75\  %\
curdu           2,cutri\durzo\\
curdu           4,(2v)\x waters y/pours/sprinkles water on y.
curdu           5,water\(vt)\- put water on... /water-do/\curdu\2 7

x.6 Net from RJL, Metaphors and Semantic Neutrality

One of the first problems I want to deal with in reformatting definitions
is that of semantic neutrality. Whenever I try to discuss Loglan with a
prospective new recruit, I describe the language in terms of its unique
features, especially its unambiguity. Almost always, though, upon turning
to the dictionary, the recruit notices that the Loglan primitives and
complexes represent concepts that are abstract, and often ambiguous. The
most concrete words are usually complexes; the longer ones are most
concrete. Most of the primitive's English definitions carry all the
ambiguity and cultural stereotypes of 20th century American usage. To the
lay recruit, this almost always seems to cancel the theoretical
attraction of the multi-lingual roots found in primitives, and the
supposed cultural neutrality necessary to test Whorfian concepts.

Presuming that the Eaton tables justifiably present a set of culturally
neutral concepts that must be represented by primitives and short
complexes, it is necessary in Loglan to define these Eaton terms to
preserve that cultural neutrality. Some of the most common Loglan
primitives do not do so. Cluva (formerly clivu) represents the concept of
love. But the English 'I love you' conveys only one meaning of the word.
'I love my work' is another. There are also meanings never adequately
conveyed in English, like the Greek concepts of Eros and Agape. My
limited linguistics library shows over 20 Hebrew and Greek words
representing the concept of love as used in English in the Bible. This
would render the most basic abstract discussion of the topic totally tied
to whatever prior cultural associations the Loglanist speaker has for the
word 'love'.

Similarly, the English word 'run' has dozens of denotations, of which
prano, as defined in the current dictionary, could represent any of
several. Even the arguments for various predicates, which can limit the
scope of meaning, leave prano ambiguous. Colloquial English "I'm going to
run out to the store" usually does not mean the same as "I'm going to run
in the Boston Marathon" or "Will you run this upstairs". In Loglan,
hopefully, each of these meanings will have a different predicate
available. And the primitive has to be defined so as to be sufficiently
abstract to cover all such meanings consistent with the Eaton concept,
while avoiding (preferably) the connotative English jargon of, say, "run
this computer program", which does not fit most interpretations of the
abstract concept or the Loglan argument structure. Note that Eaton, and
most concept frequency analyses are probably limited in that they do not
account for common semantic misuse. The commonalty of the word 'run'
includes all the definitions listed in the dictionary, and related words
that are passed over (such as 'convey quickly', 'execute', or 'stocking
defect'), are less common in such analyses even though the concepts may
be more common.

Zipf has already struck English, causing ambiguity. It will probably
affect Loglan implicitly in ways that cannot be planned for. We must
therefore try to allow room for Zipfean effects, while not preserving
those implicit in English already. Yet, attempts to decide Zipfean
effects in advance will tend to put heavy English undertones in the
language. Since most Loglanists are Engliish-speakers, it is essential
for Whorfian testing to preserve the unique concepts in the language
against this cultural bias. For these reasons, I tend to be opposed to
changes which are added only to make Loglan seem more like 'natural
languages'. Let those who use the language determine that there is a need
for such changes before making them.

Incidentally, I asked JCB about the abstract concept expressed by cluva,
in this context. He defined it in the abstract sense of "strong emotional
attachment". Since this definition doesn't even convey a sense of
'positive' feeling, I would be reluctant to use cluva to express my
feelings toward a lady from a different cultural background. In a related
example, an earlier Lognet referred to two different metaphors for
'politics' which reflected two different philosophies, and in fact two
different English denotations of the word.

Thus, we need a cultural neutral definition of all primitives, and
preferably those Eaton concepts expressed as complexes. And we need a set
of derived complexes, hopefully including the primitive in the metaphor,
to express the various denotations of the English words representing the
concept in a Loglan-English dictionary. Presumably another, possibly
distinct set of complexes would be needed to write a Loglan-Chinese or
Loglan-Hebrew dictionary. But that is luckily not my problem.

