grammar_police 0.1.0
Sign up to get free protection for your applications and to get access to all the features.
- data/.DS_Store +0 -0
- data/.gitignore +4 -0
- data/Gemfile +4 -0
- data/Rakefile +2 -0
- data/c/.DS_Store +0 -0
- data/c/link-grammar.c +65 -0
- data/c/link-grammar.h +60 -0
- data/c/link-grammar.o +0 -0
- data/c/link-grammar.so +0 -0
- data/c/link-grammar/.DS_Store +0 -0
- data/c/link-grammar/.deps/analyze-linkage.Plo +198 -0
- data/c/link-grammar/.deps/and.Plo +202 -0
- data/c/link-grammar/.deps/api.Plo +244 -0
- data/c/link-grammar/.deps/build-disjuncts.Plo +212 -0
- data/c/link-grammar/.deps/command-line.Plo +201 -0
- data/c/link-grammar/.deps/constituents.Plo +201 -0
- data/c/link-grammar/.deps/count.Plo +202 -0
- data/c/link-grammar/.deps/disjunct-utils.Plo +126 -0
- data/c/link-grammar/.deps/disjuncts.Plo +123 -0
- data/c/link-grammar/.deps/error.Plo +121 -0
- data/c/link-grammar/.deps/expand.Plo +133 -0
- data/c/link-grammar/.deps/extract-links.Plo +198 -0
- data/c/link-grammar/.deps/fast-match.Plo +200 -0
- data/c/link-grammar/.deps/idiom.Plo +200 -0
- data/c/link-grammar/.deps/jni-client.Plo +217 -0
- data/c/link-grammar/.deps/link-parser.Po +1 -0
- data/c/link-grammar/.deps/massage.Plo +202 -0
- data/c/link-grammar/.deps/post-process.Plo +202 -0
- data/c/link-grammar/.deps/pp_knowledge.Plo +202 -0
- data/c/link-grammar/.deps/pp_lexer.Plo +201 -0
- data/c/link-grammar/.deps/pp_linkset.Plo +200 -0
- data/c/link-grammar/.deps/prefix.Plo +102 -0
- data/c/link-grammar/.deps/preparation.Plo +202 -0
- data/c/link-grammar/.deps/print-util.Plo +200 -0
- data/c/link-grammar/.deps/print.Plo +201 -0
- data/c/link-grammar/.deps/prune.Plo +202 -0
- data/c/link-grammar/.deps/read-dict.Plo +223 -0
- data/c/link-grammar/.deps/read-regex.Plo +123 -0
- data/c/link-grammar/.deps/regex-morph.Plo +131 -0
- data/c/link-grammar/.deps/resources.Plo +203 -0
- data/c/link-grammar/.deps/spellcheck-aspell.Plo +1 -0
- data/c/link-grammar/.deps/spellcheck-hun.Plo +115 -0
- data/c/link-grammar/.deps/string-set.Plo +198 -0
- data/c/link-grammar/.deps/tokenize.Plo +160 -0
- data/c/link-grammar/.deps/utilities.Plo +222 -0
- data/c/link-grammar/.deps/word-file.Plo +201 -0
- data/c/link-grammar/.deps/word-utils.Plo +212 -0
- data/c/link-grammar/.libs/analyze-linkage.o +0 -0
- data/c/link-grammar/.libs/and.o +0 -0
- data/c/link-grammar/.libs/api.o +0 -0
- data/c/link-grammar/.libs/build-disjuncts.o +0 -0
- data/c/link-grammar/.libs/command-line.o +0 -0
- data/c/link-grammar/.libs/constituents.o +0 -0
- data/c/link-grammar/.libs/count.o +0 -0
- data/c/link-grammar/.libs/disjunct-utils.o +0 -0
- data/c/link-grammar/.libs/disjuncts.o +0 -0
- data/c/link-grammar/.libs/error.o +0 -0
- data/c/link-grammar/.libs/expand.o +0 -0
- data/c/link-grammar/.libs/extract-links.o +0 -0
- data/c/link-grammar/.libs/fast-match.o +0 -0
- data/c/link-grammar/.libs/idiom.o +0 -0
- data/c/link-grammar/.libs/jni-client.o +0 -0
- data/c/link-grammar/.libs/liblink-grammar-java-symbols.expsym +31 -0
- data/c/link-grammar/.libs/liblink-grammar-java.4.dylib +0 -0
- data/c/link-grammar/.libs/liblink-grammar-java.4.dylib.dSYM/Contents/Info.plist +20 -0
- data/c/link-grammar/.libs/liblink-grammar-java.4.dylib.dSYM/Contents/Resources/DWARF/liblink-grammar-java.4.dylib +0 -0
- data/c/link-grammar/.libs/liblink-grammar-java.a +0 -0
- data/c/link-grammar/.libs/liblink-grammar-java.dylib +0 -0
- data/c/link-grammar/.libs/liblink-grammar-symbols.expsym +194 -0
- data/c/link-grammar/.libs/liblink-grammar.4.dylib +0 -0
- data/c/link-grammar/.libs/liblink-grammar.4.dylib.dSYM/Contents/Info.plist +20 -0
- data/c/link-grammar/.libs/liblink-grammar.4.dylib.dSYM/Contents/Resources/DWARF/liblink-grammar.4.dylib +0 -0
- data/c/link-grammar/.libs/liblink-grammar.a +0 -0
- data/c/link-grammar/.libs/liblink-grammar.dylib +0 -0
- data/c/link-grammar/.libs/liblink-grammar.la +41 -0
- data/c/link-grammar/.libs/liblink-grammar.lai +41 -0
- data/c/link-grammar/.libs/massage.o +0 -0
- data/c/link-grammar/.libs/post-process.o +0 -0
- data/c/link-grammar/.libs/pp_knowledge.o +0 -0
- data/c/link-grammar/.libs/pp_lexer.o +0 -0
- data/c/link-grammar/.libs/pp_linkset.o +0 -0
- data/c/link-grammar/.libs/prefix.o +0 -0
- data/c/link-grammar/.libs/preparation.o +0 -0
- data/c/link-grammar/.libs/print-util.o +0 -0
- data/c/link-grammar/.libs/print.o +0 -0
- data/c/link-grammar/.libs/prune.o +0 -0
- data/c/link-grammar/.libs/read-dict.o +0 -0
- data/c/link-grammar/.libs/read-regex.o +0 -0
- data/c/link-grammar/.libs/regex-morph.o +0 -0
- data/c/link-grammar/.libs/resources.o +0 -0
- data/c/link-grammar/.libs/spellcheck-aspell.o +0 -0
- data/c/link-grammar/.libs/spellcheck-hun.o +0 -0
- data/c/link-grammar/.libs/string-set.o +0 -0
- data/c/link-grammar/.libs/tokenize.o +0 -0
- data/c/link-grammar/.libs/utilities.o +0 -0
- data/c/link-grammar/.libs/word-file.o +0 -0
- data/c/link-grammar/.libs/word-utils.o +0 -0
- data/c/link-grammar/Makefile +900 -0
- data/c/link-grammar/Makefile.am +202 -0
- data/c/link-grammar/Makefile.in +900 -0
- data/c/link-grammar/analyze-linkage.c +1317 -0
- data/c/link-grammar/analyze-linkage.h +24 -0
- data/c/link-grammar/and.c +1603 -0
- data/c/link-grammar/and.h +27 -0
- data/c/link-grammar/api-structures.h +362 -0
- data/c/link-grammar/api-types.h +72 -0
- data/c/link-grammar/api.c +1887 -0
- data/c/link-grammar/api.h +96 -0
- data/c/link-grammar/autoit/.DS_Store +0 -0
- data/c/link-grammar/autoit/README +10 -0
- data/c/link-grammar/autoit/_LGTest.au3 +22 -0
- data/c/link-grammar/autoit/_LinkGrammar.au3 +545 -0
- data/c/link-grammar/build-disjuncts.c +487 -0
- data/c/link-grammar/build-disjuncts.h +21 -0
- data/c/link-grammar/command-line.c +458 -0
- data/c/link-grammar/command-line.h +15 -0
- data/c/link-grammar/constituents.c +1836 -0
- data/c/link-grammar/constituents.h +26 -0
- data/c/link-grammar/corpus/.DS_Store +0 -0
- data/c/link-grammar/corpus/.deps/cluster.Plo +1 -0
- data/c/link-grammar/corpus/.deps/corpus.Plo +1 -0
- data/c/link-grammar/corpus/Makefile +527 -0
- data/c/link-grammar/corpus/Makefile.am +46 -0
- data/c/link-grammar/corpus/Makefile.in +527 -0
- data/c/link-grammar/corpus/README +17 -0
- data/c/link-grammar/corpus/cluster.c +286 -0
- data/c/link-grammar/corpus/cluster.h +32 -0
- data/c/link-grammar/corpus/corpus.c +483 -0
- data/c/link-grammar/corpus/corpus.h +46 -0
- data/c/link-grammar/count.c +828 -0
- data/c/link-grammar/count.h +25 -0
- data/c/link-grammar/disjunct-utils.c +261 -0
- data/c/link-grammar/disjunct-utils.h +27 -0
- data/c/link-grammar/disjuncts.c +138 -0
- data/c/link-grammar/disjuncts.h +13 -0
- data/c/link-grammar/error.c +92 -0
- data/c/link-grammar/error.h +35 -0
- data/c/link-grammar/expand.c +67 -0
- data/c/link-grammar/expand.h +13 -0
- data/c/link-grammar/externs.h +22 -0
- data/c/link-grammar/extract-links.c +625 -0
- data/c/link-grammar/extract-links.h +16 -0
- data/c/link-grammar/fast-match.c +309 -0
- data/c/link-grammar/fast-match.h +17 -0
- data/c/link-grammar/idiom.c +373 -0
- data/c/link-grammar/idiom.h +15 -0
- data/c/link-grammar/jni-client.c +779 -0
- data/c/link-grammar/jni-client.h +236 -0
- data/c/link-grammar/liblink-grammar-java.la +42 -0
- data/c/link-grammar/liblink-grammar.la +41 -0
- data/c/link-grammar/link-features.h +37 -0
- data/c/link-grammar/link-features.h.in +37 -0
- data/c/link-grammar/link-grammar-java.def +31 -0
- data/c/link-grammar/link-grammar.def +194 -0
- data/c/link-grammar/link-includes.h +465 -0
- data/c/link-grammar/link-parser.c +849 -0
- data/c/link-grammar/massage.c +329 -0
- data/c/link-grammar/massage.h +13 -0
- data/c/link-grammar/post-process.c +1113 -0
- data/c/link-grammar/post-process.h +45 -0
- data/c/link-grammar/pp_knowledge.c +376 -0
- data/c/link-grammar/pp_knowledge.h +14 -0
- data/c/link-grammar/pp_lexer.c +1920 -0
- data/c/link-grammar/pp_lexer.h +19 -0
- data/c/link-grammar/pp_linkset.c +158 -0
- data/c/link-grammar/pp_linkset.h +20 -0
- data/c/link-grammar/prefix.c +482 -0
- data/c/link-grammar/prefix.h +139 -0
- data/c/link-grammar/preparation.c +412 -0
- data/c/link-grammar/preparation.h +20 -0
- data/c/link-grammar/print-util.c +87 -0
- data/c/link-grammar/print-util.h +32 -0
- data/c/link-grammar/print.c +1085 -0
- data/c/link-grammar/print.h +16 -0
- data/c/link-grammar/prune.c +1864 -0
- data/c/link-grammar/prune.h +17 -0
- data/c/link-grammar/read-dict.c +1785 -0
- data/c/link-grammar/read-dict.h +29 -0
- data/c/link-grammar/read-regex.c +161 -0
- data/c/link-grammar/read-regex.h +12 -0
- data/c/link-grammar/regex-morph.c +126 -0
- data/c/link-grammar/regex-morph.h +17 -0
- data/c/link-grammar/resources.c +180 -0
- data/c/link-grammar/resources.h +23 -0
- data/c/link-grammar/sat-solver/.DS_Store +0 -0
- data/c/link-grammar/sat-solver/.deps/fast-sprintf.Plo +1 -0
- data/c/link-grammar/sat-solver/.deps/sat-encoder.Plo +1 -0
- data/c/link-grammar/sat-solver/.deps/util.Plo +1 -0
- data/c/link-grammar/sat-solver/.deps/variables.Plo +1 -0
- data/c/link-grammar/sat-solver/.deps/word-tag.Plo +1 -0
- data/c/link-grammar/sat-solver/Makefile +527 -0
- data/c/link-grammar/sat-solver/Makefile.am +29 -0
- data/c/link-grammar/sat-solver/Makefile.in +527 -0
- data/c/link-grammar/sat-solver/clock.hpp +33 -0
- data/c/link-grammar/sat-solver/fast-sprintf.cpp +26 -0
- data/c/link-grammar/sat-solver/fast-sprintf.hpp +7 -0
- data/c/link-grammar/sat-solver/guiding.hpp +244 -0
- data/c/link-grammar/sat-solver/matrix-ut.hpp +79 -0
- data/c/link-grammar/sat-solver/sat-encoder.cpp +2811 -0
- data/c/link-grammar/sat-solver/sat-encoder.h +11 -0
- data/c/link-grammar/sat-solver/sat-encoder.hpp +381 -0
- data/c/link-grammar/sat-solver/trie.hpp +118 -0
- data/c/link-grammar/sat-solver/util.cpp +23 -0
- data/c/link-grammar/sat-solver/util.hpp +14 -0
- data/c/link-grammar/sat-solver/variables.cpp +5 -0
- data/c/link-grammar/sat-solver/variables.hpp +829 -0
- data/c/link-grammar/sat-solver/word-tag.cpp +159 -0
- data/c/link-grammar/sat-solver/word-tag.hpp +162 -0
- data/c/link-grammar/spellcheck-aspell.c +148 -0
- data/c/link-grammar/spellcheck-hun.c +136 -0
- data/c/link-grammar/spellcheck.h +34 -0
- data/c/link-grammar/string-set.c +169 -0
- data/c/link-grammar/string-set.h +16 -0
- data/c/link-grammar/structures.h +498 -0
- data/c/link-grammar/tokenize.c +1049 -0
- data/c/link-grammar/tokenize.h +15 -0
- data/c/link-grammar/utilities.c +847 -0
- data/c/link-grammar/utilities.h +281 -0
- data/c/link-grammar/word-file.c +124 -0
- data/c/link-grammar/word-file.h +15 -0
- data/c/link-grammar/word-utils.c +526 -0
- data/c/link-grammar/word-utils.h +152 -0
- data/data/.DS_Store +0 -0
- data/data/Makefile +511 -0
- data/data/Makefile.am +4 -0
- data/data/Makefile.in +511 -0
- data/data/de/.DS_Store +0 -0
