[developers] cfrom/cto in cheap - bug when packing?

Ben Waldron benjamin.waldron at cl.cam.ac.uk
Thu Dec 1 22:41:22 CET 2005


Please see

http://wiki.delph-in.net/moin/LkbPreprocessing

for documention of existing preprocessing machinery. There are no 
equivalent PET or HoG pages as yet.  We aim for an independent 
preprocessing module and plan to integrate the LKB's embedded 
preprocessor into PET via ECL.

Please feel free to contribute to the online Wiki ! :)

- Ben

Francis Bond wrote:
>G'day,
>
>  
>>After consultation with Ben Waldron today, I can report that if you do the
>>following three things you can get characterization to work for the ERG
>>in the LKB, though I don't know what would be required for cheap.  So
>>if your question was just about cheap, then never mind this message.
>>    
>
>My question was mainly about cheap as I wanted the values for the RMRS
>for the dictionary batches.  I beleive the HoG people enabled this
>magic, but I haven't been able to do it here.  However, I also welcome
>it's appearance in the LKB (^_^).
>
>  
>>And note carefully that the changes below are not guaranteed to be
>>compatible with other functionality such as that provided by [incr tsdb()]
>>or the Redwoods machinery. So caveat adoptor :)
>>
>>At present, the three steps are as follows:
>>1. In erg/lkb/globals.lsp, include the line
>>   (setf *characterize-p* t)
>>2. In erg/lkb/user-fns.lsp, include the following redefinition
>>   (defun preprocess-sentence-string (str)
>>     (x-preprocess str :format :chared))
>>3. In erg/lkb/script, replace the line
>>   (read-preprocessor (lkb-pathname (parent-directory) "preprocessor.fsr"))
>>   with the line
>>   (x-read-preprocessor (lkb-pathname (parent-directory) "preprocessor.fsr"))
>>
>>I have not experimented extensively with these alterations, and direct
>>queries to Ben.
>>    
>
>Will do,
>
>--
>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