def parse(self, spec):
""" KNPの出力を読み取る
Args:
spec (str): KNP出力
"""
for string in spec.split('\n'):
if string.strip() == "":
continue
if string.startswith('#\t'):
items = string.split("\t")
if len(items) >= 3 and items[1] == "PAS":
self._pinfos.append(items[2])
elif string.startswith('#'):
if self.comment:
self.comment += "\n"
self.comment += string
match = re.match(r'# S-ID: ?(\S*)( .+)?$', self.comment)
if match:
self.sid = match.group(1)
if 'KNP++' in string and 'output:KNP' not in string:
self.juman_format = JUMAN_FORMAT.LATTICE_TOP_ONE # TODO
elif re.match(self.pattern, string):
break
elif string.startswith(';;'):
raise Exception("Error: %s" % string)
elif string.startswith('*'):
bnst = Bunsetsu(string, len(self._bnst))
self._bnst.append(bnst)
elif string.startswith('+'):
if self.juman_format != JUMAN_FORMAT.DEFAULT: # TODO
bnst = Bunsetsu(string, len(self._bnst), self.juman_format)
self._bnst.append(bnst)
self._bnst[-1].push_tag(
Tag(string, len(self.tag_list()), self.juman_format))
elif string.startswith('!!'):
synnodes = SynNodes(string)
self._bnst[-1].tag_list().push_synnodes(synnodes)
elif string.startswith('!') and not string.startswith('! ! !'):
synnode = SynNode(string)
self._bnst[-1].tag_list().push_synnode(synnode)
elif string.startswith('EOS'):
pass
else:
mrph = Morpheme(string, len(self.mrph_list()),
self.juman_format)
if len(self._bnst) == 0:
bnst = Bunsetsu("*", len(self._bnst))
self._bnst.append(bnst)
self._bnst[-1].push_mrph(mrph)
def _parse_spec(self, spec):
for string in spec.split('\n'):
if string.strip() == "":
continue
if string.startswith('#\t'):
items = string.split("\t")
if len(items) >= 3 and items[1] == "PAS":
self._pinfos.append(items[2])
elif string.startswith('#'):
self.comment += string
self.comment += "\n"
match = re.match(r'# S-ID:(.*?)[ $\n]', self.comment)
if match:
self.sid = match.group(1)
if 'KNP++' in string:
self.newstyle = True
elif re.match(self.pattern, string):
break
elif string.startswith(';;'):
sys.stderr.write("Error: %s\n" % string)
quit(1)
elif string.startswith('*'):
bnst = Bunsetsu(string, len(self._bnst))
self._bnst.append(bnst)
elif string.startswith('+'):
if self.newstyle:
bnst = Bunsetsu(string, len(self._bnst), self.newstyle)
self._bnst.append(bnst)
self._bnst[-1].push_tag(
Tag(string, len(self.tag_list()), self.newstyle))
elif string.startswith('!!'):
synnodes = SynNodes(string)
self._bnst[-1].tag_list().push_synnodes(synnodes)
elif string.startswith('!') and not string.startswith('! ! !'):
synnode = SynNode(string)
self._bnst[-1].tag_list().push_synnode(synnode)
elif string.startswith('EOS'):
pass
else:
mrph = Morpheme(string, len(self.mrph_list()), self.newstyle)
if not self._bnst:
bnst = Bunsetsu("*", len(self._bnst))
self._bnst.append(bnst)
self._bnst[-1].push_mrph(mrph)
def get_bnst_span(self, bnst: Bunsetsu) -> Tuple[int, int]:
""" 文節内の基本句すべてがまたがる、元文内における文字列範囲(スパン)を返す
例. '日本経済新聞の記者' において、 bnst('日本経済新聞の') のスパン -> (0,6)
"""
def merge_spans(continuous_spans):
if len(continuous_spans) < 2:
return continuous_spans[0]
return continuous_spans[0][0], continuous_spans[-1][1]
return merge_spans(
[self.get_tag_span(t.tag_id) for t in bnst.tag_list()])