def __init__(self, command='knp', option='-tab', rcfile='',
server=None, port=31000, timeout=30,
pattern=r'(?:^|\n)EOS($|\n)',
jumanrcfile='', juman_option='-e2 -B', juman_port=32000,
juman_command='juman', jumanpp=False):
self.use_jumanpp = (juman_command == "jumanpp") or jumanpp
assert 'EOS' in pattern
self.pattern = pattern
self.EOS = 'EOS'
# tab形式しかパースしない
assert '-tab' in option
if rcfile and not os.path.isfile(os.path.expanduser(rcfile)):
sys.stderr.write("Can't read rcfile (%s)!\n" % rcfile)
quit(1)
# Setup Juman(++)
assert port != juman_port
juman_args = {'option': juman_option, 'rcfile': jumanrcfile,
'server':server, 'port':juman_port}
if self.use_jumanpp:
self.juman = Jumanpp(**juman_args)
else:
self.juman = Juman(**juman_args)
# Setup KNP
if server is not None:
self.socket = Socket(server, port, option=option, timeout=timeout)
self.query = partial(self.socket.query, pattern=pattern)
else:
if rcfile:
option += " -r {}".format(rcfile)
self.subprocess = Subprocess(command, option=option)
self.query = partial(self.subprocess.query, pattern=pattern)
def analyzer():
bc = BertClient(ip='bertserving', output_fmt='list')
client = Elasticsearch('elasticsearch:9200')
texts = []
list_text = []
jumanpp = Juman()
query = request.args.get('q')
result = jumanpp.analysis(query)
for mrph in result.mrph_list():
texts.append(mrph.midasi)
list_text.append(" ".join(texts))
query_vector = bc.encode(list_text, is_tokenized=False)[0]
script_query = {
"script_score": {
"query": {
"match": {
"source": "tb"
}
},
"script": {
"source":
"cosineSimilarity(params.query_vector, doc['question_vector']) + 1.0",
"params": {
"query_vector": query_vector
}
}
}
}
response = client.search(index=INDEX_NAME,
body={
"size": SEARCH_SIZE,
"query": script_query
})
return jsonify(response)
def __init__(self):
"""Constructs a Tokenizer for Juman++.
"""
from pyknp import Juman
self.do_lower_case = False
self._jumanpp = Juman()
def __init__(self,
command='knp',
server=None,
port=31000,
timeout=60,
option='-tab',
rcfile='',
pattern=r'EOS',
jumancommand='juman',
jumanrcfile='',
jumanpp=False):
self.command = command
self.server = server
self.port = port
self.timeout = timeout
self.option = option
self.rcfile = rcfile
self.pattern = pattern
self.socket = None
self.subprocess = None
self.jumanpp = (jumancommand == "jumanpp") or jumanpp
if self.rcfile and not os.path.isfile(os.path.expanduser(self.rcfile)):
sys.stderr.write("Can't read rcfile (%s)!\n" % self.rcfile)
quit(1)
if (self.jumanpp):
self.juman = Jumanpp()
else:
self.juman = Juman(command=jumancommand, rcfile=jumanrcfile)
def _apply_jumanpp(self, inp: str) -> Tuple[str, str]:
jumanpp = Juman(command=self.juman, option=self.juman_option)
jumanpp_result = jumanpp.analysis(inp)
jumanpp_out = jumanpp_result.spec() + 'EOS\n'
jumanpp_conll_out = self._jumanpp2conll_one_sentence(
jumanpp_out) + 'EOS\n'
return jumanpp_out, jumanpp_conll_out
def __init__(self,
command='knp',
server=None,
port=31000,
timeout=60,
option='-tab',
rcfile='',
pattern=r'EOS',
jumancommand='jumanpp',
jumanrcfile='',
jumanoption='',
jumanpp=True):
self.command = command
self.server = server
self.port = port
self.timeout = timeout
self.options = option.split()
self.rcfile = rcfile
self.pattern = pattern
self.socket = None
self.subprocess = None
self.jumanpp = jumanpp
if self.rcfile and not os.path.isfile(os.path.expanduser(self.rcfile)):
raise Exception("Can't read rcfile (%s)!" % self.rcfile)
