web tool help

s.t. bischoff bischoff.st at GMAIL.COM
Mon Jul 28 19:40:26 UTC 2008


there is a fellow working on a non-commercial toolkit that improves on some
of the xfst/lexc stuff. the major difference is simplified notation for
things like reduplication, unification, long distance dependencies, free
morpheme order, etc. The flag system seems a little combersome in the xerox
kit, and this free kit will do away with them entirely. hopefully it will be
available soon, though the xerox kit works remarkably well. Karttunen (one
of the authors) says that "phonology and morphology are solved"...as more
people work with this fst technology we'll see.

thanks for the advice.

On Mon, Jul 28, 2008 at 3:15 PM, William J Poser <wjposer at ldc.upenn.edu>wrote:

> >For more advanced needs, it might be interesting to get xfst and lexc
> >source code and make some adaptations (I have no clue whether Xerox made
> >their source code available, nor whether their license would allow any
> >modification, though).
>
> xfst and lexc are proprietary. Xerox does not release the source.
> They put a lot of  work into optimizations such as compressing the
> compiled transducers and consider the greater speed and smaller
> footprint a selling point. However, although I am a fan of free
> software (in fact, the Xerox tools are the only
> non-free software that I use), I doubt very much that anyone would
> have a need to modify xfst or lexc since they are general purpose
> tools.
>
> For non-commercial purposes they can be obtained cheaply as they
> come on a CD with the book _Finite State Morphology_.
> (
> http://www.amazon.com/Finite-State-Morphology-Kenneth-Beesley/dp/1575864347/ref=pd_bbs_sr_1?ie=UTF8&s=books&qid=1217272389&sr=8-1
> )
> See: http://www.stanford.edu/~laurik/fsmbook/home.html<http://www.stanford.edu/%7Elaurik/fsmbook/home.html>for more info.
>
> Bill
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/ilat/attachments/20080728/64c4eb83/attachment.htm>


More information about the Ilat mailing list