;; -*- Mode: TDL; Coding: utf-8 -*- ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ; CONT type nothing := individual. ; (ERB 2001-08-06) Added ALTKEY for number names. I suspect it will be useful ; elsewhere as well. KEY, ALTKEY, and MESSAGE should probably be pulled out ; into a KEYS feature like in the LinGO grammar. ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; CONT values (MRS semantics) ;; ;; The basic semantic objects are all sub-types of thing. Right now, the only ;; type of thing that has an internal structure is index, but that will ;; probably change. For example, an event might have various times associated ;; with them or a handle may include a list of "subhandles" which are in its ;; scope. Indices are classified in the familiar HPSG way. ; (ERB 2001-08-02) Adding type conj_non_expl-ind to help MRS machinery ; deal with coordinated NPs. (MRS machinery doesn't require indices ; of this type to be bound by a quantifier.) Verbs are enforcing ; "full"ness of the indices of their arguments, but this needs to ; be compatible with being a conj_non_expl-ind, so we cross-classify. ; (MS 2003-07-09) Adapted these to matrix. ; (MS 2003-09-25) And again adapted to matrix ver.0.5 full := event-or-ref-index. full-or-zpro-ref-ind := full & index. full_ref-ind := full-or-zpro-ref-ind & ref-ind & [PNG png]. zpro_ref-ind := full-or-zpro-ref-ind. ; (MS 2004-12-13) The conj-ref-ind should be under full_ref-ind ; in order to match the type hierarchy including zero pronouns. ; This overrides the Matrix type. conj-ref-ind := conj-index & full_ref-ind. ;;; Japanese values for PNG png :+ [PN pn, GEN gender]. ; The basic semantic structure, a CONT, is just something with a handle. ; Conts are further classified into mrss (i.e., things with liszts) and ; relations (i.e., things that go on liszts). Conts are divided into psoas ; (with a distinguished event) and nom-objs (with a distinguished index). ; We use a polymorphic attribute name INDEX for both of these, to simplify ; manipulation of these objects; for example, modifying PPs assign as their ; ARG's value the INDEX of the phrase they modify, whether it's an N-bar ; (with a ref-ind value) or a VP (with an event value). Similarly useful ; for coordination. ; ERB (03-10-96) Along with implementing embedded questions, I am ; removing the feature CL-MODE in favor of message (=another name ; for psoa) for the same distinctions. (Plus the use of [MOD none] ; to distinguish and exclude relatives from certain positions.) ; Ultimately, I hope to use the type message for this purpose, but ; until certain changes are made to the internal structure of CONT ; (notably, bundling everything else as opposed to the RELS), this ; isn't possible because the hcomp-nc rule needs to copy this information ; up from head-dtr to mother. So, it will be a feature MESSAGE for now. ; The intermediate types ques-prop-msg and prop-com-msg represent the ; natural classes needed for selection by know, announce, care, check, ; etc. and the analysis of THAT (see lextypes.tdl) and the ; forthcoming analysis of tags on imperatives, respectively. ; Relations are classified according to the types of arguments they take. All ; relations have a handle. In addition, quantifier relations have a bound ; variable and a restriction, nominal relations have an instance, and event ; relations have an event. Furthermore, nominal relations and event relations ; can have additional roles, depending on their meaning. ; Since some temporal nouns can be either arguments or VP modifiers, their ; rel's ARG0 value (coindexed with INDEX of the containing MRS) must remain ; unspecified for type. Consider "Kim arrived Tuesday" vs. "Tuesday comes ; after Monday"; in the first, Tuesday must have an ARG0 index, while in the ; second, it must have a REF-IND index. ; (MS 2003-12-03) no-relation seems to be useful again, as Emily wants to ; use it in nominal-numcl-rule-nomod. no-relation := relation. ;def_rel := quant-relation & [PRED "def"]. ;udef_rel := quant-relation & [PRED "udef"]. ; Nominal relations ; (ERB 2001-11-05) For k2y purposes, making a supertype that ; groups