WikiDiscuss

On Sunday 13 February 2005 15:55, Jorge "Llambías" wrote:
> --- Pierre Abbat wrote:
> > The parser shouldn't be lexing text inside a zoi-quote; it should only be
> > looking for the delimiter.
>
> That would be much more complicated, because to recognize a delimiter
> it would have to do the full syntactic parsing rather than merely
> morphological. It's not enough to detect a zoi or a la'o, because
> these could be affected themselves by other magic words.
>
> It's probably doable, but I don't think it is worth the complications
> it introduces. I prefer word lexing to be strictly separated from
> syntactic parsing. The morphology algorithm should be able to break
> any stream into words (including non-lojban words) without having any
> clue as to the meaning or grammar of any of the words.
>
> Any string separated by spaces that contains a unrecognized character
> is considered a non-lojban word. More generally, any string separated
> by spaces (or pauses in speech) that cannot be _fully_ broken into valid
> Lojban words is considered a single non-lojban word.

The problem with that is that a single foreign word could look like two Lojban
words each of which is also a foreign word. Thus {zoi gy i got to go to togo
gy} parses as "i got to go to to go".

phma
--
1 m = 3*3*5*7*47*44351/73/293339 * Cs133