A new lexical module for NooJ

1
A new lexical module for NooJ

• Max Silberztein
• LASELDI, Université de Franche Comté

2
DELA INTEX

• DELAS aider,V3tr
• automatic construction of a DELAF
• aiderions,aider.V3trC1p
• DELAC a few tools but no full support
• manual construction of a DELACF
• cousine germaine,cousin germin.NNAHumfs
• DELAV no support except in DELAFs
• Lexicon Grammar tables (very!) frozen
  expressions / no support for simple words

3
In NooJ

• No more DELAFs
• No more DELACFs
• DELAS, DELAC DELAV are merged
• aider,VtrFLXAimer
• cousin germain,NNAHumFLXCousinGermain
• czar,tsar,NHumFLXCrayon
• Lexicon-Grammar tables are supported

4
NooJ dictionary
5
Dictionary file format

• no more inflectional information is no
  longer limited to one character
• Present PP 3rd sing fem
• Inflectional information is separated from
  category no more A31 vs N31
• ami,NFLXCousin
• vert,AFLXCousin
• no more alphabet file (corrected Unicode)
  sorting is performed according to language
  culture information.

6
Special properties

• FLXAimer (inflectional paradigm)
• NW (Non Word)
• ha,VNWFLXHave
• UNAMB (UNAMBiguous)
• cousin germain,NUNAMB

7
Inflectional description (flx)
8
Inflectional desc. for compounds
9
Inflectional grammars (nof)
10
Table view Lexicon-Grammars

• no more meta-graphs gt only one type of syntactic
  grammars LG lexical information can be used,
  e.g. ltVN0VN1gt, ltVN0Humgt ltVABLEgt etc.
• information has a type features become values of
  properties
• either explicitly
• FLXAimerTensePresentPers3Numbersing
• or implicitly, thanks to a properties definition
  file
• Present3sing

11
Properties definition files
12
Table view lexicon-grammars
13
Perspectives

• a new property, e.g. DRVIser ?
• france gt franciser, refranciserons,
  défranciseable...
• generate inflected forms during analyses, e.g.
• V_PP N_Iser_PR3s V_Ationp
• lexicon grammars for non-contiguous expressions