Stefan Mueller (Stefan.Mueller@dfki.de)
Thu, 07 May 1998 14:04:12 +0200
Hi, I do not understand why you think that the Babel grammar is a Prolog hack. The syntax of the lexical entries is similar to PATR-II. So it's basically path equations. Instead of having a '|' for disjunktion I use ';' which incedently is the disjunction in Prolog too =;-). What follows are two example lexical entries written in the Babel syntax and in PAGE syntax. Regarding the syntax the only difference are brackets. Until I started to develop grammars with PAGE I have been a vi user. Without emacs it is probably not possible to cope with all the brackets. verb_stamm(LS) :- LS isa trans_verb_erg(Ref1,_,Ref2,belebt,[m,m,p,m],p), LS:pres1 <=> "fall", LS:pres2 <=> "faell", LS:past <=> "fiel", LS:part <=> "fall", LS:sep_praef <=> auf, LS:flex_m:fk_v <=> stark, LS:flex_m:fk_v_agr:ep <=> m, LS:flex_m:fk_v_agr:evp <=> m, LS:flex_m:fk_v_agr:epp <=> p, LS:flex_m:fk_v_agr:sm <=> m, LS:flex_m:fk_v_agr:pl_n <=> m, LS:flex_m:fk_v_agr:spe <=> m, LS:flex_m:fk_v_agr:tm <=> m, LS:flex_m:fk_v_agr:em <=> m, LS:flex_m:bet1 <=> p, LS:synsem:loc:cont:reln <=> auffallen, LS:synsem:loc:cont:experiencer <=> Ref2, LS:synsem:loc:cont:topic <=> Ref1. Sign isa intrans_verb_erg(Ref,Belebt,Loc,Dir) :- Sign isa log_subj_verb, Sign:synsem:loc:cat:head:auxf <=> sein, Sign:synsem:loc:cat:head:subj <=> [NP_str], Sign:synsem:loc:cat:head:erg <=> [NP_str], Sign:synsem:loc:cat:subcat <=> [], Sign:synsem:loc:cat:vcomp <=> none, NP_str isa ref_comp_noun_synsem, NP_str:loc:cat:head:cas <=> Case, Case isa struc_case, NP_str:loc:cat:subcat <=> [], NP_str:loc:cont:ind:ref <=> Ref, NP_str:loc:cont:ind:belebt <=> Bel, belebt(Belebt,Bel), Sign:synsem:loc:cont_t:loc <=> Loc, Sign:synsem:loc:cont_t:dir <=> Dir. PAGE: AUFFALLEN-f-VAR := @final-praefix-verb() & @trans-erg-verb ($STAMM = 'auffall, $RELN = 'auffallen). trans-erg-verb($stamm,$reln) := trans-erg-verb-lex & [ morph <! [ stem $stamm ] !>, synsem.loc [ cont [ PRED $reln ] ] ]. super-trans-erg-verb-synsem-syn := log-subj-verb-synsem & [ loc [ cat [ head [ auxf haben-auxf, subj < #erg & saturated-ref-comp-noun-synsem & [ loc.cat [ head.cas struc-case ] ] >, erg < #erg > ], subcat < saturated-ref-comp-noun-synsem & [ loc.cat [ head.cas struc-case ] ] >, vcomp *undef* ] ] ]. Plus some more types ... Anyway, Babel is not intended to be a development system. I build it for me in order to test my grammar and I do not have the time to give support to other developers. So your decision not to use it was right althoug for the wrong reason. Regarding the properties of PAGE: PAGE has: a very powerful formalism. - ditributed disjunktions - possability to switch between open and closed world which is a good thing if you are working with a closed world and your analysis fails because of a missing type you can switch to open world reasoning and you see a type conjunction that lead to the unification failure. - complex types like expl-noun-synsem := noun-synsem & saturated-complement-synsem & [ loc [ cat [ head [ cas.morph-case nom-acc-val, expl + ] ], cont expletiv-pronoun-sem-type & [ INDEX [ pers third, num sg ], KEY [ VREF third, VTYPE 'event_std ] ] ] ]. I think this is not possible in ALE. You have to do this via makros + types. a good development environment - Chart Display - Console - Graphical Feature Editor (FeGramEd) PAGE does not have: - functional or relatinal constraints attached to feature structures used for instance in the new extraction theory of Bouma, Malouf & Sag, or to implement principles - You cannot change the morphology of words with lexical rules lexical rules are used online so they are basically unary projections Greetings Stefan P.S. I didn't understand what you meant by: > Contra's: > 1. I head its underlying logic is poorly understood. St. -- FB Sprachtechnologie DFKI GmbH Tel.: (+49 - 681) 302 - 5295 Stuhlsatzenhausweg 3 Fax: (+49 - 681) 302 - 5338 D-66123 Saarbruecken http://www.dfki.de/~stefan/ http://www.dfki.de/~stefan/Babel/Interaktiv/Babajava/
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:38:19 PST