together excl_rel and greeting rel. ; (ERB 2001-12-10) and interj_rel. abstr-excl_rel := relation. ;excl_rel := abstr-excl_rel. ; (ERB 2001-12-10) Let individual interjections provide ; the REL. ;interj_rel := abstr-excl_rel. ;hes_rel := relation. ;greeting_rel := abstr-excl_rel. ; (ERB 2001-10-31) Trying to keep nmod-p の from modifying ; nominalizations, so I need a top level distinction between ; contentful and other nom_rels that will allow me to say ; that の only modifies things whose KEY is contentful-nom_rel. ; (Can't use reg-nom_rel for this, because の can modify things ; that are locable_rel, and I can't make locable_rel a subtype of ; reg-nom_rel; see below.) ; (ERB 2001-10-31) This almost works, but I need a way to distinguish ; between の nominalizations and vns used as ns. For now, I'm ; going to try making a separate type for the vn nominalizations. contentful-nom-relation := noun-relation & [ARG0 full_ref-ind]. ; (ERB 2001-11-05) For k2y purposes, I need all of the nominalise ; relations to have a common supertype. ;basic-nominalise_rel := noun-relation. ;nominalise-relation := noun-relation. ;arg-nominalise_rel := basic-nominalise_rel & arg1-relation. ;nominalise-pred_rel := arg-nominalise_rel. ;nominalise-vn_rel := contentful-nom_rel & arg-nominalise_rel. reg-nom-relation := contentful-nom-relation. ; (ERB 2001-10-03) None of the subtypes of temp_rel are being used, ; so I'm going to comment them out for now, and redo as necessary. ; Likewise for most subtypes of name_rel. ; (ERB 2001-10-03) Reorganizing the type hierarchy for times to ; create types that the PP -> NP rule and に can select for. ; (ERB 2001-10-03) locable_rel is the requirement of the rule ; that makes expressions like 来月 and 2001年 into COP-ARG + ; elements, with an added unspec_loc_rel. ; modable is the requirement of the further rule that ; gives these guys a MOD value, so they can modify without に. ; (ERB 2001-10-03) Experimenting with making locable_rel a subtype ; of reg-nom_rel because dofw will also inherit from name_rel. ; (ERB 2001-10-04) That doesn't work because then any reg-nom_rel ; can be compatible with locable_rel. So, I need to introduce ; the feature NAMED outside of reg-nom_rel. locable-relation := contentful-nom-relation. niarg-relation := locable-relation. not-niarg-relation := locable-relation. modable-relation := locable-relation. nonmodable-relation := locable-relation. ; ERB (2001-08-28) For compositional analysis of dates ; ERB (2001-10-03) Working numbered-np_rel into the new ; hierarchy below locable_rel. numbered-np-relation := nonmodable-relation & niarg-relation & noun-arg1-relation & [ ARG0 full, ARG1 handle]. ; (MS 2004-12-02) For generation, these need PRED values. dofm-relation := numbered-np-relation & [PRED "dofm_rel"]. mofy-relation := numbered-np-relation & [PRED "mofy_rel"]. yofc-relation := numbered-np-relation & [PRED "yofc_rel"]. numbered-hour-relation := numbered-np-relation & [PRED "numbered_hour_rel"]. numbered-minute-relation := numbered-np-relation & [PRED "numbered_minute_rel"]. ; ERB (2001-10-03) For days of the week, which are like ; the numbered-np_rels, but have a name instead of a number. ; (MS 2005-11-30) I want the others to have distinguished relations as well, ; which is used to disambiguate the ni particle attached to them. dofw-relation := const-value-relation & nonmodable-relation & niarg-relation & [ PRED "dofw_rel", CARG *top* ]. loc-relation := named-relation. inst-relation := named-relation. personname-relation := named-relation. productname-relation := named-relation. ; ERB (2001-10-03) Types for other kinds of date expressions. future-date-relation := niarg-relation & modable-relation. non-future-date-relation := not-niarg-relation & modable-relation. ; (ERB 2001-08-06) Changing card_rel to look more like the LinGO ; grammar. Decided to call it integer_rel instead, since it's ; not clear to me that we want to use different numbers for cardinals ; and ordinals. ... but then it doesn't make sense to have ; times_rel and plus_rel as subtypes of integer_rel. So ; there will be integer_rel, const_rel, times_rel and