- data/data/de/4.0.affix +7 -0
- data/data/de/4.0.dict +474 -0
- data/data/de/Makefile +387 -0
- data/data/de/Makefile.am +9 -0
- data/data/de/Makefile.in +387 -0
- data/data/en/.DS_Store +0 -0
- data/data/en/4.0.affix +26 -0
- data/data/en/4.0.batch +1002 -0
- data/data/en/4.0.biolg.batch +411 -0
- data/data/en/4.0.constituent-knowledge +127 -0
- data/data/en/4.0.dict +8759 -0
- data/data/en/4.0.dict.m4 +6928 -0
- data/data/en/4.0.enwiki.batch +14 -0
- data/data/en/4.0.fixes.batch +2776 -0
- data/data/en/4.0.knowledge +306 -0
- data/data/en/4.0.regex +225 -0
- data/data/en/4.0.voa.batch +114 -0
- data/data/en/Makefile +554 -0
- data/data/en/Makefile.am +19 -0
- data/data/en/Makefile.in +554 -0
- data/data/en/README +173 -0
- data/data/en/tiny.dict +157 -0
- data/data/en/words/.DS_Store +0 -0
- data/data/en/words/Makefile +456 -0
- data/data/en/words/Makefile.am +78 -0
- data/data/en/words/Makefile.in +456 -0
- data/data/en/words/currency +205 -0
- data/data/en/words/currency.p +28 -0
- data/data/en/words/entities.given-bisex.sing +39 -0
- data/data/en/words/entities.given-female.sing +4141 -0
- data/data/en/words/entities.given-male.sing +1633 -0
- data/data/en/words/entities.locations.sing +68 -0
- data/data/en/words/entities.national.sing +253 -0
- data/data/en/words/entities.organizations.sing +7 -0
- data/data/en/words/entities.us-states.sing +11 -0
- data/data/en/words/units.1 +45 -0
- data/data/en/words/units.1.dot +4 -0
- data/data/en/words/units.3 +2 -0
- data/data/en/words/units.4 +5 -0
- data/data/en/words/units.4.dot +1 -0
- data/data/en/words/words-medical.adv.1 +1191 -0
- data/data/en/words/words-medical.prep.1 +67 -0
- data/data/en/words/words-medical.v.4.1 +2835 -0
- data/data/en/words/words-medical.v.4.2 +2848 -0
- data/data/en/words/words-medical.v.4.3 +3011 -0
- data/data/en/words/words-medical.v.4.4 +3036 -0
- data/data/en/words/words-medical.v.4.5 +3050 -0
- data/data/en/words/words.adj.1 +6794 -0
- data/data/en/words/words.adj.2 +638 -0
- data/data/en/words/words.adj.3 +667 -0
- data/data/en/words/words.adv.1 +1573 -0
- data/data/en/words/words.adv.2 +67 -0
- data/data/en/words/words.adv.3 +157 -0
- data/data/en/words/words.adv.4 +80 -0
- data/data/en/words/words.n.1 +11464 -0
- data/data/en/words/words.n.1.wiki +264 -0
- data/data/en/words/words.n.2.s +2017 -0
- data/data/en/words/words.n.2.s.biolg +1 -0
- data/data/en/words/words.n.2.s.wiki +298 -0
- data/data/en/words/words.n.2.x +65 -0
- data/data/en/words/words.n.2.x.wiki +10 -0
- data/data/en/words/words.n.3 +5717 -0
- data/data/en/words/words.n.t +23 -0
- data/data/en/words/words.v.1.1 +1038 -0
- data/data/en/words/words.v.1.2 +1043 -0
- data/data/en/words/words.v.1.3 +1052 -0
- data/data/en/words/words.v.1.4 +1023 -0
- data/data/en/words/words.v.1.p +17 -0
- data/data/en/words/words.v.10.1 +14 -0
- data/data/en/words/words.v.10.2 +15 -0
- data/data/en/words/words.v.10.3 +88 -0
- data/data/en/words/words.v.10.4 +17 -0
- data/data/en/words/words.v.2.1 +1253 -0
- data/data/en/words/words.v.2.2 +1304 -0
- data/data/en/words/words.v.2.3 +1280 -0
- data/data/en/words/words.v.2.4 +1285 -0
- data/data/en/words/words.v.2.5 +1287 -0
- data/data/en/words/words.v.4.1 +2472 -0
- data/data/en/words/words.v.4.2 +2487 -0
- data/data/en/words/words.v.4.3 +2441 -0
- data/data/en/words/words.v.4.4 +2478 -0
- data/data/en/words/words.v.4.5 +2483 -0
- data/data/en/words/words.v.5.1 +98 -0
- data/data/en/words/words.v.5.2 +98 -0
- data/data/en/words/words.v.5.3 +103 -0
- data/data/en/words/words.v.5.4 +102 -0
- data/data/en/words/words.v.6.1 +388 -0
- data/data/en/words/words.v.6.2 +401 -0
- data/data/en/words/words.v.6.3 +397 -0
- data/data/en/words/words.v.6.4 +405 -0
- data/data/en/words/words.v.6.5 +401 -0
- data/data/en/words/words.v.8.1 +117 -0
- data/data/en/words/words.v.8.2 +118 -0
- data/data/en/words/words.v.8.3 +118 -0
- data/data/en/words/words.v.8.4 +119 -0
- data/data/en/words/words.v.8.5 +119 -0
- data/data/en/words/words.y +104 -0
- data/data/lt/.DS_Store +0 -0
- data/data/lt/4.0.affix +6 -0
- data/data/lt/4.0.constituent-knowledge +24 -0
- data/data/lt/4.0.dict +135 -0
- data/data/lt/4.0.knowledge +38 -0
- data/data/lt/Makefile +389 -0
- data/data/lt/Makefile.am +11 -0
- data/data/lt/Makefile.in +389 -0
- data/grammar_police.gemspec +23 -0
- data/lib/.DS_Store +0 -0
- data/lib/grammar_police.rb +11 -0
- data/lib/grammar_police/.DS_Store +0 -0
- data/lib/grammar_police/dictionary.rb +30 -0
- data/lib/grammar_police/linkage.rb +26 -0
- data/lib/grammar_police/parse_options.rb +32 -0
- data/lib/grammar_police/sentence.rb +44 -0
- data/lib/grammar_police/version.rb +3 -0
- data/tests/.DS_Store +0 -0
- data/tests/count_linkages.rb +29 -0
- data/tests/sentences.txt +86 -0
- metadata +408 -0
@@ -0,0 +1,15 @@
|
|
1
|
+
/*************************************************************************/
|
2
|
+
/* Copyright (c) 2004 */
|
3
|
+
/* Daniel Sleator, David Temperley, and John Lafferty */
|
4
|
+
/* All rights reserved */
|
5
|
+
/* */
|
6
|
+
/* Use of the link grammar parsing system is subject to the terms of the */
|
7
|
+
/* license set forth in the LICENSE file included with this software, */
|
8
|
+
/* and also available at http://www.link.cs.cmu.edu/link/license.html */
|
9
|
+
/* This license allows free redistribution and use in source and binary */
|
10
|
+
/* forms, with or without modification, subject to certain conditions. */
|
11
|
+
/* */
|
12
|
+
/*************************************************************************/
|
13
|
+
|
14
|
+
|
15
|
+
|
@@ -0,0 +1,1836 @@
|
|
1
|
+
/*************************************************************************/
|
2
|
+
/* Copyright (c) 2004 */
|
3
|
+
/* Daniel Sleator, David Temperley, and John Lafferty */
|
4
|
+
/* All rights reserved */
|
5
|
+
/* */
|
6
|
+
/* Use of the link grammar parsing system is subject to the terms of the */
|
7
|
+
/* license set forth in the LICENSE file included with this software, */
|
8
|
+
/* and also available at http://www.link.cs.cmu.edu/link/license.html */
|
9
|
+
/* This license allows free redistribution and use in source and binary */
|
10
|
+
/* forms, with or without modification, subject to certain conditions. */
|
11
|
+
/* */
|
12
|
+
/*************************************************************************/
|
13
|
+
|
14
|
+
#include <stdarg.h>
|
15
|
+
#include <string.h>
|
16
|
+
#include <link-grammar/api.h>
|
17
|
+
#include "error.h"
|
18
|
+
#include "constituents.h"
|
19
|
+
|
20
|
+
#define MAXCONSTITUENTS 8192
|
21
|
+
#define MAXSUBL 16
|
22
|
+
#define OPEN_BRACKET '['
|
23
|
+
#define CLOSE_BRACKET ']'
|
24
|
+
|
25
|
+
typedef enum {OPEN_TOK, CLOSE_TOK, WORD_TOK} CType;
|
26
|
+
typedef enum {NONE, STYPE, PTYPE, QTYPE, QDTYPE} WType;
|
27
|
+
|
28
|
+
typedef struct
|
29
|
+
{
|
30
|
+
int left;
|
31
|
+
int right;
|
32
|
+
const char * type;
|
33
|
+
char domain_type;
|
34
|
+
const char * start_link;
|
35
|
+
int start_num;
|
36
|
+
int subl;
|
37
|
+
int canon;
|
38
|
+
int valid;
|
39
|
+
#ifdef AUX_CODE_IS_DEAD
|
40
|
+
/* The only code that actually sets aux to a non-zero value is code
|
41
|
+
* followed by code that zets it to zero. -- its dead code, and so
|
42
|
+
* aux is never actually used. Comment this code out.
|
43
|
+
*/
|
44
|
+
int aux;
|
45
|
+
/* 0: it's an ordinary VP (or other type);
|
46
|
+
* 1: it's an AUX, don't print it;
|
47
|
+
* 2: it's an AUX, and print it
|
48
|
+
*/
|
49
|
+
#endif /* AUX_CODE_IS_DEAD */
|
50
|
+
} constituent_t;
|
51
|
+
|
52
|
+
/* XXX it seems like the old code worked fine with MAX_ELTS=10 */
|
53
|
+
#define MAX_ELTS 100
|
54
|
+
typedef struct
|
55
|
+
{
|
56
|
+
int num;
|
57
|
+
int e[MAX_ELTS];
|
58
|
+
int valid;
|
59
|
+
} andlist_t;
|
60
|
+
|
61
|
+
/*
|
62
|
+
* Context used to store assorted intermediate data
|
63
|
+
* when the constituent string is being generated.
|
64
|
+
*/
|
65
|
+
#define MAX_ANDS 1024
|
66
|
+
typedef struct
|
67
|
+
{
|
68
|
+
String_set * phrase_ss;
|
69
|
+
WType wordtype[MAX_SENTENCE];
|
70
|
+
int word_used[MAXSUBL][MAX_SENTENCE];
|
71
|
+
int templist[MAX_ELTS];
|
72
|
+
constituent_t constituent[MAXCONSTITUENTS];
|
73
|
+
andlist_t andlist[MAX_ANDS];
|
74
|
+
} con_context_t;
|
75
|
+
|
76
|
+
/* ================================================================ */
|
77
|
+
|
78
|
+
static inline int uppercompare(const char * s, const char * t)
|
79
|
+
{
|
80
|
+
return (FALSE == utf8_upper_match(s,t));
|
81
|
+
}
|
82
|
+
|
83
|
+
/**
|
84
|
+
* If a constituent c has a comma at either end, we exclude the
|
85
|
+
* comma. (We continue to shift the boundary until we get to
|
86
|
+
* something inside the current sublinkage)
|
87
|
+
*/
|
88
|
+
static void adjust_for_left_comma(con_context_t * ctxt, Linkage linkage, int c)
|
89
|
+
{
|
90
|
+
int w;
|
91
|
+
w = ctxt->constituent[c].left;
|
92
|
+
if (strcmp(linkage->word[w], ",") == 0)
|
93
|
+
{
|
94
|
+
w++;
|
95
|
+
while (1) {
|
96
|
+
if (ctxt->word_used[linkage->current][w] == 1) break;
|
97
|
+
w++;
|
98
|
+
}
|
99
|
+
}
|
100
|
+
ctxt->constituent[c].left = w;
|
101
|
+
}
|
102
|
+
|
103
|
+
static void adjust_for_right_comma(con_context_t *ctxt, Linkage linkage, int c)
|
104
|
+
{
|
105
|
+
int w;
|
106
|
+
w = ctxt->constituent[c].right;
|
107
|
+
if ((strcmp(linkage->word[w], ",") == 0) ||
|
108
|
+
(strcmp(linkage->word[w], "RIGHT-WALL") == 0))
|
109
|
+
{
|
110
|
+
w--;
|
111
|
+
while (1)
|
112
|
+
{
|
113
|
+
if (ctxt->word_used[linkage->current][w]==1) break;
|
114
|
+
w--;
|
115
|
+
}
|
116
|
+
}
|
117
|
+
ctxt->constituent[c].right = w;
|
118
|
+
}
|
119
|
+
|
120
|
+
static void print_constituent(con_context_t *ctxt, Linkage linkage, int c)
|
121
|
+
{
|
122
|
+
int w;
|
123
|
+
if (verbosity < 2) return;
|
124
|
+
|
125
|
+
printf(" c %2d %4s [%c] (%2d-%2d): ",
|
126
|
+
c, ctxt->constituent[c].type, ctxt->constituent[c].domain_type,
|
127
|
+
ctxt->constituent[c].left, ctxt->constituent[c].right);
|
128
|
+
for (w = ctxt->constituent[c].left; w <= ctxt->constituent[c].right; w++) {
|
129
|
+
printf("%s ", linkage->word[w]); /**PV**/
|
130
|
+
}
|
131
|
+
printf("\n");
|
132
|
+
}
|
133
|
+
|
134
|
+
/******************************************************
|
135
|
+
* These functions do the bulk of the actual
|
136
|
+
* constituent-generating; they're called once for each
|
137
|
+
* sublinkage
|
138
|
+
*********************************************************/
|
139
|
+
|
140
|
+
/**
|
141
|
+
* This function looks for constituents of type ctype1. Say it finds
|
142
|
+
* one, call it c1. It searches for the next larger constituent of
|
143
|
+
* type ctype2, call it c2. It then generates a new constituent of
|
144
|
+
* ctype3, containing all the words in c2 but not c1.
