grammar_cop 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 +8 -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/ext/.DS_Store +0 -0
- data/ext/link_grammar/.DS_Store +0 -0
- data/ext/link_grammar/extconf.rb +2 -0
- data/ext/link_grammar/link-grammar/.DS_Store +0 -0
- data/ext/link_grammar/link-grammar/.deps/analyze-linkage.Plo +198 -0
- data/ext/link_grammar/link-grammar/.deps/and.Plo +202 -0
- data/ext/link_grammar/link-grammar/.deps/api.Plo +244 -0
- data/ext/link_grammar/link-grammar/.deps/build-disjuncts.Plo +212 -0
- data/ext/link_grammar/link-grammar/.deps/command-line.Plo +201 -0
- data/ext/link_grammar/link-grammar/.deps/constituents.Plo +201 -0
- data/ext/link_grammar/link-grammar/.deps/count.Plo +202 -0
- data/ext/link_grammar/link-grammar/.deps/disjunct-utils.Plo +126 -0
- data/ext/link_grammar/link-grammar/.deps/disjuncts.Plo +123 -0
- data/ext/link_grammar/link-grammar/.deps/error.Plo +121 -0
- data/ext/link_grammar/link-grammar/.deps/expand.Plo +133 -0
- data/ext/link_grammar/link-grammar/.deps/extract-links.Plo +198 -0
- data/ext/link_grammar/link-grammar/.deps/fast-match.Plo +200 -0
- data/ext/link_grammar/link-grammar/.deps/idiom.Plo +200 -0
- data/ext/link_grammar/link-grammar/.deps/jni-client.Plo +217 -0
- data/ext/link_grammar/link-grammar/.deps/link-parser.Po +1 -0
- data/ext/link_grammar/link-grammar/.deps/massage.Plo +202 -0
- data/ext/link_grammar/link-grammar/.deps/post-process.Plo +202 -0
- data/ext/link_grammar/link-grammar/.deps/pp_knowledge.Plo +202 -0
- data/ext/link_grammar/link-grammar/.deps/pp_lexer.Plo +201 -0
- data/ext/link_grammar/link-grammar/.deps/pp_linkset.Plo +200 -0
- data/ext/link_grammar/link-grammar/.deps/prefix.Plo +102 -0
- data/ext/link_grammar/link-grammar/.deps/preparation.Plo +202 -0
- data/ext/link_grammar/link-grammar/.deps/print-util.Plo +200 -0
- data/ext/link_grammar/link-grammar/.deps/print.Plo +201 -0
- data/ext/link_grammar/link-grammar/.deps/prune.Plo +202 -0
- data/ext/link_grammar/link-grammar/.deps/read-dict.Plo +223 -0
- data/ext/link_grammar/link-grammar/.deps/read-regex.Plo +123 -0
- data/ext/link_grammar/link-grammar/.deps/regex-morph.Plo +131 -0
- data/ext/link_grammar/link-grammar/.deps/resources.Plo +203 -0
- data/ext/link_grammar/link-grammar/.deps/spellcheck-aspell.Plo +1 -0
- data/ext/link_grammar/link-grammar/.deps/spellcheck-hun.Plo +115 -0
- data/ext/link_grammar/link-grammar/.deps/string-set.Plo +198 -0
- data/ext/link_grammar/link-grammar/.deps/tokenize.Plo +160 -0
- data/ext/link_grammar/link-grammar/.deps/utilities.Plo +222 -0
- data/ext/link_grammar/link-grammar/.deps/word-file.Plo +201 -0
- data/ext/link_grammar/link-grammar/.deps/word-utils.Plo +212 -0
- data/ext/link_grammar/link-grammar/.libs/analyze-linkage.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/and.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/api.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/build-disjuncts.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/command-line.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/constituents.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/count.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/disjunct-utils.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/disjuncts.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/error.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/expand.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/extract-links.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/fast-match.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/idiom.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/jni-client.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/liblink-grammar-java-symbols.expsym +31 -0
- data/ext/link_grammar/link-grammar/.libs/liblink-grammar-java.4.dylib +0 -0
- data/ext/link_grammar/link-grammar/.libs/liblink-grammar-java.4.dylib.dSYM/Contents/Info.plist +20 -0
- data/ext/link_grammar/link-grammar/.libs/liblink-grammar-java.4.dylib.dSYM/Contents/Resources/DWARF/liblink-grammar-java.4.dylib +0 -0
- data/ext/link_grammar/link-grammar/.libs/liblink-grammar-java.a +0 -0
- data/ext/link_grammar/link-grammar/.libs/liblink-grammar-java.dylib +0 -0
- data/ext/link_grammar/link-grammar/.libs/liblink-grammar-symbols.expsym +194 -0
- data/ext/link_grammar/link-grammar/.libs/liblink-grammar.4.dylib +0 -0
- data/ext/link_grammar/link-grammar/.libs/liblink-grammar.4.dylib.dSYM/Contents/Info.plist +20 -0
- data/ext/link_grammar/link-grammar/.libs/liblink-grammar.4.dylib.dSYM/Contents/Resources/DWARF/liblink-grammar.4.dylib +0 -0
- data/ext/link_grammar/link-grammar/.libs/liblink-grammar.a +0 -0
- data/ext/link_grammar/link-grammar/.libs/liblink-grammar.dylib +0 -0
- data/ext/link_grammar/link-grammar/.libs/liblink-grammar.la +41 -0
- data/ext/link_grammar/link-grammar/.libs/liblink-grammar.lai +41 -0
- data/ext/link_grammar/link-grammar/.libs/massage.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/post-process.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/pp_knowledge.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/pp_lexer.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/pp_linkset.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/prefix.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/preparation.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/print-util.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/print.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/prune.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/read-dict.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/read-regex.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/regex-morph.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/resources.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/spellcheck-aspell.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/spellcheck-hun.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/string-set.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/tokenize.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/utilities.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/word-file.o +0 -0
- data/ext/link_grammar/link-grammar/.libs/word-utils.o +0 -0
- data/ext/link_grammar/link-grammar/Makefile +900 -0
- data/ext/link_grammar/link-grammar/Makefile.am +202 -0
- data/ext/link_grammar/link-grammar/Makefile.in +900 -0
- data/ext/link_grammar/link-grammar/analyze-linkage.c +1317 -0
- data/ext/link_grammar/link-grammar/analyze-linkage.h +24 -0
- data/ext/link_grammar/link-grammar/and.c +1603 -0
- data/ext/link_grammar/link-grammar/and.h +27 -0
- data/ext/link_grammar/link-grammar/api-structures.h +362 -0
- data/ext/link_grammar/link-grammar/api-types.h +72 -0
- data/ext/link_grammar/link-grammar/api.c +1887 -0
- data/ext/link_grammar/link-grammar/api.h +96 -0
- data/ext/link_grammar/link-grammar/autoit/.DS_Store +0 -0
- data/ext/link_grammar/link-grammar/autoit/README +10 -0
- data/ext/link_grammar/link-grammar/autoit/_LGTest.au3 +22 -0
- data/ext/link_grammar/link-grammar/autoit/_LinkGrammar.au3 +545 -0
- data/ext/link_grammar/link-grammar/build-disjuncts.c +487 -0
- data/ext/link_grammar/link-grammar/build-disjuncts.h +21 -0
- data/ext/link_grammar/link-grammar/command-line.c +458 -0