A lot of careful thought and wording needs to go into each definition.
Each definition should be reviewed for semantic neutrality by linguists
familiar enough with one or more other languages, that can catch any
English cultural bias in the concept definitions. And a lot of looking at
Webster's, as well as perhaps a few bi-lingual dictionaries, will be
necessary to create 'families' (another ambiguous word - I should use the
mathematically precise 'sets') of related predicates. These should be
reviewed by Kieran's team.

One can sense that those working on the dictionary definitions will end
up creating as many, or more words for the dictionary as Kieran's team.
And without this effort, the 'interface' between Loglan and English will
remain incomplete.


x.7 Net from RJL, Borrowings

As I described above, my limited background involvement in Loglan is tied
to discussions in San Diego with JCB. One of our most frequent discussion
topics, and one that perhaps affected many of the early GMR ideas, was
the subject of adding words to the language to discuss complex ideas of
narrowly defined meaning. One of my few attempts at Loglan translation
was to try to convert the theme song from 'Man of La Mancha', a very
emotional poem, to what seemed a dry form of Loglan. The song was full of
its own set of metaphors, and I was curious whether they would come
across in translation.

The project died in its infancy, partly because my knowledge of the
grammar was too weak to manipulate the complex structures of the poem.
But I stopped trying when I realized that more than half the words in the
poem had no Loglan equivalent. I could not even develop good metaphors
that conveyed the original metaphors. Try to come up with a metaphor for
'trumpet' that denotes a musical instrument, and connotes the spirit one
thinks of when one hears 'the sound of the trumpets of glory'. 'Loud-
cone-shaped-musical-instrument', a 5-part metaphor got me to a dry
definition of the term that did not exclude trombones, tubas, or
sousaphones. But it just didn't make it. There is really no way to invent
a satisfactory metaphor for such concrete terms. Try 'I hurl down my
gauntlet to thee' for an easier example; I solved that one with existent
Loglan words to create a usable, if weak, metaphorical 'I (x) throw to
the ground (y) my safe-make(protect)-hand-shoes(gloves)(z)'. (I still
won't attempt the grammar). The concept of challenge is completely lost,
without the cultural association.

My point is that we won't, at this point, succeed in solving such
problems. Loglan does not yet have its own culture to translate such
metaphors into. *Iglu (no longer a legal borrowing) will always be
something more than bisli hasfa (ice-house) to an Eskimo, if not to a
Japanese or Italian. And the primarily English/American Loglanists
developing the language owe it to our Loglandian descendants not to
impose on them the cultural bias of including or excluding a separate
word for igloo.

In short, I speak in opposition to making iglu-type words automatically
acceptable (at the coiner's discretion) as Loglan words. In fact, the
Loglandization of borrowings should be avoided until significant usage
proves the need. If we don't, we have to examine every other tribal
language for their specific words for their house-structures, or we
present a cultural bias toward Innuit that only an American could
understand. The use of simba for lion, is similar. Most Loglanists have
never been exposed to Swahili. It is comparatively well-known in America
because of the heritage of our black culture. So I oppose arbitrarily
incorporating any n-Prim or 'borrowed word' into the Loglan wordspace in
a formal sense without careful consideration of the cultural decisions
being made.

This cannot be done by any individual, and perhaps shouldn't be done
until there is a substantial body of speaking Loglanists, and thus a
'Loglandian culture' that can adopt such borrowings permanently. At that
point, the Loglan Academy, as the French Academy, can choose which
borrowings should be adopted based on their usage in cultural discussion.

Having said that, I know it won't work, for the same reason that French,
as now spoken, goes beyond the 'official' language, due to the borrowings
necessary to discuss modern concepts.