|
145
|
+
*/
|
146
|
+
static int gen_comp(con_context_t *ctxt, Linkage linkage,
|
147
|
+
int numcon_total, int numcon_subl,
|
148
|
+
const char * ctype1, const char * ctype2,
|
149
|
+
const char * ctype3, int x)
|
150
|
+
{
|
151
|
+
int w, w2, w3, c, c1, c2, done;
|
152
|
+
c = numcon_total + numcon_subl;
|
153
|
+
|
154
|
+
for (c1=numcon_total; c1<numcon_total + numcon_subl; c1++)
|
155
|
+
{
|
156
|
+
/* If ctype1 is NP, it has to be an appositive to continue */
|
157
|
+
if ((x==4) && (post_process_match("MX#*", ctxt->constituent[c1].start_link)==0))
|
158
|
+
continue;
|
159
|
+
|
160
|
+
/* If ctype1 is X, and domain_type is t, it's an infinitive - skip it */
|
161
|
+
if ((x==2) && (ctxt->constituent[c1].domain_type=='t'))
|
162
|
+
continue;
|
163
|
+
|
164
|
+
/* If it's domain-type z, it's a subject-relative clause;
|
165
|
+
the VP doesn't need an NP */
|
166
|
+
if (ctxt->constituent[c1].domain_type=='z')
|
167
|
+
continue;
|
168
|
+
|
169
|
+
/* If ctype1 is X or VP, and it's not started by an S, don't generate an NP
|
170
|
+
(Neither of the two previous checks are necessary now, right?) */
|
171
|
+
if ((x==1 || x==2) &&
|
172
|
+
(((post_process_match("S", ctxt->constituent[c1].start_link) == 0) &&
|
173
|
+
(post_process_match("SX", ctxt->constituent[c1].start_link) == 0) &&
|
174
|
+
(post_process_match("SF", ctxt->constituent[c1].start_link) == 0)) ||
|
175
|
+
(post_process_match("S##w", ctxt->constituent[c1].start_link) != 0)))
|
176
|
+
continue;
|
177
|
+
|
178
|
+
/* If it's an SBAR (relative clause case), it has to be a relative clause */
|
179
|
+
if ((x==3) &&
|
180
|
+
((post_process_match("Rn", ctxt->constituent[c1].start_link) == 0) &&
|
181
|
+
(post_process_match("R*", ctxt->constituent[c1].start_link) == 0) &&
|
182
|
+
(post_process_match("MX#r", ctxt->constituent[c1].start_link) == 0) &&
|
183
|
+
(post_process_match("Mr", ctxt->constituent[c1].start_link) == 0) &&
|
184
|
+
(post_process_match("MX#d", ctxt->constituent[c1].start_link) == 0)))
|
185
|
+
continue;
|
186
|
+
|
187
|
+
/* If ctype1 is SBAR (clause opener case), it has to be an f domain */
|
188
|
+
if ((x==5) && (ctxt->constituent[c1].domain_type!='f'))
|
189
|
+
continue;
|
190
|
+
|
191
|
+
/* If ctype1 is SBAR (pp opener case), it has to be a g domain */
|
192
|
+
if ((x==6) && (ctxt->constituent[c1].domain_type!='g'))
|
193
|
+
continue;
|
194
|
+
|
195
|
+
/* If ctype1 is NP (paraphrase case), it has to be started by an SI */
|
196
|
+
if ((x==7) && (post_process_match("SI", ctxt->constituent[c1].start_link)==0))
|
197
|
+
continue;
|
198
|
+
|
199
|
+
/* If ctype1 is VP (participle modifier case), it has to be
|
200
|
+
started by an Mv or Mg */
|
201
|
+
if ((x==8) && (post_process_match("M", ctxt->constituent[c1].start_link)==0))
|
202
|
+
continue;
|
203
|
+
|
204
|
+
/* If ctype1 is VP (participle opener case), it has
|
205
|
+
to be started by a COp */
|
206
|
+
if ((x==9) && (post_process_match("COp", ctxt->constituent[c1].start_link)==0))
|
207
|
+
continue;
|
208
|
+
|
209
|
+
/* Now start at the bounds of c1, and work outwards until you
|
210
|
+
find a larger constituent of type ctype2 */
|
211
|
+
if (!(strcmp(ctxt->constituent[c1].type, ctype1)==0))
|
212
|
+
continue;
|
213
|
+
|
214
|
+
if (verbosity >= 2)
|
215
|
+
printf("Generating complement constituent for c %d of type %s\n",
|
216
|
+
c1, ctype1);
|
217
|
+
done = 0;
|
218
|
+
for (w2=ctxt->constituent[c1].left; (done==0) && (w2>=0); w2--) {
|
219
|
+
for (w3=ctxt->constituent[c1].right; w3<linkage->num_words; w3++) {
|
220
|
+
for (c2=numcon_total; (done==0) &&
|
221
|
+
(c2 < numcon_total + numcon_subl); c2++) {
|
222
|
+
if (!((ctxt->constituent[c2].left==w2) &&
|
223
|
+
(ctxt->constituent[c2].right==w3)) || (c2==c1))
|
224
|
+
continue;
|
225
|
+
if (!(strcmp(ctxt->constituent[c2].type, ctype2)==0))
|
226
|
+
continue;
|
227
|
+
|
228
|
+
/* if the new constituent (c) is to the left
|
229
|
+
of c1, its right edge should be adjacent to the
|
230
|
+
left edge of c1 - or as close as possible
|
231
|
+
without going outside the current sublinkage.
|
232
|
+
(Or substituting right and left as necessary.) */
|
233
|
+
|
234
|
+
if ((x==5) || (x==6) || (x==9)) {
|
235
|
+
/* This is the case where c is to the
|
236
|
+
RIGHT of c1 */
|
237
|
+
w = ctxt->constituent[c1].right+1;
|
238
|
+
while(1) {
|
239
|
+
if (ctxt->word_used[linkage->current][w]==1)
|
240
|
+
break;
|
241
|
+
w++;
|
242
|
+
}
|
243
|
+
if (w > ctxt->constituent[c2].right)
|
244
|
+
{
|
245
|
+
done=1;
|
246
|
+
continue;
|
247
|
+
}
|
248
|
+
ctxt->constituent[c].left = w;
|
249
|
+
ctxt->constituent[c].right = ctxt->constituent[c2].right;
|
250
|
+
}
|
251
|
+
else {
|
252
|
+
w = ctxt->constituent[c1].left-1;
|
253
|
+
while(1) {
|
254
|
+
if (ctxt->word_used[linkage->current][w] == 1)
|
255
|
+
break;
|
256
|
+
w--;
|
257
|
+
}
|
258
|
+
if (w < ctxt->constituent[c2].left) {
|
259
|
+
done=1;
|
260
|
+
continue;
|
261
|
+
}
|
262
|
+
ctxt->constituent[c].right = w;
|
263
|
+
ctxt->constituent[c].left = ctxt->constituent[c2].left;
|
264
|
+
}
|
265
|
+
|
266
|
+
adjust_for_left_comma(ctxt, linkage, c1);
|
267
|
+
adjust_for_right_comma(ctxt, linkage, c1);
|
268
|
+
|
269
|
+
ctxt->constituent[c].type =
|
270
|
+
string_set_add(ctype3, ctxt->phrase_ss);
|
271
|
+
ctxt->constituent[c].domain_type = 'x';
|
272
|
+
ctxt->constituent[c].start_link =
|
273
|
+
string_set_add("XX", ctxt->phrase_ss);
|
274
|
+
ctxt->constituent[c].start_num =
|
275
|
+
ctxt->constituent[c1].start_num; /* bogus */
|
276
|
+
if (verbosity >= 2)
|
277
|
+
{
|
278
|
+
printf("Larger c found: c %d (%s); ",
|
279
|
+
c2, ctype2);
|
280
|
+
printf("Adding constituent:\n");
|
281
|
+
print_constituent(ctxt, linkage, c);
|
282
|
+
}
|
283
|
+
c++;
|
284
|
+
if (MAXCONSTITUENTS <= c)
|
285
|
+
{
|
286
|
+
err_ctxt ec;
|
287
|
+
ec.sent = linkage->sent;
|
288
|
+
err_msg(&ec, Error, "Error: Too many constituents (a).\n");
|
289
|
+
c--;
|
290
|
+
}
|
291
|
+
done = 1;
|
292
|
+
}
|
293
|
+
}
|
294
|
+
}
|
295
|
+
if (verbosity >= 2)
|
296
|
+
{
|
297
|
+
if (done == 0)
|
298
|
+
printf("No constituent added, because no larger %s " \
|
299
|
+
" was found\n", ctype2);
|
300
|
+
}
|
301
|
+
}
|
302
|
+
numcon_subl = c - numcon_total;
|
303
|
+
return numcon_subl;
|
304
|
+
}
|
305
|
+
|
306
|
+
/**
|
307
|
+
* Look for a constituent started by an MVs or MVg.
|
308
|
+
* Find any VP's or ADJP's that contain it (without going
|
309
|
+
* beyond a larger S or NP). Adjust them so that
|
310
|
+
* they end right before the m domain starts.
|
311
|
+
*/
|
312
|
+
static void adjust_subordinate_clauses(con_context_t *ctxt, Linkage linkage,
|
313
|
+
int numcon_total,
|
314
|
+
int numcon_subl)
|
315
|
+
{
|
316
|
+
int c, w, c2, w2, done;
|
317
|
+
|
318
|
+
for (c=numcon_total; c<numcon_total + numcon_subl; c++) {
|
319
|
+
if ((post_process_match("MVs", ctxt->constituent[c].start_link) == 1) ||
|
320
|
+
(post_process_match("MVg", ctxt->constituent[c].start_link)==1)) {
|
321
|
+
done=0;
|
322
|
+
for (w2=ctxt->constituent[c].left-1; (done==0) && w2>=0; w2--) {
|
323
|
+
for (c2=numcon_total; c2<numcon_total + numcon_subl; c2++) {
|
324
|
+
if (!((ctxt->constituent[c2].left==w2) &&
|
325
|
+
(ctxt->constituent[c2].right >= ctxt->constituent[c].right)))
|
326
|
+
continue;
|
327
|
+
if ((strcmp(ctxt->constituent[c2].type, "S") == 0) ||
|
328
|
+
(strcmp(ctxt->constituent[c2].type, "NP") == 0)) {
|
329
|
+
done=1;
|
330
|
+
break;
|
331
|
+
}
|
332
|
+
if ((ctxt->constituent[c2].domain_type == 'v') ||
|
333
|
+
(ctxt->constituent[c2].domain_type == 'a')) {
|
334
|
+
w = ctxt->constituent[c].left-1;
|
335
|
+
while (1) {
|
336
|
+
if (ctxt->word_used[linkage->current][w] == 1) break;
|
337
|
+
w--;
|
338
|
+
}
|
339
|
+
ctxt->constituent[c2].right = w;
|
340
|
+
|
341
|
+
if (verbosity >= 2)
|
342
|
+
printf("Adjusting constituent %d:\n", c2);
|
343
|
+
print_constituent(ctxt, linkage, c2);
|
344
|
+
}
|
345
|
+
}
|
346
|
+
}
|
347
|
+
if (strcmp(linkage->word[ctxt->constituent[c].left], ",") == 0)
|
348
|
+
ctxt->constituent[c].left++;
|
349
|
+
}
|
350
|
+
}
|
351
|
+
}
|
352
|
+
|
353
|
+
/******************************************************
|
354
|
+
* These functions are called once, after constituents
|
355
|
+
* for each sublinkage have been generated, to merge them
|
356
|
+
* together and fix up some other things.
|
357
|
+
*
|
358
|
+
********************************************************/
|
359
|
+
|
360
|
+
/**
|
361
|
+
* Here we're looking for the next andlist element to add on
|
362
|
+
* to a conjectural andlist, stored in the array templist.
|
363
|
+
* We go through the constituents, starting at "start".
|
364
|
+
*/
|
365
|
+
static int find_next_element(con_context_t *ctxt,
|
366
|
+
Linkage linkage,
|
367
|
+
int start,
|
368
|
+
int numcon_total,
|
369
|
+
int num_elements,
|
370
|
+
int num_lists)
|
371
|
+
{
|
372
|
+
int c, a, ok, c2, c3, addedone=0, n;
|
373
|
+
|
374
|
+
assert(num_elements <= MAX_ELTS, "Constutent element array overflow!\n");
|
375
|
+
|
376
|
+
n = num_lists;
|
377
|
+
for (c=start+1; c<numcon_total; c++)
|
378
|
+
{
|
379
|
+
constituent_t *cc = &ctxt->constituent[c];
|
380
|
+
|
381
|
+
if (cc->valid == 0)
|
382
|
+
continue;
|
383
|
+
if (strcmp(ctxt->constituent[ctxt->templist[0]].type, cc->type)!=0)
|
384
|
+
continue;
|
385
|
+
ok = 1;
|
386
|
+
|
387
|
+
/* We're considering adding constituent c to the andlist.
|
388
|
+
If c is in the same sublinkage as one of the other andlist
|
389
|
+
elements, don't add it. If it overlaps with one of the other
|
390
|
+
constituents, don't add it. If there's a constituent
|
391
|
+
identical to c that occurs in a sublinkage in which one of
|
392
|
+
the other elements occurs, don't add it. */
|
393
|
+
|
394
|
+
for (a=0; a<num_elements; a++)
|
395
|
+
{
|
396
|
+
int t = ctxt->templist[a];
|
397
|
+
constituent_t *ct = &ctxt->constituent[t];
|
398
|
+
|
399
|
+
if (cc->subl == ct->subl)
|
400
|
+
ok=0;
|
401
|
+
if (((cc->left < ct->left) && (cc->right > ct->left))
|
402
|
+
||
|
403
|
+
((cc->right > ct->right) && (cc->left < ct->right))
|
404
|
+
||
|
405
|
+
((cc->right > ct->right) && (cc->left < ct->right))
|
406
|
+
||
|
407
|
+
((cc->left > ct->left) && (cc->right < ct->right)))
|
408
|
+
ok=0;
|
409
|
+
|
410
|
+
for (c2=0; c2<numcon_total; c2++)
|
411
|
+
{
|
412
|
+
if (ctxt->constituent[c2].canon != cc->canon)
|
413
|
+
continue;
|
414
|
+
for (c3=0; c3<numcon_total; c3++)
|
415
|
+
{
|
416
|
+
if ((ctxt->constituent[c3].canon == ct->canon)
|
417
|
+
&& (ctxt->constituent[c3].subl == ctxt->constituent[c2].subl))
|
418
|
+
ok=0;
|
419
|
+
}
|
420
|
+
}
|
421
|
+
}
|
422
|
+
if (ok == 0) continue;
|
423
|
+
|
424
|
+
ctxt->templist[num_elements] = c;
|
425
|
+
addedone = 1;
|
426
|
+
num_lists = find_next_element(ctxt, linkage, c, numcon_total,
|
427
|
+
num_elements+1, num_lists);
|
428
|
+
|
429
|
+
/* Test for overlow of the and-list.
|
430
|
+
* With the current parser, the following will cause an
|
431
|
+
* overflow:
|
432
|
+
*
|
433
|
+
* I have not seen the grysbok, or the suni, or the dibitag, or
|
434
|
+
* the lechwi, or the aoul, or the gerenuk, or the blaauwbok,
|
435
|
+
* or the chevrotain, or lots of others, but who in the world
|
436
|
+
* could guess what they were or what they looked like, judging
|
437
|
+
* only from the names?