plus_rel. ; Just like the LinGO grammar! ; (ERB 2004-1-16) Another update to integer-relation and its subtypes ; to synch up with the ERG. ;integer-relation :< relation. integer-relation := arg1-ev-relation & [ ARG1 individual ]. ; (ERB 2002-02-14) For current version of k2y code, need const_rel ; to have an event variable. ; (MS 2003-12-10) This is now CARG (RMRS construction and matrix). ; But what should we do about ARG0? ;const_rel := integer_rel & event_rel & ; [ CONST_VALUE *top*, ; PRED 'const, ; ARG0.E.TENSE no_tense ]. ;const-relation := named-relation. const-relation := integer-relation & const-value-relation & [ PRED "card_rel" ]. ;;; FCB 2006-08-18 harmonize with ERG times-relation := integer-relation & arg123-ev-relation & [ PRED "times_rel" ]. plus-relation := integer-relation & arg123-ev-relation & [ PRED "plus_rel" ]. ; (ERB 2001-11-14) To keep the numbers-qua-numcls out of the ; first two nominal numcl rules. quantity-relation := arg1-relation & [ AMOUNT handle ]. reg-quantity-relation := quantity-relation. card-relation := quantity-relation. ; (ERB 2001-11-14) Type for the relation in numbers acting as ; labels for things, as in: ; ; 口座番号は12345です named-card-relation := card-relation & contentful-nom-relation & [PRED 'named_card]. ; (ERB 2001-10-29) I need a value for REL for integers that get ; promoted to NumClPs by the empty-cardinal-rule. I want it to ; be atomic, and to contrast with string. So, here it is: ; (ERB 2001-10-29) The REL names, written 'relation, are actually ; treated as atoms, not strings, by cheap. This means I need to ; move this contrast one level up. ; (ERB 2001-11-01) Switching to a different strategy: Three separate ; numcl2n rules. I still need to distinguish between types of numeral ; classifiers, but now I can do it based on the entire ALTKEY. So, ; I want no_rel to be a kind of mod_relation that contrasts with everything ; else. ; (MS 2003-10-05) Not used any more. ;quantifier_rel := reg-nom_rel & [QUANTITY *top*]. ; (ERB 2001-11-15) Special type for loc-names used to ; leverage Chasen's recognition of place names in YY's ; product. ;named-place_rel := named_rel. ; Pronouns pron-relation := reg-nom-relation. ;zero-pron-relation :< pron-relation. real-pron-relation := pron-relation. generic-nom-relation := pron-relation & [PRED generic-nom_rel]. place-relation := pron-relation & [PRED place_rel]. ;ad_rel :< relation. ; (ERB 2001-10-18) I'm trying to get rid of argmod1_rel. ; so adj_rel will now be a subtype of the new type adj_rel-general. ; (ERB 2002-01-14) The k2y machinery still isn't recognizing ; these. Maybe add TENSE no_tense? (Originally put this on ; adj_rel-general, but that caused problems for na-adjectives ; in prenominal position. Need to think about whether we really ; want to use a subtype of adj_rel-general for those guys, but ; moving the constraint here for now.) ; (MS 2003-09-26) Adapting to Matrix Ver.0.5: There is no ARG any more. adj-relation := arg12-ev-relation & [ ARG0.E.TENSE no_tense ]. na-adj-relation := arg1-ev-relation. ;scoped_rel := relation & ; [ SCOPE handle ]. ;scope_rel :< scoped_rel. ;mood_rel := scope_rel. ;dim_rel := relation & ; [ DARG individual ]. ;ordinal_rel := ad_rel & ; [ ORD integer ]. ;arg_rel := relation & ; [ ARG mrs-thing ]. ; (ERB 2001-10-18) Getting rid of a GLB type. ; (MS 2003-09-26) Adapting to Matrix Ver.0.5: There is no ARG any more. ;arg-event_rel := arg_rel & event_rel. ; (ERB 2001-09-03) I'm trying to get wa to take an event as its ; ARG3, so I need a supertype that doesn't require non_expl-ind. ; I think we might really need some reorganization of the upper ; part of this hierarchy, so that events are non_expl-inds anyway, ; but that wouldn't immediately solve this problem. ; (ERB 2001-09-06) In an MRS meeting with Dan, we decided that ; the topic_rel should be an arg_rel only. That is, no ARG3. The ; value of ARG is either the ARG0 of a noun or the ARG0 of a PP. ; Undoing this arg3_rel-super stuff and reverting to original ; definition above. ; (ERB 2001-10-01) It turns out that making topic_rel arg_rel only ; leads to MRSs that don't scope when wa marks NPs. So, going ; back to making it an ordinary arg_rel, with the features ARG and ; ARG3. The value of ARG3 has to be possibly an event. Reinstating ; arg3_rel-super. ; (ERB 2001-10-03) The wa that attaches to complementizer to needs to be ; able to take a handle, so I will loosen this up from [ ARG3 index ]. ; (ERB 2001-10-10) We need to distinguish the relations belonging to ; prepositions, adverbs and na-adjectives from those belonging to ; verbs and adjectives for the k2y translation. In order to do this, ; I'll need two subtypes of event_rel: v-event_rel and non-v-event_rel. ; arg1, arg2 and arg4 will all be under v-event_rel, but arg and ; arg3_rel-super will just be under v-event_rel. ; (ERB 2001-10-11) For k2y conversion, we need a supertype that includes ; all verbs and no prepositions/modifiers. Since Ps have ARG3, this can't ; in a subtype of v-event_rel. The specialization to ARG3 full for ; Ps will be on argmod3_rel. ; (MS 2003-09-26) Adapting to Matrix Ver.0.5: There is no ARG any more. ;argmod_rel := arg_rel & arg1-relation. ; (ERB 2001-10-18) I can't think of a reason that any relation ; should have both the features ARG and ARG1, so I am removing ; this type. ;modal_rel := arg12-relation & ; scope_rel. ;(MS 2003-09-26) Switched to Matrix unspec_compound_rel ;unspec_rel := arg12-relation & [PRED "unspec_rel"]. ; (ERB 2001-10-10) For time NPs used as adverbs. ; (MS 2003-09-26) Switched to Matrix unspec_compound_rel ;unspec-loc_rel := arg12-relation & [PRED "unspec-loc_rel"]. ;(MS 2001-10-23) if we're not including ZPRO's into the MRS, we should not say too much about the index of ARGs. equal-relation := arg12-ev-relation & [ ARG1 full, ARG2 event-or-ref-index, LBL handle]. ;; used in mai want2-relation := arg12-relation & [PRED "want2_rel"]. ; Preposition relations ; (MS 2003-07-09) This is different to the matrix definition. We have to see, whether it is necessary to keep it. ;j-prep-mod_rel :< arg12-relation. ; (ERB 2001-11-26) We want a special type for no-nmod and ; compounds that the downstream machinery can recognize. ; (MS 2003-09-26) Switched to Matrix prep-mod-relation. ;no_nmod := prep-mod-relation & [PRED 'no]. ;kara_nmod := prep-mod-relation. ;ni_nmod := prep-mod-relation. ;tloc_rel := prep-mod-relation. ; (ERB 2001-09-03) I want the ARG3 of the topic_rel to ; be potentially an event, so I'm changing this type. (see ; further notes above at arg2_rel-super. ; (ERB 2001-09-06) In an MRS meeting with Dan, we decided that ; the topic_rel should be an arg_rel only. That is, no ARG3. The ; value of ARG is either the ARG0 of a noun or the ARG0 of a PP. ; Can't constrain ARG to be an event then. ; (ERB 2001-10-01) It turns out that making topic_rel arg_rel only ; leads to MRSs that don't scope when wa marks NPs. So, going ; back to making it an ordinary arg_rel, with the features ARG and ; ARG3. The value of ARG is the event of the main verb. ; (ERB 2001-10-31) It seems that the reason for keeping the topic_rel ; separate from prep-mod_rel may have gone away, and it would be ; useful for k2y purposes for topic_rel to be a prep-mod_rel. So, ; seeing if that works. (If it does, it may be possible to simplify ; some things in this part of the hierachy and in the relevant ; lexsem types.) ;topic_rel := prep-mod-relation & ; [ ARG2 semarg ]. ; (ERB 2001-10-11) No one was using this type, and adverbs were getting ; the type of their relations from type inference from the features. ; It was a glb type, too. I'm going to make adv-relation a subtype of ; arg_rel & event_rel, and declare it on the adv-sem-type. ; (ERB 2001-10-18) I added arg-event_rel which inherits from ; arg_rel & event_rel and eliminates a glb type. ;modadv-relation := adv-relation & arg1-relation. ; (ERB 2001-10-22) Negation isn't working in k2ys, so I'm going ; to try making it a subtype of adv-relation for the moment, just for fun. ; (ERB 2001-10-24) It wasn't working because it wasn't using this ; rel. Instead, it was an arg_rel with REL 'neg_rel. So, I'm going ; to make nai actually use this rel. ; (ERB 2001-10-25) Changing spelling of neg_rel to neg_rel to ; match the English grammar. ; (MS 