def verb_mining():
text = read_cabocha('neko.txt.cabocha')
fout = codecs.open('mining.txt', 'w', 'utf-8')
for sentence in text:
# if を is in chunk, wo_pos represents the index of the char
for wo_pos, chunk in enumerate(sentence):
for morph in chunk.morphs:
if morph.pos1 == u'サ変接続' and chunk.case() == u'を':
verb = sentence[chunk.dst].first_verb()
if verb:
# sentence[chunk.dst] must has verb
line = chunk.surface() + verb + u'\t'
cases = []
surfaces = []
for src in (chunk.srcs + sentence[chunk.dst].srcs):
if src == wo_pos:
continue
case = sentence[src].case()
if case:
cases.append(case)
surfaces.append(sentence[src].surface())
if cases:
line = line + u' '.join(
sorted(cases)) + u'\t' + u' '.join(
sorted(surfaces,
key=lambda surface: surface[-1]))
fout.write(line + u'\n')
fout.close()
def relations():
fout = codecs.open('out42.txt', 'w', 'utf-8')
text = read_cabocha('neko.txt.cabocha')
for sentence in text:
for chunk in sentence:
line = chunk.surface()
for src in chunk.srcs:
if sentence[src].morphs:
line = line + u'\t' + sentence[src].surface()
line = line + '\n'
fout.write(line)
fout.close()
def noun_to_noun_path():
text = read_cabocha('neko.txt.cabocha')
fout = codecs.open('n2n.txt', 'w', 'utf-8')
for sentence in text:
for i, chunk1 in enumerate(list(sentence)):
for j, chunk2 in enumerate(list(sentence[i + 1:]), i + 1):
if chunk1.has_noun() and chunk2.has_noun():
start_chunk = Chunk([], chunk1.dst, chunk1.srcs)
end_chunk = Chunk([], chunk2.dst, chunk2.srcs)
for morph in chunk1.morphs:
if morph.pos == u'名詞':
if not u'X' in map(lambda m: m.surface,
start_chunk.morphs):
start_chunk.morphs.append(
Morph(u'X', u'X', u'名詞', u''))
else:
start_chunk.morphs.append(morph)
for morph in chunk2.morphs:
if morph.pos == u'名詞':
if not 'Y' in map(lambda m: m.surface,
end_chunk.morphs):
end_chunk.morphs.append(
Morph(u'Y', u'Y', u'名詞', u''))
else:
end_chunk.morphs.append(morph)
path = [start_chunk]
while path[-1].dst != -1 and path[-1] != chunk2:
path.append(sentence[path[-1].dst])
if path[-1] == chunk2:
# path i->j found
path = path[:-1]
line = u' -> '.join(
map(lambda chunk: chunk.surface(), path))
line = line + ' -> Y'
fout.write(line + u'\n')
else:
# path i->j not found, path represents the path i->root
# find k, where path i->root and j->root intersects
pathj = [end_chunk] # path j->root
while not pathj[-1].dst in map(lambda c: c.dst, path):
pathj.append(sentence[pathj[-1].dst])
k = pathj[-1].dst
k_in_path = 0
for pos, chunk in enumerate(path):
if chunk.dst == pathj[-1].dst:
k_in_path = pos + 1
line = ' -> '.join(
map(lambda c: c.surface(),
path[:k_in_path])) + ' | ' + ' -> '.join(
map(lambda c: c.surface(), pathj)
) + ' | ' + sentence[k].surface()
fout.write(line + u'\n')
fout.close()
def visualize(cabocha):
sentences = read_cabocha(cabocha)
g = pydot.Dot()
for sentence in sentences:
for chunk in sentence:
if chunk.morphs:
e_dst = chunk.surface()
for src in chunk.srcs:
if sentence[src].morphs:
e_src = sentence[src].surface()
g.add_edge(pydot.Edge(e_src, e_dst))
g.write_jpeg(cabocha + '.jpg')
def relations_nv():
fout = codecs.open('out43.txt', 'w', 'utf-8')
text = read_cabocha('neko.txt.cabocha')
for sentence in text:
for chunk in sentence:
if chunk.has_verb():
line = chunk.surface()
tmp = u''
for src in chunk.srcs:
if sentence[src].morphs and sentence[src].has_noun():
tmp = tmp + u'\t' + sentence[src].surface()
if tmp:
line = line + tmp + '\n'
fout.write(line)
fout.close()
def extract_verb_patterns():
text = read_cabocha('neko.txt.cabocha')
fout = codecs.open('patterns.txt', 'w', 'utf-8')
for sentence in text:
for chunk in sentence:
verb = chunk.first_verb()
if verb:
line = verb + '\t'
cases = []
for src in chunk.srcs:
case = sentence[src].case()
if case:
cases.append(case)
if cases:
line = line + u' '.join(sorted(cases))
fout.write(line + '\n')
def path_noun_to_root():
text = read_cabocha('neko.txt.cabocha')
paths = []
for sentence in text:
for chunk in sentence:
path = []
for morph in chunk.morphs:
if morph.pos == u'名詞':
path.append(chunk)
if path:
while path[-1].dst != -1:
# print path[-1].surface()
path.append(sentence[path[-1].dst])
if len(path) > 1:
paths.append(path)
return paths
def extract_verb_frames():
text = read_cabocha('neko.txt.cabocha')
fout = codecs.open('frames.txt', 'w', 'utf-8')
for sentence in text:
for chunk in sentence:
verb = chunk.first_verb()
if verb:
line = verb + '\t'
cases = []
surfaces = []
for src in chunk.srcs:
case = sentence[src].case()
if case:
cases.append(case)
surfaces.append(sentence[src].surface())
if cases:
line = line + u' '.join(sorted(cases)) + '\t' + u' '.join(
sorted(surfaces, key=lambda surface: surface[-1]))
fout.write(line + '\n')
def visualize_from_text(text, name):
tmp = codecs.open(name + '.txt', 'w', 'utf-8')
tmp.write(text)
tmp.close()
p = Popen("cabocha -f1 < " + name + ".txt > " + name + ".txt.cabocha", shell=True)
p.wait()
sentences = read_cabocha(name + '.txt.cabocha')
Popen("rm " + name + ".txt", shell=True)
Popen("rm " + name + ".txt.cabocha", shell=True)
g = pydot.Dot()
for sentence in sentences:
for chunk in sentence:
if chunk.morphs:
e_dst = chunk.surface()
for src in chunk.srcs:
if sentence[src].morphs:
e_src = sentence[src].surface()
print e_src + u'->' + e_dst
g.add_edge(pydot.Edge(e_src, e_dst))
g.write_jpg(name + '.jpg')