README for src/data/italian (Italian CCGbank)

(1) Put the ctu files in the directory data (or create symbolic links), with these names:

newspaper.ctu
civil_law.ctu
jrc_acquis.ctu

(2) Type: make ccgbank

(3) Alternatively, "manually". Converting TUT-Constut format into Prolog format

scripts/_convert data/<FILE>.ctu data/newspaper.pl

Generating CCG derivations (with SWI Prolog)

swipl -c prolog/tut2ccg.pl data/newspaper.pl > data/newspaper.ccg

TO DO LIST

• redirect error messages to a specific file
• apposition (all cases need to be checked)
• check punctuation (e.g. comma should modify previous constituent (?))
• check with Cristina analyis of V-502 (relative clause with pro-drop)
• check with Cristina analyis of EVALITA-NEWSPAPER-90 (head should be arg?)
• Interrogative pronouns require type raising. (Example: ALB-154) ccg(leaf(Pos,Ind,Tok), a, Cat, t(Cat,Pos,Ind,Tok)):- intpro(Pos), Cat = (s:whq/(s:dcl\np)).
• Sometimes special characters, sometimes quotes (Example: ALB-271)
• Check analysis of CHIAM-17 (Doctor J)
• Open Parenth. without closing (A-12)
• Check with Cristina: There are three cases for ..-CONTIN+NUM (see prolog/binarise.pl). Is this supposed to be like this?
• Type raising for past participles? (ALB-26)

INTERESTING CASES (still to be done)

• V-301 Vediamo cosa si puo` fare (s\np)/s s:whq/(s:dcl/np) np (s:dcl\np)/(s:inf\np) (s:inf\np)/np ----------------------------------B> (s:dcl\np)/np -------- s:X/(s:X\np) -------------------------------------------------B> s:dcl/np ----------------------------------------------------------------------B> s:whq Vediamo cosa si puo` fare (s\np)/s s:whq/(s:ynq/np) np (s:q/(s:inf\np))\np (s:inf\np)/np ---------------------------------< s:ynq/(s:inf\np) -------------------------------------------------B> s:ynq/np ----------------------------------------------------------------------B> s:whq

• Extraposition (CHIAM-49) I wonder extraposed subjects should be dealt with by crossed type-raising, as in:

  comincio` l'attesa s\np np --------Tx s\(s\np) ------------------< s