- data/ext/link_grammar/link-grammar/command-line.h +15 -0
- data/ext/link_grammar/link-grammar/constituents.c +1836 -0
- data/ext/link_grammar/link-grammar/constituents.h +26 -0
- data/ext/link_grammar/link-grammar/corpus/.DS_Store +0 -0
- data/ext/link_grammar/link-grammar/corpus/.deps/cluster.Plo +1 -0
- data/ext/link_grammar/link-grammar/corpus/.deps/corpus.Plo +1 -0
- data/ext/link_grammar/link-grammar/corpus/Makefile +527 -0
- data/ext/link_grammar/link-grammar/corpus/Makefile.am +46 -0
- data/ext/link_grammar/link-grammar/corpus/Makefile.in +527 -0
- data/ext/link_grammar/link-grammar/corpus/README +17 -0
- data/ext/link_grammar/link-grammar/corpus/cluster.c +286 -0
- data/ext/link_grammar/link-grammar/corpus/cluster.h +32 -0
- data/ext/link_grammar/link-grammar/corpus/corpus.c +483 -0
- data/ext/link_grammar/link-grammar/corpus/corpus.h +46 -0
- data/ext/link_grammar/link-grammar/count.c +828 -0
- data/ext/link_grammar/link-grammar/count.h +25 -0
- data/ext/link_grammar/link-grammar/disjunct-utils.c +261 -0
- data/ext/link_grammar/link-grammar/disjunct-utils.h +27 -0
- data/ext/link_grammar/link-grammar/disjuncts.c +138 -0
- data/ext/link_grammar/link-grammar/disjuncts.h +13 -0
- data/ext/link_grammar/link-grammar/error.c +92 -0
- data/ext/link_grammar/link-grammar/error.h +35 -0
- data/ext/link_grammar/link-grammar/expand.c +67 -0
- data/ext/link_grammar/link-grammar/expand.h +13 -0
- data/ext/link_grammar/link-grammar/externs.h +22 -0
- data/ext/link_grammar/link-grammar/extract-links.c +625 -0
- data/ext/link_grammar/link-grammar/extract-links.h +16 -0
- data/ext/link_grammar/link-grammar/fast-match.c +309 -0
- data/ext/link_grammar/link-grammar/fast-match.h +17 -0
- data/ext/link_grammar/link-grammar/idiom.c +373 -0
- data/ext/link_grammar/link-grammar/idiom.h +15 -0
- data/ext/link_grammar/link-grammar/jni-client.c +779 -0
- data/ext/link_grammar/link-grammar/jni-client.h +236 -0
- data/ext/link_grammar/link-grammar/liblink-grammar-java.la +42 -0
- data/ext/link_grammar/link-grammar/liblink-grammar.la +41 -0
- data/ext/link_grammar/link-grammar/link-features.h +37 -0
- data/ext/link_grammar/link-grammar/link-features.h.in +37 -0
- data/ext/link_grammar/link-grammar/link-grammar-java.def +31 -0
- data/ext/link_grammar/link-grammar/link-grammar.def +194 -0
- data/ext/link_grammar/link-grammar/link-includes.h +465 -0
- data/ext/link_grammar/link-grammar/link-parser.c +849 -0
- data/ext/link_grammar/link-grammar/massage.c +329 -0
- data/ext/link_grammar/link-grammar/massage.h +13 -0
- data/ext/link_grammar/link-grammar/post-process.c +1113 -0
- data/ext/link_grammar/link-grammar/post-process.h +45 -0
- data/ext/link_grammar/link-grammar/pp_knowledge.c +376 -0
- data/ext/link_grammar/link-grammar/pp_knowledge.h +14 -0
- data/ext/link_grammar/link-grammar/pp_lexer.c +1920 -0
- data/ext/link_grammar/link-grammar/pp_lexer.h +19 -0
- data/ext/link_grammar/link-grammar/pp_linkset.c +158 -0
- data/ext/link_grammar/link-grammar/pp_linkset.h +20 -0
- data/ext/link_grammar/link-grammar/prefix.c +482 -0
- data/ext/link_grammar/link-grammar/prefix.h +139 -0
- data/ext/link_grammar/link-grammar/preparation.c +412 -0
- data/ext/link_grammar/link-grammar/preparation.h +20 -0
- data/ext/link_grammar/link-grammar/print-util.c +87 -0
- data/ext/link_grammar/link-grammar/print-util.h +32 -0
- data/ext/link_grammar/link-grammar/print.c +1085 -0
- data/ext/link_grammar/link-grammar/print.h +16 -0
- data/ext/link_grammar/link-grammar/prune.c +1864 -0
- data/ext/link_grammar/link-grammar/prune.h +17 -0
- data/ext/link_grammar/link-grammar/read-dict.c +1785 -0
- data/ext/link_grammar/link-grammar/read-dict.h +29 -0
- data/ext/link_grammar/link-grammar/read-regex.c +161 -0
- data/ext/link_grammar/link-grammar/read-regex.h +12 -0
- data/ext/link_grammar/link-grammar/regex-morph.c +126 -0
- data/ext/link_grammar/link-grammar/regex-morph.h +17 -0
- data/ext/link_grammar/link-grammar/resources.c +180 -0
- data/ext/link_grammar/link-grammar/resources.h +23 -0
- data/ext/link_grammar/link-grammar/sat-solver/.DS_Store +0 -0
- data/ext/link_grammar/link-grammar/sat-solver/.deps/fast-sprintf.Plo +1 -0
- data/ext/link_grammar/link-grammar/sat-solver/.deps/sat-encoder.Plo +1 -0
- data/ext/link_grammar/link-grammar/sat-solver/.deps/util.Plo +1 -0
- data/ext/link_grammar/link-grammar/sat-solver/.deps/variables.Plo +1 -0
- data/ext/link_grammar/link-grammar/sat-solver/.deps/word-tag.Plo +1 -0
- data/ext/link_grammar/link-grammar/sat-solver/Makefile +527 -0
- data/ext/link_grammar/link-grammar/sat-solver/Makefile.am +29 -0
- data/ext/link_grammar/link-grammar/sat-solver/Makefile.in +527 -0
- data/ext/link_grammar/link-grammar/sat-solver/clock.hpp +33 -0
- data/ext/link_grammar/link-grammar/sat-solver/fast-sprintf.cpp +26 -0
- data/ext/link_grammar/link-grammar/sat-solver/fast-sprintf.hpp +7 -0
- data/ext/link_grammar/link-grammar/sat-solver/guiding.hpp +244 -0
- data/ext/link_grammar/link-grammar/sat-solver/matrix-ut.hpp +79 -0
- data/ext/link_grammar/link-grammar/sat-solver/sat-encoder.cpp +2811 -0
- data/ext/link_grammar/link-grammar/sat-solver/sat-encoder.h +11 -0
- data/ext/link_grammar/link-grammar/sat-solver/sat-encoder.hpp +381 -0
- data/ext/link_grammar/link-grammar/sat-solver/trie.hpp +118 -0
- data/ext/link_grammar/link-grammar/sat-solver/util.cpp +23 -0
- data/ext/link_grammar/link-grammar/sat-solver/util.hpp +14 -0
- data/ext/link_grammar/link-grammar/sat-solver/variables.cpp +5 -0
- data/ext/link_grammar/link-grammar/sat-solver/variables.hpp +829 -0
- data/ext/link_grammar/link-grammar/sat-solver/word-tag.cpp +159 -0
- data/ext/link_grammar/link-grammar/sat-solver/word-tag.hpp +162 -0
- data/ext/link_grammar/link-grammar/spellcheck-aspell.c +148 -0
- data/ext/link_grammar/link-grammar/spellcheck-hun.c +136 -0
- data/ext/link_grammar/link-grammar/spellcheck.h +34 -0
- data/ext/link_grammar/link-grammar/string-set.c +169 -0
- data/ext/link_grammar/link-grammar/string-set.h +16 -0
- data/ext/link_grammar/link-grammar/structures.h +498 -0
- data/ext/link_grammar/link-grammar/tokenize.c +1049 -0
- data/ext/link_grammar/link-grammar/tokenize.h +15 -0
- data/ext/link_grammar/link-grammar/utilities.c +847 -0
- data/ext/link_grammar/link-grammar/utilities.h +281 -0
- data/ext/link_grammar/link-grammar/word-file.c +124 -0
- data/ext/link_grammar/link-grammar/word-file.h +15 -0
- data/ext/link_grammar/link-grammar/word-utils.c +526 -0
- data/ext/link_grammar/link-grammar/word-utils.h +152 -0
- data/ext/link_grammar/link_grammar.c +202 -0
- data/ext/link_grammar/link_grammar.h +99 -0
- data/grammar_cop.gemspec +24 -0
- data/lib/.DS_Store +0 -0
- data/lib/grammar_cop.rb +9 -0
- data/lib/grammar_cop/.DS_Store +0 -0
- data/lib/grammar_cop/dictionary.rb +19 -0
- data/lib/grammar_cop/linkage.rb +30 -0
- data/lib/grammar_cop/parse_options.rb +32 -0
- data/lib/grammar_cop/sentence.rb +36 -0
- data/lib/grammar_cop/version.rb +3 -0
- data/test/.DS_Store +0 -0
- data/test/grammar_cop_test.rb +27 -0
- metadata +407 -0
@@ -0,0 +1,46 @@
|
|
1
|
+
/*
|
2
|
+
* corpus.h
|
3
|
+
*
|
4
|
+
* Data for corpus statistics, used to provide a parse ranking
|
5
|
+
* to drive the SAT solver, as well as parse ranking with the
|
6
|
+
* ordinary solver.