if distutils.spawn.find_executable(self.command) is None:
raise Exception("Can't find KNP command: %s" % self.command)
self.juman = Juman(command=jumancommand,
rcfile=jumanrcfile,
option=jumanoption,
jumanpp=self.jumanpp)
def main(DATA_ROOT):
text_files = Path(DATA_ROOT).glob('**/*.txt')
for text_file in text_files:
with open(text_file) as f:
content = f.read()
content = re.sub(r"=+(.*?)=+", "\g<1>", content)
content = re.sub(r"^\n", "", content, flags=re.MULTILINE)
content = content.replace('<block>', '')
content = content.replace('<math-element>', '')
# In this case, 。 can be removed safely
sentences = re.split(r"[。\n]", content)
sentences = [line for line in sentences if len(line) != 0]
sentences = [''.join(line.split()) for line in sentences]
# Remove sentence which is not properly parsed
val_sentences = []
offsets = []
juman = Juman()
for sentence in tqdm(sentences):
# Try to parse
try:
result = juman.analysis(sentence)
except ValueError:
print(sentence)
except Exception as e:
raise e
current = 0
offset = [0 for _ in range(len(sentence))]
for mrph in result.mrph_list():
current = current + len(mrph.midasi)
try:
offset[current - 1] = 1
except IndexError as e:
print(sentence)
print(current)
for _mrph in result.mrph_list():
print(_mrph.midasi)
raise e
except Exception as e:
raise e
val_sentences.append(sentence)
offsets.append(offset)
results = (sentences, offsets)
file_name = text_file.name[:-4] + '.pickle'
dic = text_file.parent
with open(Path(dic, file_name), 'wb') as f:
pickle.dump(results, f)
def run(self):
data = self.load()
jumanpp = Juman()
output = []
for _, row in data.iterrows():
zenkaku = jaconv.h2z(row["sentence"], ascii=True, digit=True)
splited = [
mrph.midasi for mrph in jumanpp.analysis(zenkaku).mrph_list()
]
if self.task_name == 'QA_B':
qa_zenkaku = jaconv.h2z(
f"{row['target']}の{row['aspect']}は{row['sentiment']}",
ascii=True,
digit=True,
)
else:
qa_zenkaku = " "
qa_splited = [
mrph.midasi
for mrph in jumanpp.analysis(qa_zenkaku).mrph_list()
]
output.append({
"context": " ".join(splited),
"qa": " ".join(qa_splited),
"label": 1
})
self.dump(pd.DataFrame(output))
def initialize(fword, tword, modelfn, start, debug):
juman = Juman()
# parse and check from_word
ms_f = juman.analysis(fword).mrph_list()
if len(ms_f) > 1:
print(u'{} is parsed multiple words'.format(fword))
exit(1)
wm_f = ms_f[0]
if not wm_f.repname:
print(u'no repname with {}'.format(fword))
exit(1)
fword = wm_f.repname
# parse and check to_word
ms_t = juman.analysis(tword).mrph_list()
if len(ms_t) > 1:
print(u'{} is parsed multiple words'.format(tword))
exit(1)
wm_t = ms_t[0]
if not wm_t.repname:
print(u'no repname with {}'.format(tword))
exit(1)
tword = wm_t.repname
# load and check model
print(u'loading model...')
if modelfn.split('.')[-1] == 'model':
model = Word2Vec.load(modelfn)
elif modelfn.split('.')[-1] == 'bin':
model = Word2Vec.load_word2vec_format(modelfn, binary=True, unicode_errors='ignore')
if fword not in model.vocab:
raise KeyError(u'{} is not found in the model'.format(fword))
exit(1)
elif tword not in model.vocab:
raise KeyError(u'{} is not found in the model'.format(tword))
exit(1)
model.save('hs0.100m.500.5.18mgt100.model')
t1 = time.clock() - start
if debug:
printtime(t1)
print(u'constructing id2vocab map...')
id2vocab = {}
for i, v in enumerate(model.vocab):
id2vocab[i] = v
t2 = time.clock() - t1
if debug:
printtime(t2)
print(u'constructing V...')