This brings me to the related problem of jargon. Loglan should avoid this
vice of English like the plague. In the interests of concise speech,
Americans probably coin more words per day than Kieran's group will do in
3 months. Most of these will be accepted as, and be indistinguishable
from, normal English. And they will cause incipient confusion among the
unknowing listener (part of the intent, I'm sure). Loglan has had phases
of the trend towards jargon in its evolution. Back in TL1 and TL2, there
was an effort to coin primitives for various computer terms. God spare us
from bit-byte-nibble-core-dump-and-patch. (all of which are English
metaphors that may or may not come across in any other language). Then
there were color words. And I remember soksu, the carefully derived (from
all 8 base languages) word for oak. As if there is only one kind of named
tree family worth having a primitive for. (not to mention that there are
dozens of distinct species of oak). I note that 1 mod 3 format for
borrowings, that JCB originally proposed in response to my jargon
argument, has died, and borrowings are morphologically distinguishable
from primitives and complexes only by recognizing that they are not of
the latter form, and are not one of the 'known' c-Primitives. This is
equivalent to asking an English speaker to identify as jargon and/or
borrowings from other languages simply by the fact that one doesn't know
the meaning. When I was younger, 'compiler', and 'deja vu', representing
jargon and borrowed words, were as Greek to me as 'quantitative', which
is a Loglan primitive, 'australopithecine', a proposed borrowing, and
'morphology', much used among Loglanists but not yet in the dictionary.

I have not yet decided whether the approach to acronyms in TL6
sufficiently discourages them to render them less used in the jargon
filled military-industrial-government complex that I serve each day. I
doubt it, and the easy use of acronyms in Loglan would make it prey to
the same tendency as American English to abbreviate without pause. This
particular problem is less worrisome to me, though, since Loglan has the
option, at least, of writing acronyms as words. What if it were the only
option? In the government documents I read, it would enhance
understanding immensely if the extra effort to spell out an acronym led
to fewer acronyms being used. Of course, the acronymic concept may be
cultural basic to Loglan, since the free variables da, de, di, do, and du
are acronymic in nature, but necessary parts of Loglan grammar.
Incidentally, jargon has already hit Loglandia, since acronyms like GPA,
GMR, jcb, MacGram, MacTeach, and pc are so common as to be manipulated as
words, even by me.

Whatever the decision on the problems of borrowings, existing n-
Primitives, 1 mod 3, jargon, and acronyms, as dictionary foreman, I'm
stuck with the problem of determining which borrowings to put into the
dictionary, and how to determine the preferred word forms for various
concepts. I see no easy algorithm, so I will repropose some of the ideas
I came up with 6 years ago, before GMR made it impossible to coherently
discuss the problem.

I propose four morphological/grammatical forms exist for words and
concepts derived from sources other than primitives and metaphors. These
forms are:

     a. True-borrowing
     b. Designator-form
     c. Pseudo-complex
     d. Approved borrowing

True borrowings are those words and or phrases used in Loglan speech with
no pretense or attempt to call them 'Loglan'. The grammar supports this
currently with quotations li ... lu, and  lae. (See TL4/2 4.6.4). These
obviously would not belong in a Loglan dictionary (except for the
grammatical lexeme words).

Designator form would be the first step in incorporating a borrowed
predicate into Loglan. If one wishes to 'borrow' a word not currently in
the language such as 'gauntlet' or 'trumpet', one could form a borrowing
in designator format, be spared the overhead prescribed in TL4 for true
borrowings. (TL4 requires that one specify that the  English word
'stingy' be associated with the Loglan katli, as well as enveloped in
li...lu, and prefixed by lae.) Note that when introducing a new
borrowing, say iglu, under the current corpus, one is not required to
identify that an iglu has any relation to either a residence or
architectural form. Designator form would be restricted only to, and
defined as, predicates, as are the current borrowings. They would be
grammatically and morphologically similar to names, except that a
different Little Word (the designator) instead of la would be used to
designate the word as a new or infrequent borrowing. Lae may be usable
here - I leave it to the grammarians to determine if that would be
unambiguous - or another Little Word, preferably similar to la and lae
must be assigned. Stress and the ending consonant-pause would be required
to be the same as for names, and spelling would have to use Loglan
phonetics. However, the user would not have to worry about consonant
pairs or possible resolution as a complex because the designator-
consonant-pause would render the word inherently resolvable. The listener
would know the word was a borrowing. The user could apply the word
grammatically as any predicate or argument (as presumably a name can be).
The listener would be warned that the word is an infrequent borrowing and
would therefore be prompted to ask its meaning if it is not already
known.