|
7
|
+
*
|
8
|
+
* Copyright (c) 2008, 2009 Linas Vepstas <linasvepstas@gmail.com>
|
9
|
+
*/
|
10
|
+
|
11
|
+
#ifndef _LINKGRAMMAR_CORPUS_H
|
12
|
+
#define _LINKGRAMMAR_CORPUS_H
|
13
|
+
|
14
|
+
#ifdef USE_CORPUS
|
15
|
+
|
16
|
+
#include "../api-types.h"
|
17
|
+
#include "../link-includes.h"
|
18
|
+
|
19
|
+
Corpus * lg_corpus_new(void);
|
20
|
+
void lg_corpus_delete(Corpus *);
|
21
|
+
|
22
|
+
void lg_corpus_score(Sentence, Linkage_info *);
|
23
|
+
double lg_corpus_disjunct_score(Linkage linkage, int w);
|
24
|
+
void lg_corpus_linkage_senses(Linkage);
|
25
|
+
|
26
|
+
Sense * lg_get_word_sense(Linkage_info *, int word);
|
27
|
+
Sense * lg_sense_next(Sense *);
|
28
|
+
int lg_sense_get_index(Sense *);
|
29
|
+
const char * lg_sense_get_subscripted_word(Sense *);
|
30
|
+
const char * lg_sense_get_disjunct(Sense *);
|
31
|
+
const char * lg_sense_get_sense(Sense *);
|
32
|
+
double lg_sense_get_score(Sense *);
|
33
|
+
void lg_sense_delete(Linkage_info *);
|
34
|
+
|
35
|
+
#else /* USE_CORPUS */
|
36
|
+
|
37
|
+
static inline void lg_corpus_score(Sentence s, Linkage_info *li) {}
|
38
|
+
static inline void lg_corpus_linkage_senses(Linkage l) {}
|
39
|
+
static inline Sense * lg_get_word_sense(Linkage_info *lif, int word) { return NULL; }
|
40
|
+
static inline Sense * lg_sense_next(Sense *s ) {return NULL; }
|
41
|
+
static inline const char * lg_sense_get_sense(Sense *s) { return NULL; }
|
42
|
+
static inline double lg_sense_get_score(Sense *s) { return 0.0; }
|
43
|
+
static inline double lg_corpus_disjunct_score(Linkage linkage, int w) { return 998.0; }
|
44
|
+
#endif /* USE_CORPUS */
|
45
|
+
|
46
|
+
#endif /* _LINKGRAMMAR_CORPUS_H */
|
@@ -0,0 +1,828 @@
|
|
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 "api.h"
|
15
|
+
#include "disjunct-utils.h"
|
16
|
+
|
17
|
+
/* This file contains the exhaustive search algorithm. */
|
18
|
+
|
19
|
+
typedef struct Table_connector_s Table_connector;
|
20
|
+
struct Table_connector_s
|
21
|
+
{
|
22
|
+
short lw, rw;
|
23
|
+
Connector *le, *re;
|
24
|
+
short cost;
|
25
|
+
s64 count;
|
26
|
+
Table_connector *next;
|
27
|
+
};
|
28
|
+
|
29
|
+
struct count_context_s
|
30
|
+
{
|
31
|
+
char ** deletable;
|
32
|
+
char ** effective_dist;
|
33
|
+
Word * local_sent;
|
34
|
+
int null_block;
|
35
|
+
int islands_ok;
|
36
|
+
int null_links;
|
37
|
+
int table_size;
|
38
|
+
int log2_table_size;
|
39
|
+
Table_connector ** table;
|
40
|
+
Resources current_resources;
|
41
|
+
};
|
42
|
+
|
43
|
+
static void free_table(count_context_t *ctxt)
|
44
|
+
{
|
45
|
+
int i;
|
46
|
+
Table_connector *t, *x;
|
47
|
+
|
48
|
+
for (i=0; i<ctxt->table_size; i++)
|
49
|
+
{
|
50
|
+
for(t = ctxt->table[i]; t!= NULL; t=x)
|
51
|
+
{
|
52
|
+
x = t->next;
|
53
|
+
xfree((void *) t, sizeof(Table_connector));
|
54
|
+
}
|
55
|
+
}
|
56
|
+
xfree(ctxt->table, ctxt->table_size * sizeof(Table_connector*));
|
57
|
+
ctxt->table = NULL;
|
58
|
+
ctxt->table_size = 0;
|
59
|
+
}
|
60
|
+
|
61
|
+
static void init_table(Sentence sent)
|
62
|
+
{
|
63
|
+
int shift;
|
64
|
+
/* A piecewise exponential function determines the size of the
|
65
|
+
* hash table. Probably should make use of the actual number of
|
66
|
+
* disjuncts, rather than just the number of words.
|
67
|
+
*/
|
68
|
+
count_context_t *ctxt = sent->count_ctxt;
|
69
|
+
|
70
|
+
if (ctxt->table) free_table(ctxt);
|
71
|
+
|
72
|
+
if (sent->length >= 10)
|
73
|
+
{
|
74
|
+
shift = 12 + (sent->length) / 6 ;
|
75
|
+
}
|
76
|
+
else
|
77
|
+
{
|
78
|
+
shift = 12;
|
79
|
+
}
|
80
|
+
|
81
|
+
/* Clamp at max 4*(1<<24) == 64 MBytes */
|
82
|
+
if (24 < shift) shift = 24;
|
83
|
+
ctxt->table_size = (1 << shift);
|
84
|
+
ctxt->log2_table_size = shift;
|
85
|
+
ctxt->table = (Table_connector**)
|
86
|
+
xalloc(ctxt->table_size * sizeof(Table_connector*));
|
87
|
+
memset(ctxt->table, 0, ctxt->table_size*sizeof(Table_connector*));
|
88
|
+
}
|
89
|
+
|
90
|
+
int x_match(Sentence sent, Connector *a, Connector *b)
|
91
|
+
{
|
92
|
+
return do_match(sent, a, b, 0, 0);
|
93
|
+
}
|
94
|
+
|
95
|
+
void count_set_effective_distance(Sentence sent)
|
96
|
+
{
|
97
|
+
sent->count_ctxt->effective_dist = sent->effective_dist;
|
98
|
+
}
|
99
|
+
|
100
|
+
void count_unset_effective_distance(Sentence sent)
|
101
|
+
{
|
102
|
+
sent->count_ctxt->effective_dist = NULL;
|
103
|
+
}
|
104
|
+
|
105
|
+
/*
|
106
|
+
* Returns TRUE if s and t match according to the connector matching
|
107
|
+
* rules. The connector strings must be properly formed, starting with
|
108
|
+
* zero or more upper case letters, followed by some other letters, and
|
109
|
+
* The algorithm is symmetric with respect to a and b.