V = []
for v in model.vocab:
V.append(model[v])
V = np.vstack(V)
t3 = time.clock() - t2
if debug:
printtime(t3)
return fword, tword, model, V, id2vocab, t3
def string_word_point(self, df):
jumanpp = Juman(jumanpp=False)
tmp_word =[]
df_time_word = pd.DataFrame(index=[], columns=['time','word']) #単語と時間のｄｆ
df_word_point = pd.DataFrame(index=[], columns=['word','point'])#単語とその出現数のｄｆ
df_time_point = pd.DataFrame(index=[], columns=['time','point'])#時間とその時のコメント数のｄｆ
df_time_www_point = pd.DataFrame(index=[], columns=['time','point'])#時間とその時のwww数のｄｆ
df_time_hakusyu_point = pd.DataFrame(index=[], columns=['time','point'])#時間とその時の拍手数のｄｆ
df_URL_point = pd.DataFrame(index=[], columns=['URL','point'])#URLまとめdf
#print(df_word_point)
for i in range(len(df)):
#URLだったら追加
url=URL_hanbetu(df['comment'][i])
if url != False:
tmp = self.my_index(df_URL_point['URL'],url)
df_URL_point = self.make_df_append(df_URL_point,tmp,url)
#print("記号削除前")
#print(df_word_point)
#記号削除中
print(df['comment'][i])
df['comment'][i] = self.my_delete(df['comment'][i])
# h:m:s -> hms に変更
tmp_time = self.strtime_to_inttime(df['time'][i])
#時間ごとのコメント数計算
tmp = self.my_index(df_time_point['time'],tmp_time)
df_time_point = self.make_df_append(df_time_point,tmp,tmp_time)
#wwwがあったら1追加なかったら0追加
print(url)
if False != self.www_hanbetu(df['comment'][i]) and url == False:
df_time_www_point = self.make_df_append(df_time_www_point,tmp,tmp_time)
else:
if False == tmp :
df_time_www_point = df_time_www_point.append({'time': tmp_time, 'point': 0}, ignore_index=True)
#拍手があったら1追加なかったら0追加
if False != self.hakusyu_hanbetu(df['comment'][i]):
df_time_hakusyu_point = self.make_df_append(df_time_hakusyu_point,tmp,tmp_time)
else:
if False == tmp :
df_time_hakusyu_point = df_time_hakusyu_point.append({'time': tmp_time, 'point': 0}, ignore_index=True)
#構文解析
result = jumanpp.analysis(df['comment'][i])
#print(result)
#分析結果からdf作成
for token in result.mrph_list():
tmp_word = token.midasi
#名詞の出現数計算
if 0 != self.word_Classification(token.hinsi):
#名詞なら
if self.word_Classification(token.hinsi) == '名詞':
tmp = self.my_index(df_word_point['word'],tmp_word)
df_word_point = self.make_df_append(df_word_point,tmp,tmp_word)
#名詞とその時の時間
df_time_word = df_time_word.append({'time':tmp_time,'word': tmp_word}, ignore_index=True)
return df_time_word,df_word_point,df_time_point,df_time_www_point, df_time_hakusyu_point,df_URL_point
def __init__(self, bert_model: str, fine_tuned_model: str, jumanpp_command: str):
self.jumanpp = Juman(command=jumanpp_command)
self.tokenizer = BertTokenizer.from_pretrained(bert_model, do_lower_case=False)
self.model = BertPosNegClassifier(bert_model)
state_dict = torch.load(fine_tuned_model, map_location=torch.device('cpu'))
self.model.load_state_dict({k.replace('module.', ''): v for k, v in state_dict.items()})
self.model.eval()
def __init__(
self,
preprocessor=None,
stopwords=[],
):
self.jumanpp = Juman()
self.preprocessor = preprocessor
self.stopwords = stopwords
def counter(text, d):
jumanapp = Juman()
result = jumanapp.analysis(text)
for mrph in result.mrph_list():