2002-12-03) neg_rel needs an arg1 that can be accessed, e.g. in tabenaide morau. ; (MS 2003-07-11) For the Japanese analysis, it also needs an ARG2 ; (MS 2004-02-20) and also an ARG3, in the case of passive? neg-relation := adv-relation & arg1-relation & [ARG0 individual]. ;cant_help_rel :< j-neg_rel. objchange-relation := arg12-relation. tai-relation := objchange-relation & [PRED tai_v_want_rel]. ; (MS 2001-11-22) RARERU needs an ARG1, if it shall be able to combine with TAI. rareru-relation := arg12-ev-relation & [PRED rareru_v_can_rel]. pass-relation := arg12-relation & [PRED 'passive]. adversative-relation := arg12-relation & [PRED 'adversative]. ; Degree relations ;degree_rel :< relation. ; (ERB 2001-08-02) Changed type of C-ARG to conj_non_expl-ind to ; help MRS machinery scope sentences with coordinated NPs. This ; probably means that we'll need a different type of conjunction_rel ; for S coordination. (Or this constraint will need to be lowered.) ; (ERB 2001-08-02) Added features L-LBL and R-LBL so that ; the quantifiers won't be allowed to scope independently. ; (ERB 2001-09-12) Turns out those features were superfluous... the ; grammar shouldn't be imposing those constraints. It really belongs ; in the scoping machinery. ; (ERB 2002-02-14) In case this helps with k2ys. ; (MS 2003-09-25) Adapted this to matrix ver.05 ;discourse-relation := subord-or-conj-relation & [PRED "if-then-rel"]. ;in-order-to_rel := subord-or-conj-relation & [PRED "in-order-to_rel"]. ; for k2ys it is helpful for the unspec rel for noun noun ; compounds to not be a prep-mod-relation. Giving it the same ; features as the English grammar. ; (MS 2003-09-25) there is a new definition in matrix version 0.5 appos-id-relation := arg12-relation & [PRED "appos-id_rel"]. title-id-relation := arg12-relation & [PRED "title-id_rel", ARG0.E.TENSE no_tense]. scopal-adv-relation := arg1-ev-relation & [ARG1 handle]. ;;; FCB 2006-01-26 --- made into arg0-relation so that RMRSs can be built interj-relation := arg0-relation. ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ; predsorts quant_rel := predsort. udef_rel := quant_rel. def_rel := udef_rel. indef_rel := udef_rel. whq_rel := def_rel. hes_rel := predsort. pron_rel := predsort. generic-nom_rel := predsort. place_rel := predsort. quantity_rel := predsort. date_rel := predsort. yofc_rel := date_rel. mofy_rel := date_rel. person_n_rel := predsort. adv_rel := predsort. noun_rel := predsort. no_rel := predsort. tai_v_want_rel := predsort. rareru_v_can_rel := predsort. garu_v_seem_rel := predsort. pass_v_rel := predsort. ; FCB relations for numbers ;card_rel := named_rel. ;plus_rel := card_rel. ;times_rel := card_rel. ;(MS 2005-11-29) these were still not defined. ; Recognized when trying to generate. degree_rel := predsort. unspec_adj_rel := predsort. compound_rel := predsort. unspec_rel := predsort. number_rel := predsort. ; (ERB 2004-1-18) Copied from ERG for measure NP numeral ; classifier phrases. ; (FCB 2005-08-09) changed DARG to ARG1 ; (trying to reduce the types of arguments) ; ARG1 is the attribute being measured ; ARG2 is the unit of measure degree-relation := arg12-relation. ; (CH 2004-09-09) ; introduce vv-relation, which is the supertype of all relations holding between ; V1 and V2 of lexical V-V compound. vv_rel := predsort. vv-manner_rel := vv_rel. vv-relation := arg12-ev-relation & [PRED vv_rel]. ; (CH 2004-09-09) ; introduce vv-prefix-v1-relation. vv-prefix-v1_rel := predsort. vv-prefix-v1-relation := arg1-ev-relation & [PRED vv-prefix-v1_rel]. ;;; ;;; (FCB & DPF 2008-08-06) New types for topic/focus markers ;;; focus_d_rel := predsort. _wa_d_rel := focus_d_rel. _mo_d_rel := focus_d_rel. ;;; (TK 08-08-28) Added following _dake_d_rel := focus_d_rel. _shika_d_rel := focus_d_rel. _bakari_d_rel := focus_d_rel. _nado_d_rel := focus_d_rel. _made_d_rel := focus_d_rel. _koso_d_rel := focus_d_rel. _sura_d_rel := focus_d_rel. _demo_d_rel := focus_d_rel. _nomi_d_rel := focus_d_rel. _sae_d_rel := focus_d_rel. _datte_d_rel := focus_d_rel. _ga_d_rel := focus_d_rel. _colon_d_rel := _wa_d_rel. _comma_d_rel := _wa_d_rel. _equal_d_rel := _wa_d_rel.