Pseudo-complexes are a new concept that I have recently come up with, so
their form may need resolvability analysis to determine what is
acceptable. The concept is based on the idea of borrowing-complexes
mentioned by JCB on page 37 of TL6/1. This form would be initially used
for the large families of borrowings that are required to support normal,
but non-jargon, conversation and metaphorical literary speech. Without
it, the typical TL or Scientific American translation, as well as the
poetry of 'Man of La Mancha', would be filled either with quoted
borrowings, constructs similar to names that convey no meaning other than
their grammatical nature, or coined borrowings of the arknidia form that
carry no more meaning to the listener/reader than the designator form,
but seem like 'real words'; i.e. jargon. The form of a pseudo-complex may
still require a Little Word prefix like designator-form (I defer to the
grammarians and morphologists). To be easily recognizable as opposed to
merely resolvable, I would prefer a form such as 1 mod 3 with at least
one CC pair, followed by a legal complex form that represents a primitive
or metaphor which 'types' or 'classes' the meaning of the word. Thus, an
acceptable word would be igluhaa (igloo-house, the common usage of igloo)
or iglutektosensi (igloo-architecture, a less likely but plausible usage
of the root). Iglumao (igloo-maker), the borrowing complex form suggested
by JCB, would be ambiguous, indicating only that an iglu is a 'made
thing'. But igluhasmao would be a preferred form. It is not clear to me
whether meanings would be totally clear - the latter word could be
interpreted as a borrowed form of 'house-builder' as well as its intended
meaning. But then all complexes are based on metaphors that are subject
to misinterpretation.

The pseudo-complex form would be true Loglan with a clear metaphorical
emphasis while retaining its obvious background as a borrowing. Most
borrowings that we put in the dictionary should be of this type. A list
of animal borrowings would have a borrowed 1 mod 3 (preferably, but see
below) root followed by -nia to indicate the meaning. Plants would have -
herba appended since there is no affix assigned. (This might lead to such
common suffixes being assigned an affix even though they cannot be
justified for an affix based on Eaton concepts.) Trees would end with -
tri, so that oak could be rendered as oksutri. For scientific-related
terminology, however, I would prefer to stick to Greek roots and Latin
taxonomic names unless the root can be legitimately claimed to be
International in scientific usage (This can be determined by looking at a
German, Russian, or Japanese translation of a scientific article and
seeing whether the English root has been borrowed. 'Astronaut' would not
be a basis for a borrowing - the Russian 'cosmonaut' form is more
internationally recognizable and is equally based on a classical root.
(It is accepted in English text, whereas 'astronaut' is not used in the
Russian.)) Four (or possibly five, if German is counted) of the eight
basic languages use a significant number of Greek and Latin roots. Thus
we would lose simba, or perhaps merge it with the root leo to form
simbleo- that would be metaphorized to form simbleonia. I would consider
an argument that roots need not be 1 mod 3 if the pseudo-complex remains
easily resolvable. But in practice, using 1 mod 3 roots is a preferable
convention, allowing easy recognition by non-computer resolvers.

