;;; -*- Mode: tdl; Coding: utf-8; indent-tabs-mode: nil; -*- ;;; ;;; this file is part of the ERG. it should be kept in the same directory as ;;; the accompanying files `english.set', `global.set', etc. ;;; encoding := utf-8. include "global". ;; ;; suppress results of regular morphology rules if irregular form exists ;; irregular-forms-only. lex-entries-can-fail. ;; ;; names of attributes not to pass from daughter to mother in parsing ;; deleted-daughters := ARGS HD-DTR NH-DTR LCONJ-DTR RCONJ-DTR DTR1 DTR2 DTRA DTRB DTR. ;; ;; names of attributes to filter for packing parser ;; packing-restrictor := RELS HCONS LTOP RNAME. ;; ;; name of type containing quick check structure ;; qc-structure := $qc_paths_set. ;; ;; based on selectional dependencies between lexical entries, reduce the chart ;; right after lexical look-up: `chart-dependencies' is a list of pairs of ;; paths into lexical entries. the type of the node at the end of the first ;; path in one lexical entry makes that entry depend on the existence of some ;; other lexical entry that has that same type as the value of the node at the ;; end of the second path. ;; ;; _fix_me_ ;; not entirely sure, but it must (in principle) be possible to saturate a ;; dependency from lexical and grammar rules. say, a lexical entry selected ;; for something nominalized, and that relation was introduced by a lexical ;; rule, in turn. unless this is the case already, compute static list of all ;; relations introduced by rules (which, presumably, requires another setting ;; to declare how to find constructional semantic contributions; C-CONT) and ;; consider all such dependencies on lexical entries always saturated. ;; (11-oct-02; oe) unidirectional-chart-dependencies. chart-dependencies := "SYNSEM.LKEYS.--+COMPKEY" "SYNSEM.LOCAL.CAT.HEAD.MINORS.MIN" "SYNSEM.LKEYS.--+OCOMPKEY" "SYNSEM.LOCAL.CAT.HEAD.MINORS.MIN" "SYNSEM.LKEYS.--COMPHD" "SYNSEM.LOCAL.CAT.HEAD" "SYNSEM.LOCAL.--+ARGIND" "SYNSEM.--SIND" . ;; ;; as of august 2003, PET includes ECL and the generic (R)MRS code; hence, we ;; need to include the MRS settings from the grammar. ;; postload-lisp-files := "lkb/mrsglobals.lsp". ;; ;; and as of august 2006, the MRS variable property mapping (VPM) can now be ;; activated as a configuration parameter. (26-aug-06; oe) ;; vpm := "semi". ;;; ;;; following are a number of settings for the new (as of late 2008) token ;;; mapping and lexical filtering support in PET. ;;; ;; ;; first, the general format of chart mapping rules, much like MRS transfer. ;; chart-mapping-context-path := "+CONTEXT". chart-mapping-input-path := "+INPUT". chart-mapping-output-path := "+OUTPUT". chart-mapping-position-path := "+POSITION". ;; ;; in lexical instatiation, the list of tokens activating a lexical entry (be ;; it native or generic) are unified into the lexical entry under this path. ;; furthermore, the parser will also give us a `pointer' to the last token. ;; lexicon-tokens-path := "TOKENS.+LIST". lexicon-last-token-path := "TOKENS.+LAST". ;; ;; furthermore, for the various input formats, we need to declare how parts of ;; input descriptions correspond to the grammar-internal feature geometry; in ;; the YY input format, for example, token feature structures (aka input items ;; PET-internally) are created from various parts of the token description. ;; token-form-path := "+FORM". ; [required] string for lexical lookup token-id-path := "+ID". ; [optional] list of external ids token-from-path := "+FROM". ; [optional] surface start position token-to-path := "+TO". ; [optional] surface end position token-postags-path := "+TNT.+TAGS". ; [optional] list of POS tags token-posprobs-path := "+TNT.+PRBS". ; [optional] list of POS probabilities ;; ;; finally, declare TDL status values for the various new entity types ;; generic-lexentry-status-values := generic-lex-entry. token-mapping-rule-status-values := token-mapping-rule. lexical-filtering-rule-status-values := lexical-filtering-rule. ;; ;; as of mid-2011, PET now includes native support for the Regular Expression ;; Pre-Processor (REPP) ;; include "repp". ;; ;; optionally (as of mid-2011), PET can call out to a PoS tagger to augment ;; the token sequence that will be input to parsing. the following are the ;; default settings for use of TnT. note that the external command will be ;; invoked through a shell, hence environment variables will be expanded. ;; ;; the specifics of interfacing to external PoS tagger(s) (and supertaggers, ;; NE recognizers, et al.) are still work in progress. it seems desirable to ;; support pipelining multiple such preprocessors, but then it still needs to ;; be determined how to integrate information coming from multiple sources. ;; one approach would be to just associate multiple lists of ;; pairs with each input item PET-internally, and then define one path per ;; preprocessor to project these lists into token feature structures; on this ;; view, integration of information would be left to token mapping. there is ;; a certain attraction, on the other hand, in assuming there is one single ;; (unified) opinion on PoS tags, supertags, and NE labels. on this alternate ;; view, 'stacking' of taggers (as we found useful with GENIA+TnT) would need ;; to be accomplished PET-internally, i.e. the rules of how to integrate the ;; opinions of multiple taggers would need to go into the configuration. ;; taggers := tnt. tnt-command := "${LOGONROOT}/bin/tnt". tnt-arguments := "-v0 -z100 ${LOGONROOT}/coli/tnt/models/wsj -". ;; ;; it appears that TnT (somewhat to our surprise) takes into account context ;; from preceding sentences, when applied to a sequence of inputs. this might ;; well be a good idea, in principle, but in DELPH-IN there is a very strong ;; assumption of processing being sentence-centered. when running under ;; [incr tsdb()] control, for example, with multiple processing clients each ;; instance will see non-consecutive sequences of sentences. hence, provide ;; a means of injecting unambiguous sentence break tokens (e.g. and