|
110
|
+
*
|
111
|
+
* It works as follows: The labels must match. The priorities must be
|
112
|
+
* compatible (both THIN_priority, or one UP_priority and one DOWN_priority).
|
113
|
+
* The sequence of upper case letters must match exactly. After these comes
|
114
|
+
* a sequence of lower case letters "*"s or "^"s. The matching algorithm
|
115
|
+
* is different depending on which of the two priority cases is being
|
116
|
+
* considered. See the comments below.
|
117
|
+
*/
|
118
|
+
int do_match(Sentence sent, Connector *a, Connector *b, int aw, int bw)
|
119
|
+
{
|
120
|
+
const char *s, *t;
|
121
|
+
int x, y, dist;
|
122
|
+
count_context_t *ctxt;
|
123
|
+
|
124
|
+
if (a->label != b->label) return FALSE;
|
125
|
+
|
126
|
+
s = a->string;
|
127
|
+
t = b->string;
|
128
|
+
|
129
|
+
while(isupper((int)*s) || isupper((int)*t))
|
130
|
+
{
|
131
|
+
if (*s != *t) return FALSE;
|
132
|
+
s++;
|
133
|
+
t++;
|
134
|
+
}
|
135
|
+
|
136
|
+
ctxt = sent->count_ctxt;
|
137
|
+
|
138
|
+
x = a->priority;
|
139
|
+
y = b->priority;
|
140
|
+
|
141
|
+
/* Probably not necessary, as long as
|
142
|
+
* effective_dist[0][0]=0 and is defined */
|
143
|
+
if (aw == 0 && bw == 0) {
|
144
|
+
dist = 0;
|
145
|
+
} else {
|
146
|
+
assert(aw < bw, "match() did not receive params in the natural order.");
|
147
|
+
dist = ctxt->effective_dist[aw][bw];
|
148
|
+
}
|
149
|
+
/* printf("M: a=%4s b=%4s ap=%d bp=%d aw=%d bw=%d a->ll=%d b->ll=%d dist=%d\n",
|
150
|
+
s, t, x, y, aw, bw, a->length_limit, b->length_limit, dist); */
|
151
|
+
if (dist > a->length_limit || dist > b->length_limit) return FALSE;
|
152
|
+
|
153
|
+
if ((x == THIN_priority) && (y == THIN_priority))
|
154
|
+
{
|
155
|
+
/*
|
156
|
+
Remember that "*" matches anything, and "^" matches nothing
|
157
|
+
(except "*"). Otherwise two characters match if and only if
|
158
|
+
they're equal. ("^" can be used in the dictionary just like
|
159
|
+
any other connector.)
|
160
|
+
*/
|
161
|
+
while ((*s!='\0') && (*t!='\0')) {
|
162
|
+
if ((*s == '*') || (*t == '*') ||
|
163
|
+
((*s == *t) && (*s != '^'))) {
|
164
|
+
s++;
|
165
|
+
t++;
|
166
|
+
} else return FALSE;
|
167
|
+
}
|
168
|
+
return TRUE;
|
169
|
+
} else if ((x==UP_priority) && (y==DOWN_priority)) {
|
170
|
+
/*
|
171
|
+
As you go up (namely from x to y) the set of strings that
|
172
|
+
match (in the normal THIN sense above) should get no larger.
|
173
|
+
Read the comment in and.c to understand this.
|
174
|
+
In other words, the y string (t) must be weaker (or at least
|
175
|
+
no stronger) that the x string (s).
|
176
|
+
|
177
|
+
This code is only correct if the strings are the same
|
178
|
+
length. This is currently true, but perhaps for safty
|
179
|
+
this assumption should be removed.
|
180
|
+
*/
|
181
|
+
while ((*s!='\0') && (*t!='\0')) {
|
182
|
+
if ((*s == *t) || (*s == '*') || (*t == '^')) {
|
183
|
+
s++;
|
184
|
+
t++;
|
185
|
+
} else return FALSE;
|
186
|
+
}
|
187
|
+
return TRUE;
|
188
|
+
}
|
189
|
+
else if ((y == UP_priority) && (x == DOWN_priority))
|
190
|
+
{
|
191
|
+
while ((*s!='\0') && (*t!='\0'))
|
192
|
+
{
|
193
|
+
if ((*s == *t) || (*t == '*') || (*s == '^'))
|
194
|
+
{
|
195
|
+
s++;
|
196
|
+
t++;
|
197
|
+
}
|
198
|
+
else
|
199
|
+
return FALSE;
|
200
|
+
}
|
201
|
+
return TRUE;
|
202
|
+
}
|
203
|
+
else
|
204
|
+
return FALSE;
|
205
|
+
}
|
206
|
+
|
207
|
+
/**
|
208
|
+
* Stores the value in the table. Assumes it's not already there.
|
209
|
+
*/
|
210
|
+
static Table_connector * table_store(count_context_t *ctxt,
|
211
|
+
int lw, int rw,
|
212
|
+
Connector *le, Connector *re,
|
213
|
+
int cost, s64 count)
|
214
|
+
{
|
215
|
+
Table_connector *t, *n;
|
216
|
+
int h;
|
217
|
+
|
218
|
+
n = (Table_connector *) xalloc(sizeof(Table_connector));
|
219
|
+
n->count = count;
|
220
|
+
n->lw = lw; n->rw = rw; n->le = le; n->re = re; n->cost = cost;
|
221
|
+
h = pair_hash(ctxt->log2_table_size,lw, rw, le, re, cost);
|
222
|
+
t = ctxt->table[h];
|
223
|
+
n->next = t;
|
224
|
+
ctxt->table[h] = n;
|
225
|
+
return n;
|
226
|
+
}
|
227
|
+
|
228
|
+
/** returns the pointer to this info, NULL if not there */
|
229
|
+
static Table_connector *
|
230
|
+
find_table_pointer(count_context_t *ctxt,
|
231
|
+
int lw, int rw,
|
232
|
+
Connector *le, Connector *re,
|
233
|
+
int cost)
|
234
|
+
{
|
235
|
+
Table_connector *t;
|
236
|
+
int h = pair_hash(ctxt->log2_table_size,lw, rw, le, re, cost);
|
237
|
+
t = ctxt->table[h];
|
238
|
+
for (; t != NULL; t = t->next) {
|
239
|
+
if ((t->lw == lw) && (t->rw == rw) && (t->le == le) && (t->re == re)
|
240
|
+
&& (t->cost == cost)) return t;
|
241
|
+
}
|
242
|
+
|
243
|
+
/* Create a new connector only if resources are exhausted.
|
244
|
+
* (???) Huh? I guess we're in panic parse mode in that case.
