diff options
-rw-r--r-- | configure.ac | 5 | ||||
-rw-r--r-- | decoder/dict.h | 10 | ||||
-rwxr-xr-x | gi/pipeline/local-gi-pipeline.pl | 3 | ||||
-rwxr-xr-x | gi/pyp-topics/scripts/spans2labels.py | 47 |
4 files changed, 37 insertions, 28 deletions
diff --git a/configure.ac b/configure.ac index a7f4cfe7..e6119f29 100644 --- a/configure.ac +++ b/configure.ac @@ -8,9 +8,10 @@ AC_PROG_CXX AC_LANG_CPLUSPLUS BOOST_REQUIRE BOOST_PROGRAM_OPTIONS +BOOST_REGEX CPPFLAGS="$CPPFLAGS $BOOST_CPPFLAGS" -LDFLAGS="$LDFLAGS $BOOST_PROGRAM_OPTIONS_LDFLAGS" -LIBS="$LIBS $BOOST_PROGRAM_OPTIONS_LIBS" +LDFLAGS="$LDFLAGS $BOOST_PROGRAM_OPTIONS_LDFLAGS $BOOST_REGEX_LDFLAGS" +LIBS="$LIBS $BOOST_PROGRAM_OPTIONS_LIBS $BOOST_REGEX_LIBS" AC_CHECK_HEADER(boost/math/special_functions/digamma.hpp, [AC_DEFINE([HAVE_BOOST_DIGAMMA], [], [flag for boost::math::digamma])]) diff --git a/decoder/dict.h b/decoder/dict.h index bc3a904a..39baf6ed 100644 --- a/decoder/dict.h +++ b/decoder/dict.h @@ -8,6 +8,8 @@ #include <vector> #include <boost/functional/hash.hpp> +#include <boost/regex.hpp> +#include <boost/algorithm/string/regex.hpp> #include "wordid.h" @@ -36,7 +38,7 @@ class Dict { std::string word= ""; for (std::vector<std::string>::const_iterator it=words.begin(); it != words.end(); ++it) { - if (it != words.begin()) word += "__"; + if (it != words.begin()) word += "|||"; word += *it; } @@ -49,6 +51,12 @@ class Dict { return words_[id-1]; } + inline std::vector<std::string> AsVector(const WordID& id) const { + std::vector<std::string> result; + boost::algorithm::split_regex(result, Convert(id), boost::regex("\\|\\|\\|")); + return result; + } + void clear() { words_.clear(); d_.clear(); } private: diff --git a/gi/pipeline/local-gi-pipeline.pl b/gi/pipeline/local-gi-pipeline.pl index af83beb8..be91f9ad 100755 --- a/gi/pipeline/local-gi-pipeline.pl +++ b/gi/pipeline/local-gi-pipeline.pl @@ -138,7 +138,8 @@ sub label_spans_with_topics { print STDERR "$OUT_SPANS exists, reusing...\n"; } else { safesystem("$ZCAT $IN_CLUSTERS > $OUTPUT/clusters.txt") or die "Failed to unzip"; - safesystem("$EXTRACTOR --base_phrase_spans -i $CORPUS -c $ITEMS_IN_MEMORY -L $BASE_PHRASE_MAX_SIZE -S $CONTEXT_SIZE | $S2L $OUTPUT/phrases.index $OUTPUT/contexts.index $OUTPUT/clusters.txt > $OUT_SPANS") or die "Failed to label spans"; +# safesystem("$EXTRACTOR --base_phrase_spans -i $CORPUS -c $ITEMS_IN_MEMORY -L $BASE_PHRASE_MAX_SIZE -S $CONTEXT_SIZE | $S2L $OUTPUT/phrases.index $OUTPUT/contexts.index $OUTPUT/clusters.txt > $OUT_SPANS") or die "Failed to label spans"; + safesystem("$EXTRACTOR --base_phrase_spans -i $CORPUS -c $ITEMS_IN_MEMORY -L $BASE_PHRASE_MAX_SIZE -S $CONTEXT_SIZE | $S2L $OUTPUT/clusters.txt $CONTEXT_SIZE > $OUT_SPANS") or die "Failed to label spans"; unlink("$OUTPUT/clusters.txt") or warn "Failed to remove $OUTPUT/clusters.txt"; safesystem("paste -d ' ' $CORPUS $OUT_SPANS > $OUTPUT/corpus.src_trg_al") or die "Couldn't paste"; } diff --git a/gi/pyp-topics/scripts/spans2labels.py b/gi/pyp-topics/scripts/spans2labels.py index 409fda92..3221dbf0 100755 --- a/gi/pyp-topics/scripts/spans2labels.py +++ b/gi/pyp-topics/scripts/spans2labels.py @@ -3,44 +3,43 @@ import sys from operator import itemgetter -if len(sys.argv) != 4: - print "Usage: spans2labels.py phrase_index context_index phrase_context_index" +if len(sys.argv) <= 2: + print "Usage: spans2labels.py phrase_context_index [order]" exit(1) -phrase_index = dict(map(lambda x: (x[1].strip(),x[0]), enumerate(file(sys.argv[1], 'r').readlines()))) -context_index = dict(map(lambda x: (x[1].strip(),x[0]), enumerate(file(sys.argv[2], 'r').readlines()))) +order=1 +if len(sys.argv) > 2: + order = int(sys.argv[2]) phrase_context_index = {} -for i,line in enumerate(file(sys.argv[3], 'r').readlines()): - for c,l in map(lambda x: x.split(':'), line.split()[1:]): - phrase_context_index[(int(i),int(c))] = l +for line in file(sys.argv[1], 'r'): + phrase,tail= line.split('\t') + contexts = tail.split(" ||| ") + assert len(contexts) % 2 == 0 + for i in range(0, len(contexts), 2): + category = contexts[i+1].split("=")[1].strip() + phrase_context_index[(phrase,contexts[i])] = category +# print (phrase,contexts[i]), category for line in sys.stdin: line_segments = line.split('|||') - source = ['<s>'] + line_segments[0].split() + ['</s>'] - target = ['<s>'] + line_segments[1].split() + ['</s>'] + source = ['<s>' for x in range(order)] + line_segments[0].split() + ['</s>' for x in range(order)] + target = ['<s>' for x in range(order)] + line_segments[1].split() + ['</s>' for x in range(order)] phrases = [ [int(i) for i in x.split('-')] for x in line_segments[2].split()] -# for x in source[1:-1]: -# print x, -# print "|||", -# for x in target[1:-1]: -# print x, print "|||", for s1,s2,t1,t2 in phrases: - s1 += 1 - s2 += 1 - t1 += 1 - t2 += 1 + s1 += order + s2 += order + t1 += order + t2 += order phrase = reduce(lambda x, y: x+y+" ", target[t1:t2], "").strip() - context = "%s <PHRASE> %s" % (target[t1-1], target[t2]) + left_context = reduce(lambda x, y: x+y+" ", target[t1-order:t1], "") + right_context = reduce(lambda x, y: x+y+" ", target[t2:t2+order], "").strip() + context = "%s<PHRASE> %s" % (left_context, right_context) - pi = phrase_index[phrase] - ci = context_index[context] - label = phrase_context_index[(pi,ci)] + label = phrase_context_index[(phrase,context)] print "%s-%s:X%s" % (t1-1,t2-1,label), -# print phrase, pi, context, ci -# print phrase_context_index[(pi,ci)] print |