That last form is reserved for those roots that through common Loglan
usage become commonly accepted throughout Loglandia. In effect, these
would be Loglandian I-Primitives. Thus iglu-, which initially would be
required to be defined metaphorically with -haa, could become a stand-
alone Loglan primitive if the usage and acceptability of this particular
borrowing became universal. This status should be reserved for only the
most frequently used borrowings, and should require adoption by the
Loglan Academy before they could be added to a dictionary. I see no words
except perhaps the N-Primitives and I-Primitives in the current
dictionary and those that were remade as part of GMR that could have this
rule applied. Of course, if forms for roots other than 1 mod 3 were
permitted, or if roots could end in consonants (a question for the
resolution algorithm experts), the new dotci could be written as
dotcypiu.

Thus a typical borrowing might progress through all four forms as its
usage and general acceptance became common. Yet it would always preserve
its distinctive borrowed flavor, as well as its Zipfean emphasis on word
length vs. usage frequency. It also eases my problem of determining what
types and forms of borrowed words should be included in the dictionary,
as well as which lists are appropriate in the Appendix.

x.8 Net from RJL, Arguments

No, I am not referring to what the last entry is likely to cause.

 Rather, I am referring to the one most significantly undecipherable
aspect of Loglan not remedied by GMR - the multiple arguments permitted
to each primitive and/or complex. The best way to explain s by example:
mutzavkao (the former vedzafka - evildoer) has the current dictionary
definition "X acts wickedly in doing Y, by moral standard W". A clear
definition, if one uses the dictionary. But the moral standard argument
'W' is an invention of the previous dictionary team. None of the three
components of the metaphor used in the complex refers to a moral
standard. If I were hearing the metaphor for the first time, I would make
the guess that I would use one of the components arguments, probably that
of kakto, the last one. But W in kakto (and zavlo, for that matter)
delineates a purpose, not a moral standard. In usage, I could get an
erroneous (or at least confusing) idea of the intent of the metaphor if I
had never seen it before. (I will note in passing that fu zavlo or fu
kakto are acceptable, if limited, translations of English purpose, that
are not in the dictionary. Shouldn't they be?)

The dictionary editor thus has two problems to solve:
     What is the argument structure of a newly defined metaphor
(hopefully algorithmically determined as easily as the word was built
from the metaphor).
     What passive argument exchanges are defined. (Shouldn't they all be,
if possible,, since we have no usage history?)

In addition, argument structures should be made parallel and simple for
primitives to allow them to be memorized as easily as the word itself.
Dorja (war) and kamda (fight) have similar 3-place argument structures.
But cteki (tax) and lilfa (law), which are imposed by Y on W for purpose
H (in monetary or barter units Q, or is it - on goods Q?) That plausible
meaning that one might guess when learning the two primitives, is correct
for neither. Coupled with the non-parallel structure is that each is
permitted to be used without specifying all the arguments.

There is probably no easy cure for the argument structure of primitives,
though we can try to specify sufficient arguments to cover most
conceivable contexts. We will have to, of course, compare with Eaton's
concepts. But the metaphor argument algorithm is a must, as is any
consensus on passive exchanges and permissibly omitted argument forms to
be included in the dictionary. Any ideas?

Note, possibly unrelated to the dictionary work:

I still have some uncertainty on the grammatical uses of multi-argument
predicates. Specifically, I am wondering if it is permitted to use more
than one such predicate in an utterance. I do not see the Little Words
that allow me to refer to arguments of more than one predicate in a
single utterance. Thus, if I wish to speak of an evil tax, using
mutzavkao and cteki, I see no way to refer to the moral standard W of the
first predicate, if the subject of my utterance is the latter predicate,
and I am using its arguments throughout the discussion. Is this covered?
I may have missed something in the TL7 corpus. I could always combine the
two metaphorically, hopefully retaining all the arguments of each
according to whatever the metaphorical argument algorithm allows. But
this may lead to requiring undefined place-holding little words. I have
an admittedly contrived example that combines 35 primitives of various
argument structures into a single desired utterance - even using
metaphors doesn't help much, and breaking up the utterance in the
contrived context would be confusing, and would also cause difficulties
in emphasis. I'm not including it here, but if anyone answers my
grammatical question, I may make it available if I still can't figure out
how to translate it.