|
438
|
+
*/
|
439
|
+
if (MAX_ANDS <= num_lists)
|
440
|
+
{
|
441
|
+
err_ctxt ec;
|
442
|
+
ec.sent = linkage->sent;
|
443
|
+
err_msg(&ec, Error, "Error: Constituent overflowed andlist!\n");
|
444
|
+
return MAX_ANDS;
|
445
|
+
}
|
446
|
+
}
|
447
|
+
|
448
|
+
if (addedone == 0 && num_elements > 1)
|
449
|
+
{
|
450
|
+
for (a=0; a<num_elements; a++) {
|
451
|
+
ctxt->andlist[num_lists].e[a] = ctxt->templist[a];
|
452
|
+
ctxt->andlist[num_lists].num = num_elements;
|
453
|
+
}
|
454
|
+
num_lists++;
|
455
|
+
}
|
456
|
+
return num_lists;
|
457
|
+
}
|
458
|
+
|
459
|
+
static int merge_constituents(con_context_t *ctxt, Linkage linkage, int numcon_total)
|
460
|
+
{
|
461
|
+
int c1, c2=0, c3, ok, a, n, a2, n2, match, listmatch, a3;
|
462
|
+
int num_lists, num_elements;
|
463
|
+
int leftend, rightend;
|
464
|
+
|
465
|
+
for (c1=0; c1<numcon_total; c1++)
|
466
|
+
{
|
467
|
+
ctxt->constituent[c1].valid = 1;
|
468
|
+
|
469
|
+
/* Find and invalidate any constituents with negative length */
|
470
|
+
if(ctxt->constituent[c1].right < ctxt->constituent[c1].left)
|
471
|
+
{
|
472
|
+
if(verbosity >= 2)
|
473
|
+
{
|
474
|
+
err_ctxt ec;
|
475
|
+
ec.sent = linkage->sent;
|
476
|
+
err_msg(&ec, Warn,
|
477
|
+
"Warning: Constituent %d has negative length. Deleting it.\n", c1);
|
478
|
+
}
|
479
|
+
ctxt->constituent[c1].valid = 0;
|
480
|
+
}
|
481
|
+
ctxt->constituent[c1].canon = c1;
|
482
|
+
}
|
483
|
+
|
484
|
+
/* First go through and give each constituent a canonical number
|
485
|
+
(the index number of the lowest-numbered constituent
|
486
|
+
identical to it) */
|
487
|
+
|
488
|
+
for (c1 = 0; c1 < numcon_total; c1++)
|
489
|
+
{
|
490
|
+
if (ctxt->constituent[c1].canon != c1) continue;
|
491
|
+
for (c2 = c1 + 1; c2 < numcon_total; c2++)
|
492
|
+
{
|
493
|
+
if ((ctxt->constituent[c1].left == ctxt->constituent[c2].left) &&
|
494
|
+
(ctxt->constituent[c1].right == ctxt->constituent[c2].right) &&
|
495
|
+
(strcmp(ctxt->constituent[c1].type, ctxt->constituent[c2].type) == 0))
|
496
|
+
{
|
497
|
+
ctxt->constituent[c2].canon = c1;
|
498
|
+
}
|
499
|
+
}
|
500
|
+
}
|
501
|
+
|
502
|
+
/* If constituents A and B in different sublinkages X and Y
|
503
|
+
* have one endpoint in common, but A is larger at the other end,
|
504
|
+
* and B has no duplicate in X, then declare B invalid. (Example:
|
505
|
+
* " [A [B We saw the cat B] and the dog A] "
|
506
|
+
*/
|
507
|
+
for (c1 = 0; c1 < numcon_total; c1++)
|
508
|
+
{
|
509
|
+
if (ctxt->constituent[c1].valid == 0) continue;
|
510
|
+
for (c2 = 0; c2 < numcon_total; c2++)
|
511
|
+
{
|
512
|
+
if (ctxt->constituent[c2].subl == ctxt->constituent[c1].subl) continue;
|
513
|
+
ok = 1;
|
514
|
+
/* Does c2 have a duplicate in the sublinkage containing c1?
|
515
|
+
If so, bag it */
|
516
|
+
for (c3 = 0; c3 < numcon_total; c3++)
|
517
|
+
{
|
518
|
+
if ((ctxt->constituent[c2].canon == ctxt->constituent[c3].canon) &&
|
519
|
+
(ctxt->constituent[c3].subl == ctxt->constituent[c1].subl))
|
520
|
+
ok = 0;
|
521
|
+
}
|
522
|
+
for (c3 = 0; c3 < numcon_total; c3++)
|
523
|
+
{
|
524
|
+
if ((ctxt->constituent[c1].canon == ctxt->constituent[c3].canon) &&
|
525
|
+
(ctxt->constituent[c3].subl == ctxt->constituent[c2].subl))
|
526
|
+
ok = 0;
|
527
|
+
}
|
528
|
+
if (ok == 0) continue;
|
529
|
+
if ((ctxt->constituent[c1].left == ctxt->constituent[c2].left) &&
|
530
|
+
(ctxt->constituent[c1].right > ctxt->constituent[c2].right) &&
|
531
|
+
(strcmp(ctxt->constituent[c1].type, ctxt->constituent[c2].type) == 0))
|
532
|
+
{
|
533
|
+
ctxt->constituent[c2].valid = 0;
|
534
|
+
}
|
535
|
+
|
536
|
+
if ((ctxt->constituent[c1].left < ctxt->constituent[c2].left) &&
|
537
|
+
(ctxt->constituent[c1].right == ctxt->constituent[c2].right) &&
|
538
|
+
(strcmp(ctxt->constituent[c1].type, ctxt->constituent[c2].type) == 0))
|
539
|
+
{
|
540
|
+
ctxt->constituent[c2].valid = 0;
|
541
|
+
}
|
542
|
+
}
|
543
|
+
}
|
544
|
+
|
545
|
+
/* Now go through and find duplicates; if a pair is found,
|
546
|
+
* mark one as invalid. (It doesn't matter if they're in the
|
547
|
+
* same sublinkage or not)
|
548
|
+
*/
|
549
|
+
for (c1 = 0; c1 < numcon_total; c1++)
|
550
|
+
{
|
551
|
+
if (ctxt->constituent[c1].valid == 0) continue;
|
552
|
+
for (c2 = c1 + 1; c2 < numcon_total; c2++)
|
553
|
+
{
|
554
|
+
if (ctxt->constituent[c2].canon == ctxt->constituent[c1].canon)
|
555
|
+
ctxt->constituent[c2].valid = 0;
|
556
|
+
}
|
557
|
+
}
|
558
|
+
|
559
|
+
/* Now we generate the and-lists. An and-list is a set of mutually
|
560
|
+
* exclusive constituents. Each constituent in the list may not
|
561
|
+
* be present in the same sublinkage as any of the others.
|
562
|
+
*/
|
563
|
+
num_lists = 0;
|
564
|
+
for (c1 = 0; c1 < numcon_total; c1++)
|
565
|
+
{
|
566
|
+
if (ctxt->constituent[c1].valid == 0) continue;
|
567
|
+
num_elements = 1;
|
568
|
+
ctxt->templist[0] = c1;
|
569
|
+
num_lists = find_next_element(ctxt, linkage, c1, numcon_total,
|
570
|
+
num_elements, num_lists);
|
571
|
+
|
572
|
+
/* If we're overflowing, then punt */
|
573
|
+
if (MAX_ANDS <= num_lists)
|
574
|
+
break;
|
575
|
+
}
|
576
|
+
|
577
|
+
if (verbosity >= 2)
|
578
|
+
{
|
579
|
+
printf("And-lists:\n");
|
580
|
+
for (n=0; n<num_lists; n++)
|
581
|
+
{
|
582
|
+
printf(" %d: ", n);
|
583
|
+
for (a=0; a < ctxt->andlist[n].num; a++)
|
584
|
+
{
|
585
|
+
printf("%d ", ctxt->andlist[n].e[a]);
|
586
|
+
}
|
587
|
+
printf("\n");
|
588
|
+
}
|
589
|
+
}
|
590
|
+
|
591
|
+
/* Now we prune out any andlists that are subsumed by other
|
592
|
+
* andlists--e.g. if andlist X contains constituents A and B,
|
593
|
+
* and Y contains A B and C, we throw out X
|
594
|
+
*/
|
595
|
+
for (n = 0; n < num_lists; n++)
|
596
|
+
{
|
597
|
+
ctxt->andlist[n].valid = 1;
|
598
|
+
for (n2 = 0; n2 < num_lists; n2++)
|
599
|
+
{
|
600
|
+
if (n2 == n) continue;
|
601
|
+
if (ctxt->andlist[n2].num < ctxt->andlist[n].num)
|
602
|
+
continue;
|
603
|
+
|
604
|
+
listmatch = 1;
|
605
|
+
for (a = 0; a < ctxt->andlist[n].num; a++)
|
606
|
+
{
|
607
|
+
match = 0;
|
608
|
+
for (a2 = 0; a2 < ctxt->andlist[n2].num; a2++)
|
609
|
+
{
|
610
|
+
if (ctxt->andlist[n2].e[a2] == ctxt->andlist[n].e[a])
|
611
|
+
match = 1;
|
612
|
+
}
|
613
|
+
if (match == 0) listmatch = 0;
|
614
|
+
/* At least one element was not matched by n2 */
|
615
|
+
}
|
616
|
+
if (listmatch == 1) ctxt->andlist[n].valid = 0;
|
617
|
+
}
|
618
|
+
}
|
619
|
+
|
620
|
+
/* If an element of an andlist contains an element of another
|
621
|
+
* andlist, it must contain the entire andlist.
|
622
|
+
*/
|
623
|
+
for (n = 0; n < num_lists; n++)
|
624
|
+
{
|
625
|
+
if (ctxt->andlist[n].valid == 0)
|
626
|
+
continue;
|
627
|
+
for (a = 0; (a < ctxt->andlist[n].num) && (ctxt->andlist[n].valid); a++)
|
628
|
+
{
|
629
|
+
for (n2 = 0; (n2 < num_lists) && (ctxt->andlist[n].valid); n2++)
|
630
|
+
{
|
631
|
+
if ((n2 == n) || (ctxt->andlist[n2].valid == 0))
|
632
|
+
continue;
|
633
|
+
for (a2 = 0; (a2 < ctxt->andlist[n2].num) && (ctxt->andlist[n].valid); a2++)
|
634
|
+
{
|
635
|
+
c1 = ctxt->andlist[n].e[a];
|
636
|
+
c2 = ctxt->andlist[n2].e[a2];
|
637
|
+
if (c1 == c2)
|
638
|
+
continue;
|
639
|
+
if (!((ctxt->constituent[c2].left <= ctxt->constituent[c1].left) &&
|
640
|
+
(ctxt->constituent[c2].right >= ctxt->constituent[c1].right)))
|
641
|
+
continue;
|
642
|
+
if (verbosity >= 2)
|
643
|
+
printf("Found that c%d in list %d is bigger " \
|
644
|
+
"than c%d in list %d\n", c2, n2, c1, n);
|
645
|
+
ok = 1;
|
646
|
+
|
647
|
+
/* An element of n2 contains an element of n.
|
648
|
+
* Now, we check to see if that element of n2
|
649
|
+
* contains ALL the elements of n.
|
650
|
+
* If not, n is invalid.
|
651
|
+
*/
|
652
|
+
for (a3 = 0; a3 < ctxt->andlist[n].num; a3++)
|
653
|
+
{
|
654
|
+
c3 = ctxt->andlist[n].e[a3];
|
655
|
+
if ((ctxt->constituent[c2].left>ctxt->constituent[c3].left) ||
|
656
|
+
(ctxt->constituent[c2].right<ctxt->constituent[c3].right))
|
657
|
+
ok = 0;
|
658
|
+
}
|
659
|
+
if (ok != 0)
|
660
|
+
continue;
|
661
|
+
ctxt->andlist[n].valid = 0;
|
662
|
+
if (verbosity >= 2)
|
663
|
+
{
|
664
|
+
printf("Eliminating andlist, " \
|
665
|
+
"n=%d, a=%d, n2=%d, a2=%d: ",
|
666
|
+
n, a, n2, a2);
|
667
|
+
for (a3 = 0; a3 < ctxt->andlist[n].num; a3++)
|
668
|
+
{
|
669
|
+
printf("%d ", ctxt->andlist[n].e[a3]);
|
670
|
+
}
|
671
|
+
printf("\n");
|
672
|
+
}
|
673
|
+
}
|
674
|
+
}
|
675
|
+
}
|
676
|
+
}
|
677
|
+
|
678
|
+
if (verbosity >= 2)
|
679
|
+
{
|
680
|
+
printf("And-lists after pruning:\n");
|
681
|
+
for (n=0; n<num_lists; n++) {
|
682
|
+
if (ctxt->andlist[n].valid==0)
|
683
|
+
continue;
|
684
|
+
printf(" %d: ", n);
|
685
|
+
for (a=0; a<ctxt->andlist[n].num; a++) {
|
686
|
+
printf("%d ", ctxt->andlist[n].e[a]);
|
687
|
+
}
|
688
|
+
printf("\n");
|
689
|
+
}
|
690
|
+
}
|
691
|
+
|
692
|
+
c1 = numcon_total;
|
693
|
+
for (n = 0; n < num_lists; n++)
|
694
|
+
{
|
695
|
+
if (ctxt->andlist[n].valid == 0) continue;
|
696
|
+
leftend = 256;
|
697
|
+
rightend = -1;
|
698
|
+
for (a = 0; a < ctxt->andlist[n].num; a++)
|
699
|
+
{
|
700
|
+
c2 = ctxt->andlist[n].e[a];
|
701
|
+
if (ctxt->constituent[c2].left < leftend)
|
702
|
+
{
|
703
|
+
leftend = ctxt->constituent[c2].left;
|
704
|
+
}
|
705
|
+
if (ctxt->constituent[c2].right > rightend)
|
706
|
+
{
|
707
|
+
rightend=ctxt->constituent[c2].right;
|
708
|
+
}
|
709
|
+
}
|
710
|
+
|
711
|
+
ctxt->constituent[c1].left = leftend;
|
712
|
+
ctxt->constituent[c1].right = rightend;
|
713
|
+
ctxt->constituent[c1].type = ctxt->constituent[c2].type;
|
714
|
+
ctxt->constituent[c1].domain_type = 'x';
|
715
|
+
ctxt->constituent[c1].valid = 1;
|
716
|
+
ctxt->constituent[c1].start_link = ctxt->constituent[c2].start_link; /* bogus */
|
717
|
+
ctxt->constituent[c1].start_num = ctxt->constituent[c2].start_num; /* bogus */
|
718
|
+
|
719
|
+
#ifdef AUX_CODE_IS_DEAD /* See comments above */
|
720
|
+
/* If a constituent within the andlist is an aux (aux==1),
|
721
|
+
* set aux for the whole-list constituent to 2, also set
|
722
|
+
* aux for the smaller constituent to 2, meaning they'll both
|
723
|
+
* be printed (as an "X"). (If aux is 2 for the smaller
|
724
|
+
* constituent going in, the same thing should be done,
|
725
|
+
* though I doubt this ever happens.)
