;;; -*- Mode: TDL; Package: LKB; Coding: utf-8 -*- ;;; ;;; Based on: ;;; JACY Grammar: ;;; Developed by Melanie Siegel, Francis Bond ; (MS 2004-04-02) Empty case particles. Try to restrict: to case particle function and lexical nouns. ; I'm sure this will add loads of ambiguity. Only for parsing spoken Japanese! ; (MS 2004-12-14) RELS and HCONS and HOOK should come from the argument. ; (MS 2004-12-16) Maybe this adds too much ambiguity. Taking it out again. ; (FCB 2006-19-31) can we restrict it to adjacent arguments ? ;pp_np_rule-case := empty-preposition-type & ; [SYNSEM [LOCAL [CAT [HEAD empty-case-p_head & [CASE ga-or-wo], ; VAL.COMPS < >], ; CONT [HOOK #hook, ; RELS #rels, ; HCONS #hcons]], ; NON-LOCAL.ROBUST +], ; C-CONT.RELS , ; ARGS.FIRST.SYNSEM [LEX +, ; LOCAL [CONT [HOOK #hook, ; RELS #rels, ; HCONS #hcons], ; CAT.VAL saturated]]]. ; (MS 2004-12-22) Why did we have two rules for adjectives and non-adjectives? I'll take that out. rel-cl-sbj-gap-rule := rel-cl-gap-rule & [SYNSEM.NON-LOCAL.UNDERSPEC_ARG -, ARGS <[SYNSEM.LOCAL.CAT [HEAD.ADJ bool, VAL [SUBJ opt-1-arg & [FIRST.LOCAL.CONT.HOOK.INDEX #sbjind], COMPS onull]]], [SYNSEM.LOCAL.CONT.HOOK.INDEX #sbjind ]>]. rel-cl-obj1-gap-rule := rel-cl-gap-rule & [SYNSEM.NON-LOCAL.UNDERSPEC_ARG -, ARGS <[SYNSEM [LOCAL [CAT [VAL [SUBJ onull, COMPS opt-1-arg & [FIRST.LOCAL.CONT.HOOK.INDEX #obj1ind, REST onull]]]]]], [SYNSEM.LOCAL.CONT.HOOK.INDEX #obj1ind ]>]. rel-cl-obj2-gap-rule := rel-cl-gap-rule & [SYNSEM.NON-LOCAL.UNDERSPEC_ARG -, ARGS <[SYNSEM.LOCAL.CAT.VAL [SUBJ onull, COMPS 1-plus-list & [REST.FIRST.LOCAL.CONT.HOOK.INDEX #obj2ind]]], [SYNSEM.LOCAL.CONT.HOOK.INDEX #obj2ind ]>]. ; (ERB 2001-12-13) For cases where the splitter can't break apart ; two sentences for lack of punctuation. This rule allows two sentences ; be coordinated with no overt marking of it. In pet/japanese.set ; it is marked as a spanning-only-rule so that it can only apply at ; the very top of the tree. I guess that won't be true in the lkb?? ; (ERB 2001-12-13) This is causing some memory problems with cheap. ; Perhaps the problem is that nothing is requiring the second daughter ; to be a sentence. Adding that. ; (ERB 2001-12-13) Still causing memory problems, so commenting this ; out (as well as its ranking in japanese.set) until I can consult ; with Uli on Monday. ; (ERB 2001-12-17) Uli fixed the problem: the spanning-only-rules ; condition wasn't really being checked. But this is still producing ; way too many parses. Part of the problem is that te forms are ; [H-TENSE tensed] instead of [H-TENSE te] (in order to get the pure ; aspect analysis of the aspect auxiliaries). So, try blocking the ; te forms in the left daughter here by requiring [FIN +]. ; (ERB 2002-01-02) Don't want this to be [NUCL +] on the ; second daughter, if hadj rules are going to be [NUCL -]. ; (MS 2002-1-7) I'll restrict the rule a little bit further to non-adjectives. ; Also, both verbs are MAIN-PRD +. This prevents NA to be the first verb. ; MODUS nonoperative prevents NAI to be the first verb. It has to be seen if that is too restrictive... ; (MS 2002-1-7) It is: "表示できませんでした申し訳ございません. ; (MS 2002-1-10) Another restriction is EMPTY - in the head of the first argument. ; This prevents an empty-cop-id-rule or an empty-cop-rule to be applied to this. ; (ERB 2002-01-22) NUCL conj now on supertype. ; (ERB 2002-01-29) If we want runon_s with SAPs on the end, we need the ; head-dtr here to be utt-valid instead of verb_head. (Remembering that ; in cheap at least this rule is restricted to spanning-only-edges.) ; (MS 2002-05-28) If we want the first sentence to have an SAP, we need utt-valid there, too. ; (MS 2002-06-05) Maybe that's not a good idea, leads to spurious ambiguity in, e.g., ; ご飯を食べるかよい ; (MS 2002-06-19) But we seem to need this anyway: "口座振替依頼書送ってください,お願いします" ; (MS 2002-06-20) Trying to restrict this with NUCL nucl_plus. ; (FCB 2005-06-28) take this one out for the time being ; runon_s := sentence-coord-type & ; [SYNSEM [LOCAL [BAR +, ; CAT.VAL saturated], ; NON-LOCAL.ROBUST +], ; ARGS <[SYNSEM.LOCAL [CAT [HEAD utt-valid & [H-TENSE present-past-or-future, ; FIN +, ; EMPTY -, ; ADJ -, ; MAIN-PRD +], ; VAL saturated], ; BAR +, ; NUCL nucl_plus], ; J-NEEDS-AFFIX -], ; [SYNSEM synsem & [LOCAL [BAR +, ; CAT [ HEAD utt-valid & ; [ H-TENSE present-past-or-future, ; MAIN-PRD +, ; EMPTY -], ; VAL saturated]]], ; J-NEEDS-AFFIX -]>].