|
245
|
+
*/
|
246
|
+
if ((ctxt->current_resources != NULL) &&
|
247
|
+
resources_exhausted(ctxt->current_resources))
|
248
|
+
{
|
249
|
+
return table_store(ctxt, lw, rw, le, re, cost, 0);
|
250
|
+
}
|
251
|
+
else return NULL;
|
252
|
+
}
|
253
|
+
|
254
|
+
/** returns the count for this quintuple if there, -1 otherwise */
|
255
|
+
s64 table_lookup(Sentence sent,
|
256
|
+
int lw, int rw, Connector *le, Connector *re, int cost)
|
257
|
+
{
|
258
|
+
Table_connector *t = find_table_pointer(sent->count_ctxt, lw, rw, le, re, cost);
|
259
|
+
|
260
|
+
if (t == NULL) return -1; else return t->count;
|
261
|
+
}
|
262
|
+
|
263
|
+
/**
|
264
|
+
* Stores the value in the table. Unlike table_store, it assumes
|
265
|
+
* it's already there
|
266
|
+
*/
|
267
|
+
static void table_update(count_context_t *ctxt, int lw, int rw,
|
268
|
+
Connector *le, Connector *re,
|
269
|
+
int cost, s64 count)
|
270
|
+
{
|
271
|
+
Table_connector *t = find_table_pointer(ctxt, lw, rw, le, re, cost);
|
272
|
+
|
273
|
+
assert(t != NULL, "This entry is supposed to be in the table.");
|
274
|
+
t->count = count;
|
275
|
+
}
|
276
|
+
|
277
|
+
/**
|
278
|
+
* Returns 0 if and only if this entry is in the hash table
|
279
|
+
* with a count value of 0.
|
280
|
+
*/
|
281
|
+
static s64 pseudocount(Sentence sent,
|
282
|
+
int lw, int rw, Connector *le, Connector *re, int cost)
|
283
|
+
{
|
284
|
+
s64 count;
|
285
|
+
count = table_lookup(sent, lw, rw, le, re, cost);
|
286
|
+
if (count == 0) return 0; else return 1;
|
287
|
+
}
|
288
|
+
|
289
|
+
static s64 do_count(Sentence sent, int lw, int rw,
|
290
|
+
Connector *le, Connector *re, int null_count)
|
291
|
+
{
|
292
|
+
Disjunct * d;
|
293
|
+
s64 total, pseudototal;
|
294
|
+
int start_word, end_word, w;
|
295
|
+
s64 leftcount, rightcount;
|
296
|
+
int lcost, rcost, Lmatch, Rmatch;
|
297
|
+
|
298
|
+
Match_node * m, *m1;
|
299
|
+
Table_connector *t;
|
300
|
+
|
301
|
+
count_context_t *ctxt = sent->count_ctxt;
|
302
|
+
|
303
|
+
if (null_count < 0) return 0; /* can this ever happen?? */
|
304
|
+
|
305
|
+
t = find_table_pointer(ctxt, lw, rw, le, re, null_count);
|
306
|
+
|
307
|
+
if (t == NULL) {
|
308
|
+
/* Create the table entry with a tentative null count of 0.
|
309
|
+
* This count must be updated before we return. */
|
310
|
+
t = table_store(ctxt, lw, rw, le, re, null_count, 0);
|
311
|
+
} else {
|
312
|
+
return t->count;
|
313
|
+
}
|
314
|
+
|
315
|
+
if (rw == 1+lw)
|
316
|
+
{
|
317
|
+
/* lw and rw are neighboring words */
|
318
|
+
/* You can't have a linkage here with null_count > 0 */
|
319
|
+
if ((le == NULL) && (re == NULL) && (null_count == 0))
|
320
|
+
{
|
321
|
+
t->count = 1;
|
322
|
+
}
|
323
|
+
else
|
324
|
+
{
|
325
|
+
t->count = 0;
|
326
|
+
}
|
327
|
+
return t->count;
|
328
|
+
}
|
329
|
+
|
330
|
+
if ((le == NULL) && (re == NULL))
|
331
|
+
{
|
332
|
+
if (!ctxt->islands_ok && (lw != -1))
|
333
|
+
{
|
334
|
+
/* If we don't allow islands (a set of words linked together
|
335
|
+
* but separate from the rest of the sentence) then the
|
336
|
+
* null_count of skipping n words is just n */
|
337
|
+
if (null_count == ((rw-lw-1) + ctxt->null_block-1)/ctxt->null_block)
|
338
|
+
{
|
339
|
+
/* If null_block=4 then the null_count of
|
340
|
+
1,2,3,4 nulls is 1; and 5,6,7,8 is 2 etc. */
|
341
|
+
t->count = 1;
|
342
|
+
}
|
343
|
+
else
|
344
|
+
{
|
345
|
+
t->count = 0;
|
346
|
+
}
|
347
|
+
return t->count;
|
348
|
+
}
|
349
|
+
if (null_count == 0)
|
350
|
+
{
|
351
|
+
/* There is no solution without nulls in this case. There is
|
352
|
+
* a slight efficiency hack to separate this null_count==0
|
353
|
+
* case out, but not necessary for correctness */
|
354
|
+
t->count = 0;
|
355
|
+
}
|
356
|
+
else
|
357
|
+
{
|
358
|
+
total = 0;
|
359
|
+
w = lw+1;
|
360
|
+
for (d = ctxt->local_sent[w].d; d != NULL; d = d->next)
|
361
|
+
{
|
362
|
+
if (d->left == NULL)
|
363
|
+
{
|
364
|
+
total += do_count(sent, w, rw, d->right, NULL, null_count-1);
|
365
|
+
}
|
366
|
+
}
|
367
|
+
total += do_count(sent, w, rw, NULL, NULL, null_count-1);
|
368
|
+
t->count = total;
|
369
|
+
}
|
370
|
+
return t->count;
|
371
|
+
}
|
372
|
+
|
373
|
+
if (le == NULL)
|
374
|
+
{
|
375
|
+
start_word = lw+1;
|
376
|
+
}
|
377
|
+
else
|
378
|
+
{
|
379
|
+
start_word = le->word;
|
380
|
+
}
|
381
|
+
|
382
|
+
if (re == NULL)
|
383
|
+
{
|
384
|
+
end_word = rw;
|
385
|
+
}
|
386
|
+
else
|
387
|
+
{
|
388
|
+
end_word = re->word +1;
|
389
|
+
}
|
390
|
+
|
391
|
+
total = 0;
|
392
|
+
|
393
|
+
for (w = start_word; w < end_word; w++)
|
394
|
+
{
|
395
|
+
m1 = m = form_match_list(sent, w, le, lw, re, rw);
|
396
|
+
for (; m!=NULL; m=m->next)
|
397
|
+
{
|
398
|
+
d = m->d;
|
399
|
+
for (lcost = 0; lcost <= null_count; lcost++)
|
400
|
+
{
|
401
|
+
rcost = null_count - lcost;
|
402
|
+
/* Now lcost and rcost are the costs we're assigning
|
403
|
+
* to those parts respectively */
|
404
|
+
|
405
|
+
/* Now, we determine if (based on table only) we can see that
|
406
|
+
the current range is not parsable. */
|
407
|
+
Lmatch = (le != NULL) && (d->left != NULL) &&
|
408
|
+
do_match(sent, le, d->left, lw, w);
|
409
|
+
Rmatch = (d->right != NULL) && (re != NULL) &&
|
410
|
+
do_match(sent, d->right, re, w, rw);
|
411
|
+
|
412
|
+
rightcount = leftcount = 0;
|
413
|
+
if (Lmatch)
|
414
|
+
{
|
415
|
+
leftcount = pseudocount(sent, lw, w, le->next, d->left->next, lcost);
|
416
|
+
if (le->multi) leftcount += pseudocount(sent, lw, w, le, d->left->next, lcost);
|
417
|
+
if (d->left->multi) leftcount += pseudocount(sent, lw, w, le->next, d->left, lcost);
|
418
|
+
if (le->multi && d->left->multi) leftcount += pseudocount(sent, lw, w, le, d->left, lcost);
|
419
|
+
}
|
420
|
+
|
421
|
+
if (Rmatch)
|
422
|
+
{
|
423
|
+