|
726
|
+
*/
|
727
|
+
for (a = 0; a < ctxt->andlist[n].num; a++)
|
728
|
+
{
|
729
|
+
c2 = ctxt->andlist[n].e[a];
|
730
|
+
if ((ctxt->constituent[c2].aux == 1) || (ctxt->constituent[c2].aux == 2))
|
731
|
+
{
|
732
|
+
ctxt->constituent[c1].aux = 2;
|
733
|
+
ctxt->constituent[c2].aux = 2;
|
734
|
+
}
|
735
|
+
}
|
736
|
+
#endif /* AUX_CODE_IS_DEAD */
|
737
|
+
|
738
|
+
if (verbosity >= 2)
|
739
|
+
printf("Adding constituent:\n");
|
740
|
+
print_constituent(ctxt, linkage, c1);
|
741
|
+
c1++;
|
742
|
+
}
|
743
|
+
numcon_total = c1;
|
744
|
+
return numcon_total;
|
745
|
+
}
|
746
|
+
|
747
|
+
/**
|
748
|
+
* Go through all the words. If a word is on the right end of
|
749
|
+
* an S (or SF or SX), wordtype[w]=STYPE. If it's also on the left end of a
|
750
|
+
* Pg*b, I, PP, or Pv, wordtype[w]=PTYPE. If it's a question-word
|
751
|
+
* used in an indirect question, wordtype[w]=QTYPE. If it's a
|
752
|
+
* question-word determiner, wordtype[w]=QDTYPE. Else wordtype[w]=NONE.
|
753
|
+
* (This function is called once for each sublinkage.)
|
754
|
+
*/
|
755
|
+
static void generate_misc_word_info(con_context_t * ctxt, Linkage linkage)
|
756
|
+
{
|
757
|
+
int l1, l2, w1, w2;
|
758
|
+
const char * label1, * label2;
|
759
|
+
|
760
|
+
for (w1=0; w1<linkage->num_words; w1++)
|
761
|
+
ctxt->wordtype[w1]=NONE;
|
762
|
+
|
763
|
+
for (l1=0; l1<linkage_get_num_links(linkage); l1++) {
|
764
|
+
w1=linkage_get_link_rword(linkage, l1);
|
765
|
+
label1 = linkage_get_link_label(linkage, l1);
|
766
|
+
if ((uppercompare(label1, "S")==0) ||
|
767
|
+
(uppercompare(label1, "SX")==0) ||
|
768
|
+
(uppercompare(label1, "SF")==0)) {
|
769
|
+
ctxt->wordtype[w1] = STYPE;
|
770
|
+
for (l2=0; l2<linkage_get_num_links(linkage); l2++) {
|
771
|
+
w2=linkage_get_link_lword(linkage, l2);
|
772
|
+
label2 = linkage_get_link_label(linkage, l2);
|
773
|
+
if ((w1==w2) &&
|
774
|
+
((post_process_match("Pg#b", label2)==1) ||
|
775
|
+
(uppercompare(label2, "I")==0) ||
|
776
|
+
(uppercompare(label2, "PP")==0) ||
|
777
|
+
(post_process_match("Pv", label2)==1))) {
|
778
|
+
/* Pvf, Pgf? */
|
779
|
+
ctxt->wordtype[w1] = PTYPE;
|
780
|
+
}
|
781
|
+
}
|
782
|
+
}
|
783
|
+
if (post_process_match("QI#d", label1)==1) {
|
784
|
+
ctxt->wordtype[w1] = QTYPE;
|
785
|
+
for (l2=0; l2<linkage_get_num_links(linkage); l2++) {
|
786
|
+
w2=linkage_get_link_lword(linkage, l2);
|
787
|
+
label2 = linkage_get_link_label(linkage, l2);
|
788
|
+
if ((w1==w2) && (post_process_match("D##w", label2)==1)) {
|
789
|
+
ctxt->wordtype[w1] = QDTYPE;
|
790
|
+
}
|
791
|
+
}
|
792
|
+
}
|
793
|
+
if (post_process_match("Mr", label1)==1) ctxt->wordtype[w1] = QDTYPE;
|
794
|
+
if (post_process_match("MX#d", label1)==1) ctxt->wordtype[w1] = QDTYPE;
|
795
|
+
}
|
796
|
+
}
|
797
|
+
|
798
|
+
static int last_minute_fixes(con_context_t *ctxt, Linkage linkage, int numcon_total)
|
799
|
+
{
|
800
|
+
int c, c2, global_leftend_found, adjustment_made,
|
801
|
+
global_rightend_found, lastword, newcon_total = 0;
|
802
|
+
Sentence sent;
|
803
|
+
sent = linkage_get_sentence(linkage);
|
804
|
+
|
805
|
+
for (c = 0; c < numcon_total; c++)
|
806
|
+
{
|
807
|
+
/* In a paraphrase construction ("John ran, he said"),
|
808
|
+
the paraphrasing clause doesn't get
|
809
|
+
an S. (This is true in Treebank II, not Treebank I) */
|
810
|
+
|
811
|
+
if (uppercompare(ctxt->constituent[c].start_link, "CP") == 0)
|
812
|
+
{
|
813
|
+
ctxt->constituent[c].valid = 0;
|
814
|
+
}
|
815
|
+
|
816
|
+
/* If it's a possessive with an "'s", the NP on the left
|
817
|
+
should be extended to include the "'s". */
|
818
|
+
if ((uppercompare(ctxt->constituent[c].start_link, "YS") == 0) ||
|
819
|
+
(uppercompare(ctxt->constituent[c].start_link, "YP") == 0))
|
820
|
+
{
|
821
|
+
ctxt->constituent[c].right++;
|
822
|
+
}
|
823
|
+
|
824
|
+
/* If a constituent has starting link MVpn, it's a time
|
825
|
+
expression like "last week"; label it as a noun phrase
|
826
|
+
(incorrectly) */
|
827
|
+
|
828
|
+
if (strcmp(ctxt->constituent[c].start_link, "MVpn") == 0)
|
829
|
+
{
|
830
|
+
ctxt->constituent[c].type = string_set_add("NP", ctxt->phrase_ss);
|
831
|
+
}
|
832
|
+
if (strcmp(ctxt->constituent[c].start_link, "COn") == 0)
|
833
|
+
{
|
834
|
+
ctxt->constituent[c].type = string_set_add("NP", ctxt->phrase_ss);
|
835
|
+
}
|
836
|
+
if (strcmp(ctxt->constituent[c].start_link, "Mpn") == 0)
|
837
|
+
{
|
838
|
+
ctxt->constituent[c].type = string_set_add("NP", ctxt->phrase_ss);
|
839
|
+
}
|
840
|
+
|
841
|
+
/* If the constituent is an S started by "but" or "and" at
|
842
|
+
the beginning of the sentence, it should be ignored. */
|
843
|
+
|
844
|
+
if ((strcmp(ctxt->constituent[c].start_link, "Wdc") == 0) &&
|
845
|
+
(ctxt->constituent[c].left == 2))
|
846
|
+
{
|
847
|
+
ctxt->constituent[c].valid = 0;
|
848
|
+
}
|
849
|
+
|
850
|
+
/* For prenominal adjectives, an ADJP constituent is assigned
|
851
|
+
if it's a hyphenated (Ah) or comparative (Am) adjective;
|
852
|
+
otherwise no ADJP is assigned, unless the phrase is more
|
853
|
+
than one word long (e.g. "very big"). The same with certain
|
854
|
+
types of adverbs. */
|
855
|
+
/* That was for Treebank I. For Treebank II, the rule only
|
856
|
+
seems to apply to prenominal adjectives (of all kinds).
|
857
|
+
However, it also applies to number expressions ("QP"). */
|
858
|
+
|
859
|
+
if ((post_process_match("A", ctxt->constituent[c].start_link) == 1) ||
|
860
|
+
(ctxt->constituent[c].domain_type == 'd') ||
|
861
|
+
(ctxt->constituent[c].domain_type == 'h')) {
|
862
|
+
if (ctxt->constituent[c].right-ctxt->constituent[c].left == 0)
|
863
|
+
{
|
864
|
+
ctxt->constituent[c].valid = 0;
|
865
|
+
}
|
866
|
+
}
|
867
|
+
|
868
|
+
if ((ctxt->constituent[c].domain_type == 'h') &&
|
869
|
+
(strcmp(linkage->word[ctxt->constituent[c].left - 1], "$") == 0))
|
870
|
+
{
|
871
|
+
ctxt->constituent[c].left--;
|
872
|
+
}
|
873
|
+
|
874
|
+
#ifdef AUX_CODE_IS_DEAD /* See comments at top */
|
875
|
+
/* If a constituent has type VP and its aux value is 2,
|
876
|
+
this means it's an aux that should be printed; change its
|
877
|
+
type to "X". If its aux value is 1, set "valid" to 0. (This
|
878
|
+
applies to Treebank I only) */
|
879
|
+
|
880
|
+
if (ctxt->constituent[c].aux == 2)
|
881
|
+
{
|
882
|
+
ctxt->constituent[c].type = string_set_add("X", ctxt->phrase_ss);
|
883
|
+
}
|
884
|
+
if (ctxt->constituent[c].aux == 1)
|
885
|
+
{
|
886
|
+
ctxt->constituent[c].valid = 0;
|
887
|
+
}
|
888
|
+
#endif /* AUX_CODE_IS_DEAD */
|
889
|
+
}
|
890
|
+
|
891
|
+
numcon_total = numcon_total + newcon_total;
|
892
|
+
|
893
|
+
/* If there's a global S constituent that includes everything
|
894
|
+
except a final period or question mark, extend it by one word */
|
895
|
+
|
896
|
+
for (c = 0; c < numcon_total; c++)
|
897
|
+
{
|
898
|
+
if ((ctxt->constituent[c].right == linkage->num_words -3) &&
|
899
|
+
(ctxt->constituent[c].left == 1) &&
|
900
|
+
(strcmp(ctxt->constituent[c].type, "S") == 0) &&
|
901
|
+
(strcmp(sent->word[linkage->num_words -2].string, ".") == 0))
|
902
|
+
ctxt->constituent[c].right++;
|
903
|
+
}
|
904
|
+
|
905
|
+
/* If there's no S boundary at the very left end of the sentence,
|
906
|
+
or the very right end, create a new S spanning the entire sentence */
|
907
|
+
|
908
|
+
lastword = linkage->num_words - 2;
|
909
|
+
global_leftend_found = 0;
|
910
|
+
global_rightend_found = 0;
|
911
|
+
for (c = 0; c < numcon_total; c++)
|
912
|
+
{
|
913
|
+
if ((ctxt->constituent[c].left == 1) && (strcmp(ctxt->constituent[c].type, "S") == 0) &&
|
914
|
+
(ctxt->constituent[c].valid == 1))
|
915
|
+
{
|
916
|
+
global_leftend_found = 1;
|
917
|
+
}
|
918
|
+
}
|
919
|
+
for (c = 0; c < numcon_total; c++)
|
920
|
+
{
|
921
|
+
if ((ctxt->constituent[c].right >= lastword) &&
|
922
|
+
(strcmp(ctxt->constituent[c].type, "S") == 0) && (ctxt->constituent[c].valid == 1))
|
923
|
+
{
|
924
|
+
global_rightend_found = 1;
|
925
|
+
}
|
926
|
+
}
|
927
|
+
if ((global_leftend_found == 0) || (global_rightend_found == 0))
|
928
|
+
{
|
929
|
+
c = numcon_total;
|
930
|
+
ctxt->constituent[c].left = 1;
|
931
|
+
ctxt->constituent[c].right = linkage->num_words-1;
|
932
|
+
ctxt->constituent[c].type = string_set_add("S", ctxt->phrase_ss);
|
933
|
+
ctxt->constituent[c].valid = 1;
|
934
|
+
ctxt->constituent[c].domain_type = 'x';
|
935
|
+
numcon_total++;
|
936
|
+
if (verbosity >= 2)
|
937
|
+
printf("Adding global sentence constituent:\n");
|
938
|
+
print_constituent(ctxt, linkage, c);
|
939
|
+
}
|
940
|
+
|
941
|
+
/* Check once more to see if constituents are nested (checking BETWEEN sublinkages
|
942
|
+
this time) */
|
943
|
+
|
944
|
+
while (1)
|
945
|
+
{
|
946
|
+
adjustment_made=0;
|
947
|
+
for (c = 0; c < numcon_total; c++)
|
948
|
+
{
|
949
|
+
if(ctxt->constituent[c].valid == 0) continue;
|
950
|
+
for (c2 = 0; c2 < numcon_total; c2++)
|
951
|
+
{
|
952
|
+
if(ctxt->constituent[c2].valid == 0) continue;
|
953
|
+
if ((ctxt->constituent[c].left < ctxt->constituent[c2].left) &&
|
954
|
+
(ctxt->constituent[c].right < ctxt->constituent[c2].right) &&
|
955
|
+
(ctxt->constituent[c].right >= ctxt->constituent[c2].left))
|
956
|
+
{
|
957
|
+
if (verbosity >= 2)
|
958
|
+
{
|
959
|
+
err_ctxt ec;
|
960
|
+
ec.sent = linkage->sent;
|
961
|
+
err_msg(&ec, Warn, "Warning: the constituents aren't nested! "
|
962
|
+
"Adjusting them. (%d, %d)\n", c, c2);
|
963
|
+
}
|
964
|
+
ctxt->constituent[c].left = ctxt->constituent[c2].left;
|
965
|
+
}
|
966
|
+
}
|
967
|
+
}
|
968
|
+
if (adjustment_made == 0) break;
|
969
|
+
}
|
970
|
+
return numcon_total;
|
971
|
+
}
|
972
|
+
|
973
|
+
/**
|
974
|
+
* This function generates a table, word_used[i][w], showing
|
975
|
+
* whether each word w is used in each sublinkage i; if so,
|
976
|
+
* the value for that cell of the table is 1.
