[developers] Re: Morphological ambiguity and xfst-lkb interface

Francis Bond bond at cslab.kecl.ntt.co.jp
Mon Apr 11 11:11:27 CEST 2005


>> I'm curious what :cto and :cfrom (properties of morph-edge
>> and chart-edge) are, especially since chart-edge also has
>> :from :to.  I couldn't find any enlightening comments...
> :cto and :cfrom came in when we discussed the combination of shallow and 
> deep NLP in DeepThought. The meaning is the character position of the 
> relations in the RMRS, such that the modules can communicate.

There seems to be some confusion as to whether :cfrom and :cto are
token or character positions.  The current cheap output appears to be
token-based (pet-0.99.7 with a recent JACY) but I think the HoG demo
(at least the English) uses character positions (starting from 0).

ChaSen uses bytes, which is really annoying as you get different
results depending on the encoding.

Francis Bond  <www.kecl.ntt.co.jp/icl/mtg/members/bond/>
NTT Communication Science Laboratories | Machine Translation Research Group

More information about the developers mailing list