rightcount = pseudocount(sent, w, rw, d->right->next, re->next, rcost);
|
424
|
+
if (d->right->multi) rightcount += pseudocount(sent, w,rw,d->right,re->next, rcost);
|
425
|
+
if (re->multi) rightcount += pseudocount(sent, w, rw, d->right->next, re, rcost);
|
426
|
+
if (d->right->multi && re->multi) rightcount += pseudocount(sent, w, rw, d->right, re, rcost);
|
427
|
+
}
|
428
|
+
|
429
|
+
/* total number where links are used on both sides */
|
430
|
+
pseudototal = leftcount*rightcount;
|
431
|
+
|
432
|
+
if (leftcount > 0) {
|
433
|
+
/* evaluate using the left match, but not the right */
|
434
|
+
pseudototal += leftcount * pseudocount(sent, w, rw, d->right, re, rcost);
|
435
|
+
}
|
436
|
+
if ((le == NULL) && (rightcount > 0)) {
|
437
|
+
/* evaluate using the right match, but not the left */
|
438
|
+
pseudototal += rightcount * pseudocount(sent, lw, w, le, d->left, lcost);
|
439
|
+
}
|
440
|
+
|
441
|
+
/* now pseudototal is 0 implies that we know that the true total is 0 */
|
442
|
+
if (pseudototal != 0) {
|
443
|
+
rightcount = leftcount = 0;
|
444
|
+
if (Lmatch) {
|
445
|
+
leftcount = do_count(sent, lw, w, le->next, d->left->next, lcost);
|
446
|
+
if (le->multi) leftcount += do_count(sent, lw, w, le, d->left->next, lcost);
|
447
|
+
if (d->left->multi) leftcount += do_count(sent, lw, w, le->next, d->left, lcost);
|
448
|
+
if (le->multi && d->left->multi) leftcount += do_count(sent, lw, w, le, d->left, lcost);
|
449
|
+
}
|
450
|
+
|
451
|
+
if (Rmatch) {
|
452
|
+
rightcount = do_count(sent, w, rw, d->right->next, re->next, rcost);
|
453
|
+
if (d->right->multi) rightcount += do_count(sent, w,rw,d->right,re->next, rcost);
|
454
|
+
if (re->multi) rightcount += do_count(sent, w, rw, d->right->next, re, rcost);
|
455
|
+
if (d->right->multi && re->multi) rightcount += do_count(sent, w, rw, d->right, re, rcost);
|
456
|
+
}
|
457
|
+
|
458
|
+
total += leftcount*rightcount; /* total number where links are used on both sides */
|
459
|
+
|
460
|
+
if (leftcount > 0) {
|
461
|
+
/* evaluate using the left match, but not the right */
|
462
|
+
total += leftcount * do_count(sent, w, rw, d->right, re, rcost);
|
463
|
+
}
|
464
|
+
if ((le == NULL) && (rightcount > 0)) {
|
465
|
+
/* evaluate using the right match, but not the left */
|
466
|
+
total += rightcount * do_count(sent, lw, w, le, d->left, lcost);
|
467
|
+
}
|
468
|
+
}
|
469
|
+
}
|
470
|
+
}
|
471
|
+
|
472
|
+
put_match_list(sent, m1);
|
473
|
+
}
|
474
|
+
t->count = total;
|
475
|
+
return total;
|
476
|
+
}
|
477
|
+
|
478
|
+
/**
|
479
|
+
* Returns the number of ways the sentence can be parsed with the
|
480
|
+
* specified null count. Assumes that the hash table has already been
|
481
|
+
* initialized, and is freed later. The "null_count" here is the
|
482
|
+
* number of words that are allowed to have no links to them.
|
483
|
+
*/
|
484
|
+
s64 do_parse(Sentence sent, int null_count, Parse_Options opts)
|
485
|
+
{
|
486
|
+
s64 total;
|
487
|
+
count_context_t *ctxt = sent->count_ctxt;
|
488
|
+
|
489
|
+
count_set_effective_distance(sent);
|
490
|
+
ctxt->current_resources = opts->resources;
|
491
|
+
ctxt->local_sent = sent->word;
|
492
|
+
ctxt->deletable = sent->deletable;
|
493
|
+
ctxt->null_block = opts->null_block;
|
494
|
+
ctxt->islands_ok = opts->islands_ok;
|
495
|
+
|
496
|
+
total = do_count(sent, -1, sent->length, NULL, NULL, null_count+1);
|
497
|
+
|
498
|
+
ctxt->local_sent = NULL;
|
499
|
+
ctxt->current_resources = NULL;
|
500
|
+
return total;
|
501
|
+
}
|
502
|
+
|
503
|
+
/**
|
504
|
+
CONJUNCTION PRUNING.
|
505
|
+
|
506
|
+
The basic idea is this. Before creating the fat disjuncts,
|
507
|
+
we run a modified version of the exhaustive search procedure.
|
508
|
+
Its purpose is to mark the disjuncts that can be used in any
|
509
|
+
linkage. It's just like the normal exhaustive search, except that
|
510
|
+
if a subrange of words are deletable, then we treat them as though
|
511
|
+
they were not even there. So, if we call the function in the
|
512
|
+
situation where the set of words between the left and right one
|
513
|
+
are deletable, and the left and right connector pointers
|
514
|
+
are NULL, then that range is considered to have a solution.
|
515
|
+
|
516
|
+
There are actually two procedures to implement this. One is
|
517
|
+
mark_region() and the other is region_valid(). The latter just
|
518
|
+
checks to see if the given region can be completed (within it).
|
519
|
+
The former actually marks those disjuncts that can be used in
|
520
|
+
any valid linkage of the given region.
|
521
|
+
|
522
|
+
As in the standard search procedure, we make use of the fast-match
|
523
|
+
data structure (which requires power pruning to have been done), and
|
524
|
+
we also use a hash table. The table is used differently in this case.
|
525
|
+
The meaning of values stored in the table are as follows:
|
526
|
+
|
527
|
+
-1 Nothing known (Actually, this is not stored. It's returned
|
528
|
+
by table_lookup when nothing is known.)
|
529
|
+
0 This region can't be completed (marking is therefore irrelevant)
|
530
|
+
1 This region can be completed, but it's not yet marked
|
531
|
+
2 This region can be completed, and it's been marked.
|
532
|
+
*/
|
533
|
+
|
534
|
+
static int x_prune_match(count_context_t *ctxt,
|
535
|
+
Connector *le, Connector *re, int lw, int rw)
|
536
|
+
{
|
537
|
+
int dist;
|
538
|
+
|
539
|
+
assert(lw < rw, "prune_match() did not receive params in the natural order.");
|
540
|
+
dist = ctxt->effective_dist[lw][rw];
|
541
|
+
return prune_match(dist, le, re);
|
542
|
+
}
|
543
|
+
|
544
|
+
/**
|
545
|
+
* Returns 0 if this range cannot be successfully filled in with
|
546
|
+
* links. Returns 1 if it can, and it's not been marked, and returns
|
547
|
+
* 2 if it can and it has been marked.