|
977
|
+
*/
|
978
|
+
static void count_words_used(con_context_t *ctxt, Linkage linkage)
|
979
|
+
{
|
980
|
+
int i, w, link, num_subl;
|
981
|
+
|
982
|
+
num_subl = linkage->num_sublinkages;
|
983
|
+
if(linkage->unionized == 1 && num_subl > 1) num_subl--;
|
984
|
+
|
985
|
+
if (verbosity >= 2)
|
986
|
+
printf("Number of sublinkages = %d\n", num_subl);
|
987
|
+
|
988
|
+
for (i=0; i<num_subl; i++)
|
989
|
+
{
|
990
|
+
for (w = 0; w < linkage->num_words; w++) ctxt->word_used[i][w] = 0;
|
991
|
+
linkage->current = i;
|
992
|
+
for (link = 0; link < linkage_get_num_links(linkage); link++)
|
993
|
+
{
|
994
|
+
ctxt->word_used[i][linkage_get_link_lword(linkage, link)] = 1;
|
995
|
+
ctxt->word_used[i][linkage_get_link_rword(linkage, link)] = 1;
|
996
|
+
}
|
997
|
+
if (verbosity >= 2)
|
998
|
+
{
|
999
|
+
printf("Sublinkage %d: ", i);
|
1000
|
+
for (w = 0; w < linkage->num_words; w++)
|
1001
|
+
{
|
1002
|
+
if (ctxt->word_used[i][w] == 0) printf("0 ");
|
1003
|
+
if (ctxt->word_used[i][w] == 1) printf("1 ");
|
1004
|
+
}
|
1005
|
+
printf("\n");
|
1006
|
+
}
|
1007
|
+
}
|
1008
|
+
}
|
1009
|
+
|
1010
|
+
static int add_constituent(con_context_t *ctxt, int c, Linkage linkage, Domain domain,
|
1011
|
+
int l, int r, const char * name)
|
1012
|
+
{
|
1013
|
+
int nwords = linkage->num_words-2;
|
1014
|
+
c++;
|
1015
|
+
|
1016
|
+
/* Avoid running off end, to walls. */
|
1017
|
+
if (l < 1) l=1;
|
1018
|
+
if (r > nwords) r = nwords;
|
1019
|
+
if (l > nwords) l = nwords;
|
1020
|
+
assert(l <= r, "negative constituent length!" );
|
1021
|
+
|
1022
|
+
ctxt->constituent[c].left = l;
|
1023
|
+
ctxt->constituent[c].right = r;
|
1024
|
+
ctxt->constituent[c].domain_type = domain.type;
|
1025
|
+
ctxt->constituent[c].start_link =
|
1026
|
+
linkage_get_link_label(linkage, domain.start_link);
|
1027
|
+
ctxt->constituent[c].start_num = domain.start_link;
|
1028
|
+
ctxt->constituent[c].type = string_set_add(name, ctxt->phrase_ss);
|
1029
|
+
return c;
|
1030
|
+
}
|
1031
|
+
|
1032
|
+
static const char * cons_of_domain(Linkage linkage, char domain_type)
|
1033
|
+
{
|
1034
|
+
switch (domain_type) {
|
1035
|
+
case 'a':
|
1036
|
+
return "ADJP";
|
1037
|
+
case 'b':
|
1038
|
+
return "SBAR";
|
1039
|
+
case 'c':
|
1040
|
+
return "VP";
|
1041
|
+
case 'd':
|
1042
|
+
return "QP";
|
1043
|
+
case 'e':
|
1044
|
+
return "ADVP";
|
1045
|
+
case 'f':
|
1046
|
+
return "SBAR";
|
1047
|
+
case 'g':
|
1048
|
+
return "PP";
|
1049
|
+
case 'h':
|
1050
|
+
return "QP";
|
1051
|
+
case 'i':
|
1052
|
+
return "ADVP";
|
1053
|
+
case 'k':
|
1054
|
+
return "PRT";
|
1055
|
+
case 'n':
|
1056
|
+
return "NP";
|
1057
|
+
case 'p':
|
1058
|
+
return "PP";
|
1059
|
+
case 'q':
|
1060
|
+
return "SINV";
|
1061
|
+
case 's':
|
1062
|
+
return "S";
|
1063
|
+
case 't':
|
1064
|
+
return "VP";
|
1065
|
+
case 'u':
|
1066
|
+
return "ADJP";
|
1067
|
+
case 'v':
|
1068
|
+
return "VP";
|
1069
|
+
case 'y':
|
1070
|
+
return "NP";
|
1071
|
+
case 'z':
|
1072
|
+
return "VP";
|
1073
|
+
default:
|
1074
|
+
{
|
1075
|
+
err_ctxt ec;
|
1076
|
+
ec.sent = linkage->sent;
|
1077
|
+
err_msg(&ec, Error, "Error: Illegal domain: %c\n", domain_type);
|
1078
|
+
return "";
|
1079
|
+
}
|
1080
|
+
}
|
1081
|
+
}
|
1082
|
+
|
1083
|
+
static int read_constituents_from_domains(con_context_t *ctxt, Linkage linkage,
|
1084
|
+
int numcon_total, int s)
|
1085
|
+
{
|
1086
|
+
int d, c, leftlimit, l, leftmost, rightmost, w, c2, numcon_subl = 0, w2;
|
1087
|
+
List_o_links * dlink;
|
1088
|
+
int rootright, rootleft, adjustment_made;
|
1089
|
+
Sublinkage * subl;
|
1090
|
+
const char * name;
|
1091
|
+
Domain domain;
|
1092
|
+
|
1093
|
+
subl = &linkage->sublinkage[s];
|
1094
|
+
|
1095
|
+
for (d = 0, c = numcon_total; d < subl->pp_data.N_domains; d++, c++)
|
1096
|
+
{
|
1097
|
+
domain = subl->pp_data.domain_array[d];
|
1098
|
+
rootright = linkage_get_link_rword(linkage, domain.start_link);
|
1099
|
+
rootleft = linkage_get_link_lword(linkage, domain.start_link);
|
1100
|
+
|
1101
|
+
if ((domain.type=='c') ||
|
1102
|
+
(domain.type=='d') ||
|
1103
|
+
(domain.type=='e') ||
|
1104
|
+
(domain.type=='f') ||
|
1105
|
+
(domain.type=='g') ||
|
1106
|
+
(domain.type=='u') ||
|
1107
|
+
(domain.type=='y'))
|
1108
|
+
{
|
1109
|
+
leftlimit = 0;
|
1110
|
+
leftmost = linkage_get_link_lword(linkage, domain.start_link);
|
1111
|
+
rightmost = linkage_get_link_lword(linkage, domain.start_link);
|
1112
|
+
}
|
1113
|
+
else
|
1114
|
+
{
|
1115
|
+
leftlimit = linkage_get_link_lword(linkage, domain.start_link) + 1;
|
1116
|
+
leftmost = linkage_get_link_rword(linkage, domain.start_link);
|
1117
|
+
rightmost = linkage_get_link_rword(linkage, domain.start_link);
|
1118
|
+
}
|
1119
|
+
|
1120
|
+
/* Start by assigning both left and right limits to the
|
1121
|
+
* right word of the start link. This will always be contained
|
1122
|
+
* in the constituent. This will also handle the case
|
1123
|
+
* where the domain contains no links.
|
1124
|
+
*/
|
1125
|
+
for (dlink = domain.lol; dlink != NULL; dlink = dlink->next)
|
1126
|
+
{
|
1127
|
+
l = dlink->link;
|
1128
|
+
|
1129
|
+
if ((linkage_get_link_lword(linkage, l) < leftmost) &&
|
1130
|
+
(linkage_get_link_lword(linkage, l) >= leftlimit))
|
1131
|
+
{
|
1132
|
+
leftmost = linkage_get_link_lword(linkage, l);
|
1133
|
+
}
|
1134
|
+
|
1135
|
+
if (linkage_get_link_rword(linkage, l) > rightmost)
|
1136
|
+
{
|
1137
|
+
rightmost = linkage_get_link_rword(linkage, l);
|
1138
|
+
}
|
1139
|
+
}
|
1140
|
+
|
1141
|
+
c--;
|
1142
|
+
c = add_constituent(ctxt, c, linkage, domain, leftmost, rightmost,
|
1143
|
+
cons_of_domain(linkage, domain.type));
|
1144
|
+
|
1145
|
+
if (domain.type == 'z')
|
1146
|
+
{
|
1147
|
+
c = add_constituent(ctxt, c, linkage, domain, leftmost, rightmost, "S");
|
1148
|
+
}
|
1149
|
+
if (domain.type=='c')
|
1150
|
+
{
|
1151
|
+
c = add_constituent(ctxt, c, linkage, domain, leftmost, rightmost, "S");
|
1152
|
+
}
|
1153
|
+
if ((post_process_match("Ce*", ctxt->constituent[c].start_link)==1) ||
|
1154
|
+
(post_process_match("Rn", ctxt->constituent[c].start_link)==1))
|
1155
|
+
{
|
1156
|
+
c = add_constituent(ctxt, c, linkage, domain, leftmost, rightmost, "SBAR");
|
1157
|
+
}
|
1158
|
+
if ((post_process_match("R*", ctxt->constituent[c].start_link)==1) ||
|
1159
|
+
(post_process_match("MX#r", ctxt->constituent[c].start_link)==1))
|
1160
|
+
{
|
1161
|
+
w = leftmost;
|
1162
|
+
if (strcmp(linkage->word[w], ",") == 0) w++;
|
1163
|
+
c = add_constituent(ctxt, c, linkage, domain, w, w, "WHNP");
|
1164
|
+
}
|
1165
|
+
if (post_process_match("Mj", ctxt->constituent[c].start_link) == 1)
|
1166
|
+
{
|
1167
|
+
w = leftmost;
|
1168
|
+
if (strcmp(linkage->word[w], ",") == 0) w++;
|
1169
|
+
c = add_constituent(ctxt, c, linkage, domain, w, w+1, "WHPP");
|
1170
|
+
c = add_constituent(ctxt, c, linkage, domain, w+1, w+1, "WHNP");
|
1171
|
+
}
|
1172
|
+
if ((post_process_match("Ss#d", ctxt->constituent[c].start_link)==1) ||
|
1173
|
+
(post_process_match("B#d", ctxt->constituent[c].start_link)==1))
|
1174
|
+
{
|
1175
|
+
c = add_constituent(ctxt, c, linkage, domain, rootleft, rootleft, "WHNP");
|
1176
|
+
c = add_constituent(ctxt, c, linkage, domain,
|
1177
|
+
rootleft, ctxt->constituent[c-1].right, "SBAR");
|
1178
|
+
}
|
1179
|
+
if (post_process_match("CP", ctxt->constituent[c].start_link)==1)
|
1180
|
+
{
|
1181
|
+
if (strcmp(linkage->word[leftmost], ",") == 0)
|
1182
|
+
ctxt->constituent[c].left++;
|
1183
|
+
c = add_constituent(ctxt, c, linkage, domain, 1, linkage->num_words-1, "S");
|
1184
|
+
}
|
1185
|
+
if ((post_process_match("MVs", ctxt->constituent[c].start_link)==1) ||
|
1186
|
+
(domain.type=='f'))
|
1187
|
+
{
|
1188
|
+
w = ctxt->constituent[c].left;
|
1189
|
+
if (strcmp(linkage->word[w], ",") == 0)
|
1190
|
+
w++;
|
1191
|
+
if (strcmp(linkage->word[w], "when") == 0)
|
1192
|
+
{
|
1193
|
+
c = add_constituent(ctxt, c, linkage, domain, w, w, "WHADVP");
|
1194
|
+
}
|
1195
|
+
}
|
1196
|
+
if (domain.type=='t')
|
1197
|
+
{
|
1198
|
+
c = add_constituent(ctxt, c, linkage, domain, leftmost, rightmost, "S");
|
1199
|
+
}
|
1200
|
+
if ((post_process_match("QI", ctxt->constituent[c].start_link) == 1) ||
|
1201
|
+
(post_process_match("Mr", ctxt->constituent[c].start_link) == 1) ||
|
1202
|
+
(post_process_match("MX#d", ctxt->constituent[c].start_link) == 1))
|
1203
|
+
{
|
1204
|
+
w = leftmost;
|
1205
|
+
if (strcmp(linkage->word[w], ",") == 0) w++;
|
1206
|
+
if (ctxt->wordtype[w] == NONE)
|
1207
|
+
name = "WHADVP";
|
1208
|
+
else if (ctxt->wordtype[w] == QTYPE)
|
1209
|
+
name = "WHNP";
|
1210
|
+
else if (ctxt->wordtype[w] == QDTYPE)
|
1211
|
+
name = "WHNP";
|
1212
|
+
else
|
1213
|
+
assert(0, "Unexpected word type");
|
1214
|
+
c = add_constituent(ctxt, c, linkage, domain, w, w, name);
|
1215
|
+
|
1216
|
+
if (ctxt->wordtype[w] == QDTYPE)
|
1217
|
+
{
|
1218
|
+
/* Now find the finite verb to the right, start an S */
|
1219
|
+
/* Limit w2 to sentence length. */
|
1220
|
+
// for( w2=w+1; w2 < ctxt->r_limit-1; w2++ )
|
1221
|
+
for (w2 = w+1; w2 < rightmost; w2++)
|
1222
|
+
if ((ctxt->wordtype[w2] == STYPE) || (ctxt->wordtype[w2] == PTYPE)) break;
|
1223
|
+
|
1224
|
+
/* Adjust the right boundary of previous constituent */
|
1225
|
+
ctxt->constituent[c].right = w2 - 1;
|
1226
|
+
c = add_constituent(ctxt, c, linkage, domain, w2, rightmost, "S");
|
1227
|
+
}
|
1228
|
+
}
|
1229
|
+
|
1230
|
+
if (ctxt->constituent[c].domain_type == '\0')
|
1231
|
+
{
|
1232
|
+
err_ctxt ec;
|
1233
|
+
ec.sent = linkage->sent;
|
1234
|
+
err_msg(&ec, Error, "Error: no domain type assigned to constituent\n");
|
1235
|
+
}
|
1236
|
+
if (ctxt->constituent[c].start_link == NULL)
|
1237
|
+
{
|
1238
|
+
err_ctxt ec;
|
1239
|
+
ec.sent = linkage->sent;
|
1240
|
+
err_msg(&ec, Error, "Error: no type assigned to constituent\n");
|
1241
|
+
}
|
1242
|
+
}
|
1243
|
+
|
1244
|
+
numcon_subl = c - numcon_total;
|
1245
|
+
/* numcon_subl = handle_islands(linkage, numcon_total, numcon_subl); */
|
1246
|
+
|
1247
|
+
if (verbosity >= 2)
|
1248
|
+
printf("Constituents added at first stage for subl %d:\n",
|
1249
|
+
linkage->current);
|
1250
|
+
for (c = numcon_total; c < numcon_total + numcon_subl; c++)
|
1251
|
+
{
|
1252
|
+
print_constituent(ctxt, linkage, c);
|
1253
|
+
}
|
1254
|
+
|
1255
|
+
/* Opener case - generates S around main clause.
