;;; -*- Mode: TDL; Coding: utf-8 -*- ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; file: newlexsem-types.tdl ;;; module: mrs semantics ;;; version: 01 ;;; written by: Melanie Siegel/Emily Bender ;;; last update: %G% --- %U% ;;; updated by: ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; author | date | modification ;;; ------------------|-------------|------------------------------------------ ;;;Melanie Siegel (MS)| |Emily Bender (ERB) , Takaaki Tanaka (TT) ;;; | | ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ; (ERB 2001-09-13) Moved LTOP = KEY.LBL constraint to MRS, so I'm taking it ; out of all of these types. ; (MS 2001-11-5) I'm afraid I have to undo this for the treatment of HAYAKU NEREBA YOI. ; (ERB 2001-11-26) I don't want nominal-sign-sem-type to inherit ; the constraints identifying its MOD-IND with its INDEX and its ; MOD-HAND with its LTOP because this is incorrect for the subtype ; nom-vn-sem-type. Making a subtype of nominal-sign-sem-type which ; does inherit all of these constraints. ; (MS 2003-07-09) changed due to matrix nominal-sign-sem-type := wg-handel-lex-mrs. mod-id-nominal-sign-sem-type := nominal-sign-sem-type & mod-id-wg-handel-mrs. common-noun-sem-type := mod-id-nominal-sign-sem-type & [ RELS diff-list & [LIST.FIRST reg-nom-relation ]]. ;date-noun-sem-type := mod-id-nominal-sign-sem-type & ; [ KEY #key & locable_rel, ; RELS 1-dlist & [LIST.FIRST #key ], ; HCONS ]. date-noun-sem-type := mod-id-nominal-sign-sem-type & [ HCONS ]. ; (TT 2003-08-14) temporarily added nadv-sem-type := mod-id-nominal-sign-sem-type & [ HCONS ]. ; pron-sem-type := mod-id-nominal-sign-sem-type & [HOOK [LTOP #handle, INDEX #i], RELS diff-list & , HCONS ]. ppron-sem-type := pron-sem-type & [RELS ]. ; (ERB 2001-09-13) This type identifies LTOP with the LBL of ; the arg_rel. This has the happy consequence that the LBLS of ; the arg_rel and the place_rel are the same. lpron-sem-type := mod-id-nominal-sign-sem-type & [HOOK [LTOP #hand, INDEX #ind], RELS , HCONS ]. personal-pronoun-sem-type := ppron-sem-type & [RELS.LIST.REST.FIRST.PRED def_rel]. det-sem-type := mod-id-wg-handel-mrs. lex-det-sem-type := det-sem-type. nom-sem-type := mod-id-nominal-sign-sem-type & [RELS diff-list & [LIST.FIRST noun-relation]]. nom-pred-sem-type := mod-id-nominal-sign-sem-type & [RELS diff-list & [LIST.FIRST noun-arg1-relation]]. ; (ERB 2001-10-31) I want to disallow NO from modifying NO ; nominalizations while still allowing it to modify vns used as ns. ; I'm doing this based on the KEY, so I've introduce a new kind ; of nominalise_rel for this type. (Replacing nominalise-pred_rel ; with nominalise-vn_rel.) ; (ERB 2001-11-26) We want to expose the event of the embedded ; verb as a target for modification. This means that the MOD-IND ; and the INDEX shouldn't be identified for this type. See notes ; at the vn2n rule in rule-types.tdl. ; nom-vn-sem-type := nominal-sign-sem-type. ; (FCB 2005-06-30) Move this inside the lexical rule ; [SYNSEM.LKEYS.KEYREL noun-arg1-relation]. adj-sem-type := mod-id-wg-handel-mrs & [ HOOK.INDEX #inst, RELS diff-list & [LIST.FIRST adj-relation & [ ARG1 #inst]]]. ;(MS 2004-06-25) added the linking of LBL to LTOP. na-adj-sem-type-basic := mod-id-wg-handel-mrs & [ HOOK [INDEX #event, LTOP #handle], RELS ]. na-adj-sem-type := na-adj-sem-type-basic & [ HCONS ]. ; (ERB 2001-08-06) In Smith-style number names, the numbers sometimes ; have to also provide the times_rel and the plus-relation. ; (ERB 2004-1-16) See notes at card-lex: exposing the ARG1 as the XARG. ; (FCB 2006-06-20) no MOD-IND/MOD-HAND ; FIXME check the semantics card-sem-type := mrs & [HOOK [INDEX #ind, XARG #xarg], HCONS , RELS diff-list & [LIST.FIRST const-relation & [ARG0 #ind, ARG1 #xarg]]]. ; card-sem-type := mrs & ; [MOD-IND #ind, ; MOD-HAND #hand, ; HOOK [INDEX #ind, ; LTOP #hand, ; XARG #xarg], ; HCONS , ; RELS diff-list & [LIST.FIRST const-relation & [ARG0 #ind, ; ARG1 #xarg]]]. ; (ERB 2001-08-28) Date expressions date-sem-type := mod-id-wg-handel-mrs & [HCONS , HOOK.INDEX #ind, RELS diff-list & [LIST.FIRST numbered-np-relation & [ ARG0 #ind ]]]. simple-adj-sem-type := adj-sem-type. dim-adj-sem-type := adj-sem-type & [ RELS diff-list & [LIST.FIRST adj-relation], HCONS ]. ;(MS 2004-06-24) As the n-conjunction gets an index, it should be quantified and therefore ; needs something in its HCONS. Therefore I erased the [HCONS ] condition. n-conj-sem-type := mod-id-wg-handel-mrs. ; (ERB 2001-09-13) Nothing is identifying KEY.LBL and LTOP for ; these guys. This constraint might belong somewhere higher up. ; (ERB 2001-10-10) In order for the k2y code to recognize PPs as modifiers, ; they need to be ARG0.E.TENSE no_tense. ; (ERB 2001-11-26) This should inherit constraints identifying ; MOD-IND/HAND with INDEX/LTOP. ??? prep-sem-type := mod-id-wg-handel-mrs. ; (ERB 2001-09-03) For wa, which isn't quite like the other Ps. ; (ERB 2001-10-01) ... but not as different as I thought. Adding ; back constraint that ARG = INDEX. ; (ERB 2001-10-10) In order for the k2y code to recognize PPs as modifiers, ; they need to be ARG0.E.TENSE no_tense. topic-sem-type := mod-id-wg-handel-mrs. ; (MS 2003-07-09) This seems not to be used. ;preposition-sem-type := prep-sem-type & ; [ KEY #key, ; RELS diff-list & [LIST.FIRST #key]]. ; (ERB 2001-10-10) For constructing k2ys it's helpful to have ; the unspecified relation between nouns in an nn compound not ; be an instance of prep_mod_rel. ; (ERB 2001-11-26) This should inherit constraints identifying ; MOD-IND/HAND with INDEX/LTOP. ; (ERB 2001-11-26) It also can't identify its KEY's LBL with ; its LTOP, since KEY.LBL = MOD.MOD-HAND, and we don't want ; LTOP and MOD.MOD-HAND identified. ; (ERB 2001-11-26) Experimenting (as we so desire) with using ; a prep-mod_rel again here to allow compounds with nominalizations ; as the head. ; (ERB 2001-11-26) It seems to work. We'll use a special relation ; type (same as for genitive modifiers) so that the downstream machinery ; recognize this and posit a role. nncompound-sem-type := wg-handel-lex-mrs-basic. ; (ERB 2001-11-08) On the other hand, for n+vn compounds (used ; as vns) it's useful to have it be prep_mod_rel. ; (MS 2003-07-09) This seems not to be used. ;nvncompound-sem-type := mod-id-wg-handel-mrs & ; [KEY #key & prep-mod_rel & ; [ ARG0.E.TENSE no_tense ], ; RELS diff-list & [ LIST.FIRST #key ]]. appos-id-sem-type := mod-id-wg-handel-mrs & [RELS ]. greeting-sem-type := mod-id-wg-handel-mrs & [ HCONS ]. excl-sem-type := mod-id-wg-handel-mrs & [ HCONS ]. ; (ERB 2001-12-10) Making these more like discourse adverbs like ; "well" and "so" in the LinGO grammar. (Existing analysis gives ; them only a LBL and doesn't link that handle up to anything. ; In a sufficiently long sentence, this leads to non-scoping MRSs.) ; (ERB 2001-12-10) This can't be mod-id-wg-handel mrs or even ; wg-handel-lex-mrs because its LTOP isn't its KEY.LBL. ; (ERB 2002-02-14) Try making discourse_rel a real type, not just ; REL name on subord_rel. ; (FCB 2006-06-20) Removed MOD-HAND interj-sem-type := wg-handel-lex-mrs-basic & [ HCONS ]. hes-sem-type := mod-id-wg-handel-mrs & [RELS , HCONS ]. pred-sem-type := wg-handel-lex-mrs-basic. verb-sem-type := pred-sem-type. ;(FCB 2005-04-09) no one seems to be using this ; aux-sem-type := pred-sem-type & ; [MOD-IND #ind, ; MOD-HAND #hand, ; HOOK [INDEX #ind, ; LTOP #hand]]. ; (FCB 2006-06-20) lose the MOD-IND/MOD-HAND; I think we are linking this up elsewhere. simple-verb-sem-type := verb-sem-type. ; & ; [ HOOK [INDEX #ind, ; LTOP #top], ; MOD-IND #ind, ; MOD-HAND #top]. ; (ERB 2001-10-10) In order for the k2y code to recognize adverbs as modifiers, ; they need to be ARG0.E.TENSE no_tense. ; (ERB 2001-10-11) No one was saying the type of the relation for adverbs, ; it was being given as arg_rel & event_rel by inference from the features. ; Fixing this by making it adv_rel. i-sect-adv-sem-type := mod-id-wg-handel-mrs & [ HCONS ]. ; (ERB 2001-09-06) wh adverbs will have more than one element ; on the liszt. So scopal-adv-sem-type can't be RELS 1-dlist. ; I have some more things to add to the sem type for wh adverbs, ; so I'll just a supertype without that constraint... ; (ERB 2001-10-10) In order for the k2y code to recognize adverbs as modifiers, ; they need to be ARG0.E.TENSE no_tense. ; (ERB 2001-10-11) No one was saying the type of the relation for adverbs, ; it was being given as arg_rel & event_rel by inference from the features. ; Fixing this by making it adv_rel. possibly-wh-scopal-adv-sem-type := mod-id-wg-handel-mrs. scopal-adv-sem-type := possibly-wh-scopal-adv-sem-type. ; (ERB 2001-09-06) Here's my attempt at the sem type that gives ; the decompositional semantics for wh adverbs. ; (ERB 2001-09-06) Going on the assumption for now that these ; things are actually intersective modifiers. ; (ERB 2001-10-10) In order for the k2y code to recognize adverbs as modifiers ; they need to be ARG0.E.TENSE no_tense. ; (MS 2001-10-17) That leads to problems for e.g. dou suru. ; (MS 2001-10-18) I took the identification of INDEX.ARG0 and KEY.ARG0 out. ; (ERB 2001-10-17) Like other modifiers, wh adverbs pass up the index ; of the heads they modify. This index should not infact be identified ; with their own event. ; (ERB 2001-12-05) We're not finding dou in role tables, possibly ; because the altkey has no label. Try identifying WLINK of KEY & ; ALTKEY. ; (MS 2004-12-02) For generation, these need PRED values on the prep-mod-relation. wh-adv-sem-type := mod-id-wg-handel-mrs & [ RELS , HCONS ]. ;(MS 2003-07-09) This seems not to be used. ;i-adj-sem-type := pred-sem-type & ; [ KEY #key & adj-relation, ; MOD-IND #ind, ; MOD-HAND #hand, ; HOOK [INDEX #ind, ; LTOP #hand], ; RELS [LIST.FIRST #key]]. ;simple-i-adj-sem-type := i-adj-sem-type & ; [ HOOK.INDEX #ind, ; RELS 1-dlist & [LIST.FIRST.ARG0 #ind], ; HCONS ]. ;pci-adj-sem-type := pred-sem-type & ; [ KEY #key & scope-adj-relation & [SCOPE #sc], ; MOD-IND #ind, ; MOD-HAND #hand, ; HOOK [INDEX #ind, ; LTOP #hand], ;; RELS diff-list & [LIST.FIRST #key], ;; HCONS.LIST ]. ;simple-pci-adj-sem-type := pci-adj-sem-type & ; [HOOK.INDEX #ind, ;; RELS 1-dlist & [LIST.FIRST.ARG0 #ind], ; HCONS ]. ;cond-pci-adj-sem-type := pci-adj-sem-type & ; [ LBL #op-hand, ; LTOP-H #scope, ; KEY.LBL #hand, ; RELS diff-list & [LIST [FIRST [ARG0.E.TENSE tense], ; REST [FIRST subord_rel & [R-HNDL #h2, ; L-HNDL #h1, ; LBL #hand]]]], ; HCONS ] . ; (ERB 2001-08-09) ; (ERB 2001-12-05) Need to enrich this to deal with 半 where the numeral ; classifier is going to add another plus-relation. ; (ERB 2004-1-16) I'm going to be introducing some H-CONS constraints now, because ; the numeral classifier is going to introduce the quantifier for the modified ; noun. ;num-cl-sem-type-super := mod-id-wg-handel-mrs & ; [ HCONS ]. num-cl-sem-type-super := mod-id-wg-handel-mrs. num-cl-sem-type := num-cl-sem-type-super. ; (FCB 2005-10-02) added so that vn2n's can become adverbial PPs reg-nom-arg1-relation := reg-nom-relation & noun-arg1-relation.