|
548
|
+
*/
|
549
|
+
static int region_valid(Sentence sent, int lw, int rw, Connector *le, Connector *re)
|
550
|
+
{
|
551
|
+
Disjunct * d;
|
552
|
+
int left_valid, right_valid, found;
|
553
|
+
int i, start_word, end_word;
|
554
|
+
int w;
|
555
|
+
Match_node * m, *m1;
|
556
|
+
|
557
|
+
count_context_t *ctxt = sent->count_ctxt;
|
558
|
+
|
559
|
+
i = table_lookup(sent, lw, rw, le, re, 0);
|
560
|
+
if (i >= 0) return i;
|
561
|
+
|
562
|
+
if ((le == NULL) && (re == NULL) && ctxt->deletable[lw][rw]) {
|
563
|
+
table_store(ctxt, lw, rw, le, re, 0, 1);
|
564
|
+
return 1;
|
565
|
+
}
|
566
|
+
|
567
|
+
if (le == NULL) {
|
568
|
+
start_word = lw+1;
|
569
|
+
} else {
|
570
|
+
start_word = le->word;
|
571
|
+
}
|
572
|
+
if (re == NULL) {
|
573
|
+
end_word = rw;
|
574
|
+
} else {
|
575
|
+
end_word = re->word + 1;
|
576
|
+
}
|
577
|
+
|
578
|
+
found = 0;
|
579
|
+
|
580
|
+
for (w=start_word; w < end_word; w++)
|
581
|
+
{
|
582
|
+
m1 = m = form_match_list(sent, w, le, lw, re, rw);
|
583
|
+
for (; m!=NULL; m=m->next)
|
584
|
+
{
|
585
|
+
d = m->d;
|
586
|
+
/* mark_cost++;*/
|
587
|
+
/* in the following expressions we use the fact that 0=FALSE. Could eliminate
|
588
|
+
by always saying "region_valid(...) != 0" */
|
589
|
+
left_valid = (((le != NULL) && (d->left != NULL) && x_prune_match(ctxt, le, d->left, lw, w)) &&
|
590
|
+
((region_valid(sent, lw, w, le->next, d->left->next)) ||
|
591
|
+
((le->multi) && region_valid(sent, lw, w, le, d->left->next)) ||
|
592
|
+
((d->left->multi) && region_valid(sent, lw, w, le->next, d->left)) ||
|
593
|
+
((le->multi && d->left->multi) && region_valid(sent, lw, w, le, d->left))));
|
594
|
+
if (left_valid && region_valid(sent, w, rw, d->right, re)) {
|
595
|
+
found = 1;
|
596
|
+
break;
|
597
|
+
}
|
598
|
+
right_valid = (((d->right != NULL) && (re != NULL) && x_prune_match(ctxt, d->right, re, w, rw)) &&
|
599
|
+
((region_valid(sent, w, rw, d->right->next,re->next)) ||
|
600
|
+
((d->right->multi) && region_valid(sent, w,rw,d->right,re->next)) ||
|
601
|
+
((re->multi) && region_valid(sent, w, rw, d->right->next, re)) ||
|
602
|
+
((d->right->multi && re->multi) && region_valid(sent, w, rw, d->right, re))));
|
603
|
+
if ((left_valid && right_valid) || (right_valid && region_valid(sent, lw, w, le, d->left))) {
|
604
|
+
found = 1;
|
605
|
+
break;
|
606
|
+
}
|
607
|
+
}
|
608
|
+
put_match_list(sent, m1);
|
609
|
+
if (found != 0) break;
|
610
|
+
}
|
611
|
+
table_store(ctxt, lw, rw, le, re, 0, found);
|
612
|
+
return found;
|
613
|
+
}
|
614
|
+
|
615
|
+
/**
|
616
|
+
* Mark as useful all disjuncts involved in some way to complete the
|
617
|
+
* structure within the current region. Note that only disjuncts
|
618
|
+
* strictly between lw and rw will be marked. If it so happens that
|
619
|
+
* this region itself is not valid, then this fact will be recorded
|
620
|
+
* in the table, and nothing else happens.
|
621
|
+
*/
|
622
|
+
static void mark_region(Sentence sent,
|
623
|
+
int lw, int rw, Connector *le, Connector *re)
|
624
|
+
{
|
625
|
+
|
626
|
+
Disjunct * d;
|
627
|
+
int left_valid, right_valid, i;
|
628
|
+
int start_word, end_word;
|
629
|
+
int w;
|
630
|
+
Match_node * m, *m1;
|
631
|
+
count_context_t *ctxt = sent->count_ctxt;
|
632
|
+
|
633
|
+
i = region_valid(sent, lw, rw, le, re);
|
634
|
+
if ((i==0) || (i==2)) return;
|
635
|
+
/* we only reach this point if it's a valid unmarked region, i=1 */
|
636
|
+
table_update(ctxt, lw, rw, le, re, 0, 2);
|
637
|
+
|
638
|
+
if ((le == NULL) && (re == NULL) && (ctxt->null_links) && (rw != 1+lw)) {
|
639
|
+
w = lw+1;
|
640
|
+
for (d = ctxt->local_sent[w].d; d != NULL; d = d->next) {
|
641
|
+
if ((d->left == NULL) && region_valid(sent, w, rw, d->right, NULL)) {
|
642
|
+
d->marked = TRUE;
|
643
|
+
mark_region(sent, w, rw, d->right, NULL);
|
644
|
+
}
|
645
|
+
}
|
646
|
+
mark_region(sent, w, rw, NULL, NULL);
|
647
|
+
return;
|
648
|
+
}
|
649
|
+
|
650
|
+
if (le == NULL) {
|
651
|
+
start_word = lw+1;
|
652
|
+
} else {
|
653
|
+
start_word = le->word;
|
654
|
+
}
|
655
|
+
if (re == NULL) {
|
656
|
+
end_word = rw;
|
657
|
+
} else {
|
658
|
+
end_word = re->word + 1;
|
659
|
+
}
|
660
|
+
|
661
|
+
for (w=start_word; w < end_word; w++)
|
662
|
+
{
|
663
|
+
m1 = m = form_match_list(sent, w, le, lw, re, rw);
|
664
|
+
for (; m!=NULL; m=m->next)
|
665
|
+
{
|
666
|
+
d = m->d;
|
667
|
+
/* mark_cost++;*/
|
668
|
+
left_valid = (((le != NULL) && (d->left != NULL) && x_prune_match(ctxt, le, d->left, lw, w)) &&
|
669
|
+
((region_valid(sent, lw, w, le->next, d->left->next)) ||
|
670
|
+
((le->multi) && region_valid(sent, lw, w, le, d->left->next)) ||
|
671
|
+
((d->left->multi) && region_valid(sent, lw, w, le->next, d->left)) ||
|
672
|
+
((le->multi && d->left->multi) && region_valid(sent, lw, w, le, d->left))));
|
673
|
+
right_valid = (((d->right != NULL) && (re != NULL) && x_prune_match(ctxt, d->right, re, w, rw)) &&
|
674
|
+
((region_valid(sent, w, rw, d->right->next,re->next)) ||
|
675
|
+
((d->right->multi) && region_valid(sent, w,rw,d->right,re->next)) ||
|
676
|
+
((re->multi) && region_valid(sent, w, rw, d->right->next, re)) ||
|
677
|
+
((d->right->multi && re->multi) && region_valid(sent, w, rw, d->right, re))));
|
678
|
+
|
679
|
+
/* The following if statements could be restructured to avoid superfluous calls
|
680
|
+
to mark_region. It didn't seem a high priority, so I didn't optimize this.