|
1256
|
+
(This must be done first; the S generated will be needed for
|
1257
|
+
later cases.) */
|
1258
|
+
numcon_subl =
|
1259
|
+
gen_comp(ctxt, linkage, numcon_total, numcon_subl, "SBAR", "S", "S", 5);
|
1260
|
+
|
1261
|
+
/* pp opener case */
|
1262
|
+
numcon_subl =
|
1263
|
+
gen_comp(ctxt, linkage, numcon_total, numcon_subl, "PP", "S", "S", 6);
|
1264
|
+
|
1265
|
+
/* participle opener case */
|
1266
|
+
numcon_subl =
|
1267
|
+
gen_comp(ctxt, linkage, numcon_total, numcon_subl, "S", "S", "S", 9);
|
1268
|
+
|
1269
|
+
/* Subject-phrase case; every main VP generates an S */
|
1270
|
+
numcon_subl =
|
1271
|
+
gen_comp(ctxt, linkage, numcon_total, numcon_subl, "VP", "S", "NP", 1);
|
1272
|
+
|
1273
|
+
/* Relative clause case; an SBAR generates a complement NP */
|
1274
|
+
numcon_subl =
|
1275
|
+
gen_comp(ctxt, linkage, numcon_total, numcon_subl, "SBAR", "NP", "NP", 3);
|
1276
|
+
|
1277
|
+
/* Participle modifier case */
|
1278
|
+
numcon_subl =
|
1279
|
+
gen_comp(ctxt, linkage, numcon_total, numcon_subl, "VP", "NP", "NP", 8);
|
1280
|
+
|
1281
|
+
/* PP modifying NP */
|
1282
|
+
numcon_subl =
|
1283
|
+
gen_comp(ctxt, linkage, numcon_total, numcon_subl, "PP", "NP", "NP", 8);
|
1284
|
+
|
1285
|
+
/* Appositive case */
|
1286
|
+
numcon_subl =
|
1287
|
+
gen_comp(ctxt, linkage, numcon_total, numcon_subl, "NP", "NP", "NP", 4);
|
1288
|
+
|
1289
|
+
/* S-V inversion case; an NP generates a complement VP */
|
1290
|
+
numcon_subl =
|
1291
|
+
gen_comp(ctxt, linkage, numcon_total, numcon_subl, "NP", "SINV", "VP", 7);
|
1292
|
+
|
1293
|
+
adjust_subordinate_clauses(ctxt, linkage, numcon_total, numcon_subl);
|
1294
|
+
for (c = numcon_total; c < numcon_total + numcon_subl; c++)
|
1295
|
+
{
|
1296
|
+
if ((ctxt->constituent[c].domain_type=='p') &&
|
1297
|
+
(strcmp(linkage->word[ctxt->constituent[c].left], ",")==0))
|
1298
|
+
{
|
1299
|
+
ctxt->constituent[c].left++;
|
1300
|
+
}
|
1301
|
+
}
|
1302
|
+
|
1303
|
+
/* Make sure the constituents are nested. If two constituents
|
1304
|
+
* are not nested: whichever constituent has the furthest left
|
1305
|
+
* boundary, shift that boundary rightwards to the left boundary
|
1306
|
+
* of the other one.
|
1307
|
+
*/
|
1308
|
+
while (1)
|
1309
|
+
{
|
1310
|
+
adjustment_made = 0;
|
1311
|
+
for (c = numcon_total; c < numcon_total + numcon_subl; c++)
|
1312
|
+
{
|
1313
|
+
for (c2 = numcon_total; c2 < numcon_total + numcon_subl; c2++)
|
1314
|
+
{
|
1315
|
+
if ((ctxt->constituent[c].left < ctxt->constituent[c2].left) &&
|
1316
|
+
(ctxt->constituent[c].right < ctxt->constituent[c2].right) &&
|
1317
|
+
(ctxt->constituent[c].right >= ctxt->constituent[c2].left))
|
1318
|
+
{
|
1319
|
+
/* We've found two overlapping constituents.
|
1320
|
+
If one is larger, except the smaller one
|
1321
|
+
includes an extra comma, adjust the smaller one
|
1322
|
+
to exclude the comma */
|
1323
|
+
|
1324
|
+
if ((strcmp(linkage->word[ctxt->constituent[c2].right], ",") == 0) ||
|
1325
|
+
(strcmp(linkage->word[ctxt->constituent[c2].right],
|
1326
|
+
"RIGHT-WALL") == 0))
|
1327
|
+
{
|
1328
|
+
if (verbosity >= 2)
|
1329
|
+
printf("Adjusting %d to fix comma overlap\n", c2);
|
1330
|
+
adjust_for_right_comma(ctxt, linkage, c2);
|
1331
|
+
adjustment_made = 1;
|
1332
|
+
}
|
1333
|
+
else if (strcmp(linkage->word[ctxt->constituent[c].left], ",") == 0)
|
1334
|
+
{
|
1335
|
+
if (verbosity >= 2)
|
1336
|
+
printf("Adjusting c %d to fix comma overlap\n", c);
|
1337
|
+
adjust_for_left_comma(ctxt, linkage, c);
|
1338
|
+
adjustment_made = 1;
|
1339
|
+
}
|
1340
|
+
else
|
1341
|
+
{
|
1342
|
+
if (verbosity >= 2)
|
1343
|
+
{
|
1344
|
+
err_ctxt ec;
|
1345
|
+
ec.sent = linkage->sent;
|
1346
|
+
err_msg(&ec, Warn,
|
1347
|
+
"Warning: the constituents aren't nested! "
|
1348
|
+
"Adjusting them. (%d, %d)\n", c, c2);
|
1349
|
+
}
|
1350
|
+
ctxt->constituent[c].left = ctxt->constituent[c2].left;
|
1351
|
+
}
|
1352
|
+
}
|
1353
|
+
}
|
1354
|
+
}
|
1355
|
+
if (adjustment_made == 0) break;
|
1356
|
+
}
|
1357
|
+
|
1358
|
+
#ifdef AUX_CODE_IS_DEAD
|
1359
|
+
/* The code here is ifdef-dead as it appears to be dead, as the computation it does
|
1360
|
+
* is immediately undone in the very next block.
|
1361
|
+
*/
|
1362
|
+
/* This labels certain words as auxiliaries (such as forms of "be"
|
1363
|
+
* with passives, forms of "have" wth past participles,
|
1364
|
+
* "to" with infinitives). These words start VP's which include
|
1365
|
+
* them. In Treebank I, these don't get printed unless they're part of an
|
1366
|
+
* andlist, in which case they get labeled "X". (this is why we need to
|
1367
|
+
* label them as "aux".) In Treebank II, however, they seem to be treated
|
1368
|
+
* just like other verbs, so the "aux" stuff isn't needed.
|
1369
|
+
*/
|
1370
|
+
for (c = numcon_total; c < numcon_total + numcon_subl; c++)
|
1371
|
+
{
|
1372
|
+
ctxt->constituent[c].subl = linkage->current;
|
1373
|
+
if (((ctxt->constituent[c].domain_type == 'v') &&
|
1374
|
+
(ctxt->wordtype[linkage_get_link_rword(linkage,
|
1375
|
+
ctxt->constituent[c].start_num)] == PTYPE))
|
1376
|
+
||
|
1377
|
+
((ctxt->constituent[c].domain_type == 't') &&
|
1378
|
+
(strcmp(ctxt->constituent[c].type, "VP") == 0)))
|
1379
|
+
{
|
1380
|
+
ctxt->constituent[c].aux = 1;
|
1381
|
+
}
|
1382
|
+
else
|
1383
|
+
{
|
1384
|
+
ctxt->constituent[c].aux = 0;
|
1385
|
+
}
|
1386
|
+
}
|
1387
|
+
#endif /* AUX_CODE_IS_DEAD */
|
1388
|
+
|
1389
|
+
if (MAXCONSTITUENTS <= numcon_total + numcon_subl)
|
1390
|
+
{
|
1391
|
+
err_ctxt ec;
|
1392
|
+
ec.sent = linkage->sent;
|
1393
|
+
err_msg(&ec, Error, "Error: Too many constituents (a2).\n");
|
1394
|
+
numcon_total = MAXCONSTITUENTS - numcon_subl;
|
1395
|
+
}
|
1396
|
+
for (c = numcon_total; c < numcon_total + numcon_subl; c++)
|
1397
|
+
{
|
1398
|
+
ctxt->constituent[c].subl = linkage->current;
|
1399
|
+
#ifdef AUX_CODE_IS_DEAD /* See comments at top */
|
1400
|
+
ctxt->constituent[c].aux = 0;
|
1401
|
+
#endif /* AUX_CODE_IS_DEAD */
|
1402
|
+
}
|
1403
|
+
|
1404
|
+
return numcon_subl;
|
1405
|
+
}
|
1406
|
+
|
1407
|
+
static char * exprint_constituent_structure(con_context_t *ctxt, Linkage linkage, int numcon_total)
|
1408
|
+
{
|
1409
|
+
int have_opened = 1;
|
1410
|
+
int c, w;
|
1411
|
+
int leftdone[MAXCONSTITUENTS];
|
1412
|
+
int rightdone[MAXCONSTITUENTS];
|
1413
|
+
int best, bestright, bestleft;
|
1414
|
+
Sentence sent;
|
1415
|
+
char s[100], * p;
|
1416
|
+
String * cs = string_new();
|
1417
|
+
|
1418
|
+
assert (numcon_total < MAXCONSTITUENTS, "Too many constituents (b)");
|
1419
|
+
sent = linkage_get_sentence(linkage);
|
1420
|
+
|
1421
|
+
for (c = 0; c < numcon_total; c++)
|
1422
|
+
{
|
1423
|
+
leftdone[c] = 0;
|
1424
|
+
rightdone[c] = 0;
|
1425
|
+
}
|
1426
|
+
|
1427
|
+
if (verbosity >= 2)
|
1428
|
+
printf("\n");
|
1429
|
+
|
1430
|
+
for (w = 1; w < linkage->num_words; w++)
|
1431
|
+
{
|
1432
|
+
/* Skip left wall; don't skip right wall, since it may
|
1433
|
+
have constituent boundaries */
|
1434
|
+
|
1435
|
+
while(1)
|
1436
|
+
{
|
1437
|
+
best = -1;
|
1438
|
+
bestright = -1;
|
1439
|
+
for (c = 0; c < numcon_total; c++)
|
1440
|
+
{
|
1441
|
+
if ((ctxt->constituent[c].left == w) &&
|
1442
|
+
(leftdone[c] == 0) && (ctxt->constituent[c].valid == 1) &&
|
1443
|
+
(ctxt->constituent[c].right >= bestright)) {
|
1444
|
+
best = c;
|
1445
|
+
bestright = ctxt->constituent[c].right;
|
1446
|
+
}
|
1447
|
+
}
|
1448
|
+
if (best == -1)
|
1449
|
+
break;
|
1450
|
+
|
1451
|
+
leftdone[best] = 1;
|
1452
|
+
/* have_open is a hack to avoid printing anything until
|
1453
|
+
* bracket is opened */
|
1454
|
+
if (w == 1) have_opened = 0;
|
1455
|
+
#ifdef AUX_CODE_IS_DEAD /* See comments at top */
|
1456
|
+
if (ctxt->constituent[best].aux == 1) continue;
|
1457
|
+
#endif /* AUX_CODE_IS_DEAD */
|
1458
|
+
have_opened = 1;
|
1459
|
+
append_string(cs, "%c%s ", OPEN_BRACKET, ctxt->constituent[best].type);
|
1460
|
+
}
|
1461
|
+
|
1462
|
+
/* Don't print out right wall */
|
1463
|
+
if (have_opened && (w < linkage->num_words - 1))
|
1464
|
+
{
|
1465
|
+
char *p;
|
1466
|
+
strcpy(s, sent->word[w].string);
|
1467
|
+
|
1468
|
+
/* Constituent processing will crash if the sentence contains
|
1469
|
+
* square brackets, so we have to do something ... replace
|
1470
|
+
* them with curly braces ... will have to do.