|
681
|
+
*/
|
682
|
+
|
683
|
+
if (left_valid && region_valid(sent, w, rw, d->right, re))
|
684
|
+
{
|
685
|
+
d->marked = TRUE;
|
686
|
+
mark_region(sent, w, rw, d->right, re);
|
687
|
+
mark_region(sent, lw, w, le->next, d->left->next);
|
688
|
+
if (le->multi) mark_region(sent, lw, w, le, d->left->next);
|
689
|
+
if (d->left->multi) mark_region(sent, lw, w, le->next, d->left);
|
690
|
+
if (le->multi && d->left->multi) mark_region(sent, lw, w, le, d->left);
|
691
|
+
}
|
692
|
+
|
693
|
+
if (right_valid && region_valid(sent, lw, w, le, d->left))
|
694
|
+
{
|
695
|
+
d->marked = TRUE;
|
696
|
+
mark_region(sent, lw, w, le, d->left);
|
697
|
+
mark_region(sent, w, rw, d->right->next,re->next);
|
698
|
+
if (d->right->multi) mark_region(sent, w,rw,d->right,re->next);
|
699
|
+
if (re->multi) mark_region(sent, w, rw, d->right->next, re);
|
700
|
+
if (d->right->multi && re->multi) mark_region(sent, w, rw, d->right, re);
|
701
|
+
}
|
702
|
+
|
703
|
+
if (left_valid && right_valid)
|
704
|
+
{
|
705
|
+
d->marked = TRUE;
|
706
|
+
mark_region(sent, lw, w, le->next, d->left->next);
|
707
|
+
if (le->multi) mark_region(sent, lw, w, le, d->left->next);
|
708
|
+
if (d->left->multi) mark_region(sent, lw, w, le->next, d->left);
|
709
|
+
if (le->multi && d->left->multi) mark_region(sent, lw, w, le, d->left);
|
710
|
+
mark_region(sent, w, rw, d->right->next,re->next);
|
711
|
+
if (d->right->multi) mark_region(sent, w,rw,d->right,re->next);
|
712
|
+
if (re->multi) mark_region(sent, w, rw, d->right->next, re);
|
713
|
+
if (d->right->multi && re->multi) mark_region(sent, w, rw, d->right, re);
|
714
|
+
}
|
715
|
+
}
|
716
|
+
put_match_list(sent, m1);
|
717
|
+
}
|
718
|
+
}
|
719
|
+
|
720
|
+
void delete_unmarked_disjuncts(Sentence sent)
|
721
|
+
{
|
722
|
+
int w;
|
723
|
+
Disjunct *d_head, *d, *dx;
|
724
|
+
|
725
|
+
for (w=0; w<sent->length; w++) {
|
726
|
+
d_head = NULL;
|
727
|
+
for (d=sent->word[w].d; d != NULL; d=dx) {
|
728
|
+
dx = d->next;
|
729
|
+
if (d->marked) {
|
730
|
+
d->next = d_head;
|
731
|
+
d_head = d;
|
732
|
+
} else {
|
733
|
+
d->next = NULL;
|
734
|
+
free_disjuncts(d);
|
735
|
+
}
|
736
|
+
}
|
737
|
+
sent->word[w].d = d_head;
|
738
|
+
}
|
739
|
+
}
|
740
|
+
|
741
|
+
/**
|
742
|
+
* We've already built the sentence disjuncts, and we've pruned them
|
743
|
+
* and power_pruned(GENTLE) them also. The sentence contains a
|
744
|
+
* conjunction. deletable[][] has been initialized to indicate the
|
745
|
+
* ranges which may be deleted in the final linkage.
|
746
|
+
*
|
747
|
+
* This routine deletes irrelevant disjuncts. It finds them by first
|
748
|
+
* marking them all as irrelevant, and then marking the ones that
|
749
|
+
* might be useable. Finally, the unmarked ones are removed.
|
750
|
+
*/
|
751
|
+
void conjunction_prune(Sentence sent, Parse_Options opts)
|
752
|
+
{
|
753
|
+
Disjunct * d;
|
754
|
+
int w;
|
755
|
+
count_context_t *ctxt = sent->count_ctxt;
|
756
|
+
|
757
|
+
ctxt->current_resources = opts->resources;
|
758
|
+
ctxt->deletable = sent->deletable;
|
759
|
+
count_set_effective_distance(sent);
|
760
|
+
|
761
|
+
/* We begin by unmarking all disjuncts. This would not be necessary if
|
762
|
+
whenever we created a disjunct we cleared its marked field.
|
763
|
+
I didn't want to search the program for all such places, so
|
764
|
+
I did this way. XXX FIXME, someday ...
|
765
|
+
*/
|
766
|
+
for (w=0; w<sent->length; w++) {
|
767
|
+
for (d=sent->word[w].d; d != NULL; d=d->next) {
|
768
|
+
d->marked = FALSE;
|
769
|
+
}
|
770
|
+
}
|
771
|
+
|
772
|
+
init_fast_matcher(sent);
|
773
|
+
ctxt->local_sent = sent->word;
|
774
|
+
ctxt->null_links = (opts->min_null_count > 0);
|
775
|
+
/*
|
776
|
+
for (d = sent->word[0].d; d != NULL; d = d->next) {
|
777
|
+
if ((d->left == NULL) && region_valid(sent, 0, sent->length, d->right, NULL)) {
|
778
|
+
mark_region(sent, 0, sent->length, d->right, NULL);
|
779
|
+
d->marked = TRUE;
|
780
|
+
}
|
781
|
+
}
|
782
|
+
mark_region(sent, 0, sent->length, NULL, NULL);
|
783
|
+
*/
|
784
|
+
|
785
|
+
if (ctxt->null_links) {
|
786
|
+
mark_region(sent, -1, sent->length, NULL, NULL);
|
787
|
+
} else {
|
788
|
+
for (w=0; w<sent->length; w++) {
|
789
|
+
/* consider removing the words [0,w-1] from the beginning
|
790
|
+
of the sentence */
|
791
|
+
if (ctxt->deletable[-1][w]) {
|
792
|
+
for (d = sent->word[w].d; d != NULL; d = d->next) {
|
793
|
+
if ((d->left == NULL) && region_valid(sent, w, sent->length, d->right, NULL)) {
|
794
|
+
mark_region(sent, w, sent->length, d->right, NULL);
|
795
|
+
d->marked = TRUE;
|
796
|
+
}
|
797
|
+
}
|
798
|
+
}
|
799
|
+
}
|
800
|
+
}
|
801
|
+
|
802
|
+
delete_unmarked_disjuncts(sent);
|
803
|
+
|
804
|
+
free_fast_matcher(sent);
|
805
|
+
|
806
|
+
ctxt->local_sent = NULL;
|
807
|
+
ctxt->current_resources = NULL;
|
808
|
+
ctxt->deletable = NULL;
|
809
|
+
count_unset_effective_distance(sent);
|
810
|
+
}
|
811
|
+
|
812
|
+
void init_count(Sentence sent)
|
813
|
+
{
|
814
|
+
if (NULL == sent->count_ctxt)
|
815
|
+
sent->count_ctxt = (count_context_t *) malloc (sizeof(count_context_t));
|
816
|
+
memset(sent->count_ctxt, 0, sizeof(count_context_t));
|
817
|
+
|
818
|
+
init_table(sent);
|
819
|
+
}
|
820
|
+
|
821
|
+
void free_count(Sentence sent)
|
822
|
+
{
|
823
|
+
if (NULL == sent->count_ctxt) return;
|
824
|
+
|
825
|
+
free_table(sent->count_ctxt);
|
826
|
+
free(sent->count_ctxt);
|
827
|
+
sent->count_ctxt = NULL;
|
828
|
+
}
|