|
1471
|
+
*/
|
1472
|
+
p = strchr(s, OPEN_BRACKET);
|
1473
|
+
while(p)
|
1474
|
+
{
|
1475
|
+
*p = '{';
|
1476
|
+
p = strchr(p, OPEN_BRACKET);
|
1477
|
+
}
|
1478
|
+
|
1479
|
+
p = strchr(s, CLOSE_BRACKET);
|
1480
|
+
while(p)
|
1481
|
+
{
|
1482
|
+
*p = '}';
|
1483
|
+
p = strchr(p, CLOSE_BRACKET);
|
1484
|
+
}
|
1485
|
+
|
1486
|
+
/* Now, if the first character of the word was
|
1487
|
+
originally uppercase, we put it back that way */
|
1488
|
+
if (sent->word[w].firstupper == 1)
|
1489
|
+
upcase_utf8_str(s, s, MAX_WORD);
|
1490
|
+
append_string(cs, "%s ", s);
|
1491
|
+
}
|
1492
|
+
|
1493
|
+
while(1)
|
1494
|
+
{
|
1495
|
+
best = -1;
|
1496
|
+
bestleft = -1;
|
1497
|
+
for(c = 0; c < numcon_total; c++)
|
1498
|
+
{
|
1499
|
+
if ((ctxt->constituent[c].right == w) &&
|
1500
|
+
(rightdone[c] == 0) && (ctxt->constituent[c].valid == 1) &&
|
1501
|
+
(ctxt->constituent[c].left > bestleft)) {
|
1502
|
+
best = c;
|
1503
|
+
bestleft = ctxt->constituent[c].left;
|
1504
|
+
}
|
1505
|
+
}
|
1506
|
+
if (best == -1)
|
1507
|
+
break;
|
1508
|
+
rightdone[best] = 1;
|
1509
|
+
#ifdef AUX_CODE_IS_DEAD /* See comments at top */
|
1510
|
+
if (ctxt->constituent[best].aux == 1)
|
1511
|
+
continue;
|
1512
|
+
#endif /* AUX_CODE_IS_DEAD */
|
1513
|
+
append_string(cs, "%s%c ", ctxt->constituent[best].type, CLOSE_BRACKET);
|
1514
|
+
}
|
1515
|
+
}
|
1516
|
+
|
1517
|
+
append_string(cs, "\n");
|
1518
|
+
p = string_copy(cs);
|
1519
|
+
string_delete(cs);
|
1520
|
+
return p;
|
1521
|
+
}
|
1522
|
+
|
1523
|
+
static char * do_print_flat_constituents(con_context_t *ctxt, Linkage linkage)
|
1524
|
+
{
|
1525
|
+
int num_words;
|
1526
|
+
Sentence sent;
|
1527
|
+
Postprocessor * pp;
|
1528
|
+
int s, numcon_total, numcon_subl, num_subl;
|
1529
|
+
char * q;
|
1530
|
+
|
1531
|
+
sent = linkage_get_sentence(linkage);
|
1532
|
+
ctxt->phrase_ss = string_set_create();
|
1533
|
+
pp = linkage->sent->dict->constituent_pp;
|
1534
|
+
numcon_total = 0;
|
1535
|
+
|
1536
|
+
count_words_used(ctxt, linkage);
|
1537
|
+
|
1538
|
+
num_subl = linkage->num_sublinkages;
|
1539
|
+
if (num_subl > MAXSUBL)
|
1540
|
+
{
|
1541
|
+
num_subl = MAXSUBL;
|
1542
|
+
if (verbosity >= 2)
|
1543
|
+
printf("Number of sublinkages exceeds maximum: only considering first %d sublinkages\n", MAXSUBL);
|
1544
|
+
}
|
1545
|
+
|
1546
|
+
if (linkage->unionized == 1 && num_subl > 1) num_subl--;
|
1547
|
+
for (s = 0; s < num_subl; s++)
|
1548
|
+
{
|
1549
|
+
linkage_set_current_sublinkage(linkage, s);
|
1550
|
+
linkage_post_process(linkage, pp);
|
1551
|
+
num_words = linkage_get_num_words(linkage);
|
1552
|
+
generate_misc_word_info(ctxt, linkage);
|
1553
|
+
numcon_subl = read_constituents_from_domains(ctxt, linkage, numcon_total, s);
|
1554
|
+
numcon_total = numcon_total + numcon_subl;
|
1555
|
+
if (MAXCONSTITUENTS <= numcon_total)
|
1556
|
+
{
|
1557
|
+
err_ctxt ec;
|
1558
|
+
ec.sent = linkage->sent;
|
1559
|
+
err_msg(&ec, Error, "Error: Too many constituents (c).\n");
|
1560
|
+
numcon_total = MAXCONSTITUENTS-1;
|
1561
|
+
break;
|
1562
|
+
}
|
1563
|
+
}
|
1564
|
+
numcon_total = merge_constituents(ctxt, linkage, numcon_total);
|
1565
|
+
if (MAXCONSTITUENTS <= numcon_total)
|
1566
|
+
{
|
1567
|
+
err_ctxt ec;
|
1568
|
+
ec.sent = linkage->sent;
|
1569
|
+
err_msg(&ec, Error, "Error: Too many constituents (d).\n");
|
1570
|
+
numcon_total = MAXCONSTITUENTS-1;
|
1571
|
+
}
|
1572
|
+
numcon_total = last_minute_fixes(ctxt, linkage, numcon_total);
|
1573
|
+
if (MAXCONSTITUENTS <= numcon_total)
|
1574
|
+
{
|
1575
|
+
err_ctxt ec;
|
1576
|
+
ec.sent = linkage->sent;
|
1577
|
+
err_msg(&ec, Error, "Error: Too many constituents (e).\n");
|
1578
|
+
numcon_total = MAXCONSTITUENTS-1;
|
1579
|
+
}
|
1580
|
+
q = exprint_constituent_structure(ctxt, linkage, numcon_total);
|
1581
|
+
string_set_delete(ctxt->phrase_ss);
|
1582
|
+
ctxt->phrase_ss = NULL;
|
1583
|
+
return q;
|
1584
|
+
}
|
1585
|
+
|
1586
|
+
static char * print_flat_constituents(Linkage linkage)
|
1587
|
+
{
|
1588
|
+
/* In principle, the ctxt could be allocated on stack, instead of
|
1589
|
+
* with malloc(). However, The java6 jvm (and MS Windows jvm's)
|
1590
|
+
* gives JNI clients only a small amount of stack space. Alloc'ing
|
1591
|
+
* this (rather large) structure on stack will blow up the JVM.
|
1592
|
+
* This was discovered only after much work. Bummer.
|
1593
|
+
*/
|
1594
|
+
char * p;
|
1595
|
+
con_context_t *ctxt = (con_context_t *) malloc (sizeof(con_context_t));
|
1596
|
+
memset(ctxt, 0, sizeof(con_context_t));
|
1597
|
+
p = do_print_flat_constituents(ctxt, linkage);
|
1598
|
+
free(ctxt);
|
1599
|
+
return p;
|
1600
|
+
}
|
1601
|
+
|
1602
|
+
static CType token_type (char *token)
|
1603
|
+
{
|
1604
|
+
if ((token[0] == OPEN_BRACKET) && (strlen(token) > 1))
|
1605
|
+
return OPEN_TOK;
|
1606
|
+
if ((strlen(token) > 1) && (token[strlen(token) - 1] == CLOSE_BRACKET))
|
1607
|
+
return CLOSE_TOK;
|
1608
|
+
return WORD_TOK;
|
1609
|
+
}
|
1610
|
+
|
1611
|
+
static CNode * make_CNode(char *q)
|
1612
|
+
{
|
1613
|
+
CNode * cn;
|
1614
|
+
cn = (CNode *) exalloc(sizeof(CNode));
|
1615
|
+
cn->label = (char *) exalloc(sizeof(char)*(strlen(q)+1));
|
1616
|
+
strcpy(cn->label, q);
|
1617
|
+
cn->child = cn->next = (CNode *) NULL;
|
1618
|
+
cn->next = (CNode *) NULL;
|
1619
|
+
cn->start = cn->end = -1;
|
1620
|
+
return cn;
|
1621
|
+
}
|
1622
|
+
|
1623
|
+
static CNode * parse_string(CNode * n, char **saveptr)
|
1624
|
+
{
|
1625
|
+
char *q;
|
1626
|
+
CNode *m, *last_child=NULL;
|
1627
|
+
|
1628
|
+
while ((q = strtok_r(NULL, " ", saveptr))) {
|
1629
|
+
switch (token_type(q)) {
|
1630
|
+
case CLOSE_TOK :
|
1631
|
+
q[strlen(q)-1]='\0';
|
1632
|
+
assert(strcmp(q, n->label)==0,
|
1633
|
+
"Constituent tree: Labels do not match.");
|
1634
|
+
return n;
|
1635
|
+
break;
|
1636
|
+
case OPEN_TOK:
|
1637
|
+
m = make_CNode(q+1);
|
1638
|
+
m = parse_string(m, saveptr);
|
1639
|
+
break;
|
1640
|
+
case WORD_TOK:
|
1641
|
+
m = make_CNode(q);
|
1642
|
+
break;
|
1643
|
+
default:
|
1644
|
+
assert(0, "Constituent tree: Illegal token type");
|
1645
|
+
}
|
1646
|
+
if (n->child == NULL) {
|
1647
|
+
last_child = n->child = m;
|
1648
|
+
}
|
1649
|
+
else {
|
1650
|
+
last_child->next = m;
|
1651
|
+
last_child = m;
|
1652
|
+
}
|
1653
|
+
}
|
1654
|
+
assert(0, "Constituent tree: Constituent did not close");
|
1655
|
+
return NULL;
|
1656
|
+
}
|
1657
|
+
|
1658
|
+
static void print_tree(String * cs, int indent, CNode * n, int o1, int o2)
|
1659
|
+
{
|
1660
|
+
int i, child_offset;
|
1661
|
+
CNode * m;
|
1662
|
+
|
1663
|
+
if (n == NULL) return;
|
1664
|
+
|
1665
|
+
if (indent)
|
1666
|
+
for (i = 0; i < o1; ++i)
|
1667
|
+
append_string(cs, " ");
|
1668
|
+
append_string(cs, "(%s ", n->label);
|
1669
|
+
child_offset = o2 + strlen(n->label) + 2;
|
1670
|
+
|
1671
|
+
for (m = n->child; m != NULL; m = m->next)
|
1672
|
+
{
|
1673
|
+
if (m->child == NULL)
|
1674
|
+
{
|
1675
|
+
char * p;
|
1676
|
+
/* If the original string has left or right parens in it,
|
1677
|
+
* the printed string will be messed up by these ...
|
1678
|
+
* so replace them by curly braces. What else can one do?
|
1679
|
+
*/
|
1680
|
+
p = strchr(m->label, '(');
|
1681
|
+
while(p)
|
1682
|
+
{
|
1683
|
+
*p = '{';
|
1684
|
+
p = strchr(p, '(');
|
1685
|
+
}
|
1686
|
+
|
1687
|
+
p = strchr(m->label, ')');
|
1688
|
+
while(p)
|
1689
|
+
{
|
1690
|
+
*p = '}';
|
1691
|
+
p = strchr(p, ')');
|
1692
|
+
}
|
1693
|
+
|
1694
|
+
append_string(cs, "%s", m->label);
|
1695
|
+
if ((m->next != NULL) && (m->next->child == NULL))
|
1696
|
+
append_string(cs, " ");
|
1697
|
+
}
|
1698
|
+
else
|
1699
|
+
{
|
1700
|
+
if (m != n->child)
|
1701
|
+
{
|
1702
|
+
if (indent) append_string(cs, "\n");
|
1703
|
+
else append_string(cs, " ");
|
1704
|
+
print_tree(cs, indent, m, child_offset, child_offset);
|
1705
|
+
}
|
1706
|
+
else
|
1707
|
+
{
|
1708
|
+
print_tree(cs, indent, m, 0, child_offset);
|
1709
|
+
}
|
1710
|
+
if ((m->next != NULL) && (m->next->child == NULL))
|
1711
|
+
{
|
1712
|
+
if (indent)
|
1713
|
+
{
|
1714
|
+
append_string(cs, "\n");
|
1715
|
+
for (i = 0; i < child_offset; ++i)
|
1716
|
+
append_string(cs, " ");
|
1717
|
+
}
|
1718
|
+
else append_string(cs, " ");
|
1719
|
+
}
|
1720
|
+
}
|
1721
|
+
}
|
1722
|
+
append_string(cs, ")");
|
1723
|
+
}
|
1724
|
+
|
1725
|
+
static int assign_spans(CNode * n, int start) {
|
1726
|
+
int num_words=0;
|
1727
|
+
CNode * m=NULL;
|
1728
|
+
if (n==NULL) return 0;
|
1729
|
+
n->start = start;
|
1730
|
+
if (n->child == NULL) {
|
1731
|
+
n->end = start;
|
1732
|
+
return 1;
|
1733
|
+
}
|
1734
|
+
else {
|
1735
|
+
for (m=n->child; m!=NULL; m=m->next) {
|
1736
|
+
num_words += assign_spans(m, start+num_words);
|
1737
|
+
}
|
1738
|
+
n->end = start+num_words-1;
|
1739
|
+
}
|
1740
|
+
return num_words;
|
1741
|
+
}
|
1742
|
+
|
1743
|
+
CNode * linkage_constituent_tree(Linkage linkage)
|
1744
|
+
{
|
1745
|
+
char *p, *q, *saveptr;
|
1746
|
+
int len;
|
1747
|
+
CNode * root;
|
1748
|
+
|
1749
|
+
p = print_flat_constituents(linkage);
|
1750
|
+
|
1751
|
+
len = strlen(p);
|
1752
|
+
q = strtok_r(p, " ", &saveptr);
|
1753
|
+
assert(token_type(q) == OPEN_TOK, "Illegal beginning of string");
|
1754
|
+
root = make_CNode(q+1);
|
1755
|
+
root = parse_string(root, &saveptr);
|
1756
|
+
assign_spans(root, 0);
|
1757
|
+
exfree(p, sizeof(char)*(len+1));
|
1758
|
+
return root;
|
1759
|
+
}
|
1760
|
+
|
1761
|
+
void linkage_free_constituent_tree(CNode * n)
|
1762
|
+
{
|
1763
|
+
CNode *m, *x;
|
1764
|
+
for (m=n->child; m!=NULL; m=x) {
|
1765
|
+
x=m->next;
|
1766
|
+
linkage_free_constituent_tree(m);
|
1767
|
+
}
|
1768
|
+
exfree(n->label, sizeof(char)*(strlen(n->label)+1));
|
1769
|
+
exfree(n, sizeof(CNode));
|
1770
|
+
}
|
1771
|
+
|
1772
|
+
/**
|
1773
|
+
* Print out the constituent tree.
|
1774
|
+
* mode 1: treebank-style constituent tree
|
1775
|
+
* mode 2: flat, bracketed tree [A like [B this B] A]
|
1776
|
+
* mode 3: flat, treebank-style tree (A like (B this) )
|
1777
|
+
*/
|
1778
|
+
char * linkage_print_constituent_tree(Linkage linkage, int mode)
|
1779
|
+
{
|
1780
|
+
String * cs;
|
1781
|
+
CNode * root;
|
1782
|
+
char * p;
|
1783
|
+
|
1784
|
+
if ((mode == 0) || (linkage->sent->dict->constituent_pp == NULL))
|
1785
|
+
{
|
1786
|
+
return NULL;
|
1787
|
+
}
|
1788
|
+
else if (mode == 1 || mode == 3)
|
1789
|
+
{
|
1790
|
+
cs = string_new();
|
1791
|
+
root = linkage_constituent_tree(linkage);
|
1792
|
+
print_tree(cs, (mode==1), root, 0, 0);
|
1793
|
+
linkage_free_constituent_tree(root);
|
1794
|
+
append_string(cs, "\n");
|
1795
|
+
p = string_copy(cs);
|
1796
|
+
string_delete(cs);
|
1797
|
+
return p;
|
1798
|
+
}
|
1799
|
+
else if (mode == 2)
|
1800
|
+
{
|
1801
|
+
return print_flat_constituents(linkage);
|
1802
|
+
}
|
1803
|
+
assert(0, "Illegal mode in linkage_print_constituent_tree");
|
1804
|
+
return NULL;
|
1805
|
+
}
|
1806
|
+
|
1807
|
+
void linkage_free_constituent_tree_str(char * s)
|
1808
|
+
{
|
1809
|
+
exfree(s, strlen(s)+1);
|
1810
|
+
}
|
1811
|
+
|
1812
|
+
const char * linkage_constituent_node_get_label(const CNode *n)
|
1813
|
+
{
|
1814
|
+
return n->label;
|
1815
|
+
}
|
1816
|
+
|
1817
|
+
|
1818
|
+
CNode * linkage_constituent_node_get_child(const CNode *n)
|
1819
|
+
{
|
1820
|
+
return n->child;
|
1821
|
+
}
|
1822
|
+
|
1823
|
+
CNode * linkage_constituent_node_get_next(const CNode *n)
|
1824
|
+
{
|
1825
|
+
return n->next;
|
1826
|
+
}
|
1827
|
+
|
1828
|
+
int linkage_constituent_node_get_start(const CNode *n)
|
1829
|
+
{
|
1830
|
+
return n->start;
|
1831
|
+
}
|
1832
|
+
|
1833
|
+
int linkage_constituent_node_get_end(const CNode *n)
|
1834
|
+
{
|
1835
|
+
return n->end;
|
1836
|
+
}
|