if mrph.genkei in d:
d[mrph.genkei] = d[mrph.genkei] + 1
else:
d[mrph.genkei] = 1
def morphological_analysis(self, text):
jumanpp = Juman()
ret = []
text = self.remove_special_character(text)
result = jumanpp.analysis(text) # これでスペースで単語が区切られる
for mrph in result.mrph_list():
ret += self.modification(mrph.midasi)
return ret
def juman_list(text):
jumanpp = Juman()
result = jumanpp.analysis(text)
# アルファベットは全て "En" という文字列に置き換える
wakati = [
mrph.genkei if mrph.bunrui != "アルファベット" else "En"
for mrph in result.mrph_list()
]
return ",".join(wakati)
def juman_test():
juman = Juman()
print dir(juman)
text = "テストテキスト"
utext = unicode("".join(text.split()))
print( u'"'+utext+u'"' )
juman_result = juman.analysis( utext )
for mrph in juman_result.mrph_list():
print( '> ' + mrph.midasi + ' : ' + mrph.yomi + ' : ' + mrph.genkei )
def test_juman_wrapper(self):
juman = Juman()
result = juman.analysis(u"これはペンです。")
print(','.join(mrph.midasi for mrph in result))
for mrph in result.mrph_list():
assert isinstance(mrph, pyknp.Morpheme)
print(u"見出し:%s, 読み:%s, 原形:%s, 品詞:%s, 品詞細分類:%s, 活用型:%s, 活用形:%s, 意味情報:%s, 代表表記:%s" \
% (mrph.midasi, mrph.yomi, mrph.genkei, mrph.hinsi, mrph.bunrui, mrph.katuyou1, mrph.katuyou2, mrph.imis, mrph.repname))
def test_juman_wrapper(self):
juman = Juman()
result = juman.analysis(u"これはペンです。")
print(','.join(mrph.midasi for mrph in result))
for mrph in result.mrph_list():
assert isinstance(mrph, pyknp.Morpheme)
print(u"見出し:%s, 読み:%s, 原形:%s, 品詞:%s, 品詞細分類:%s, 活用型:%s, 活用形:%s, 意味情報:%s, 代表表記:%s" \
% (mrph.midasi, mrph.yomi, mrph.genkei, mrph.hinsi, mrph.bunrui, mrph.katuyou1, mrph.katuyou2, mrph.imis, mrph.repname))
def __init__(self):
"""
日本語専用トークナイザの構築。
JUMAN++ を使用する。
"""
from pyknp import Juman
self.do_lower_case = False
self._jumanpp = Juman()
def get_repname_using_jumanpp(genkei: str, pos: str) -> str:
if pos == '助詞':
return f'{genkei}/{genkei}'
juman = Juman(option='-s 1')
mrphs = juman.analysis(genkei, juman_format=JUMAN_FORMAT.LATTICE_TOP_ONE)
# 形態素解析が誤っていないか(=1形態素になっているか)をチェック
if len(mrphs) == 1:
return mrphs[0].repname
return f'{genkei}/{genkei}'
def __init__(self):
'''initialize
Examples
--------
>>> nlp = JNLP()
None
'''
self.juman = Juman()
self.KNP = KNP(option='-tab -anaphora')
def test_juman_wrapper(self):
try:
juman = Juman(command=self.path_to_juman_command)
result = juman.analysis("これはペンです。")
logger.debug(','.join(mrph.midasi for mrph in result))
for mrph in result.mrph_list():
assert isinstance(mrph, pyknp.Morpheme)
logger.debug("見出し:%s, 読み:%s, 原形:%s, 品詞:%s, 品詞細分類:%s, 活用型:%s, 活用形:%s, 意味情報:%s, 代表表記:%s" \
% (mrph.midasi, mrph.yomi, mrph.genkei, mrph.hinsi, mrph.bunrui, mrph.katuyou1, mrph.katuyou2, mrph.imis, mrph.repname))
except ImportError:
print('skip test_juman_wrapper')
def bulk_predict(docs, batch_size=256):
"""Predict bert embeddings."""
jumanpp = Juman(jumanpp=False)
for i in range(0, len(docs), batch_size):
batch_docs = docs[i: i+batch_size]
pre_embedding_docs = []
for doc in batch_docs:
for k in range(0, len(doc['question']), MAX_TXT_LENGTH)
result = jumanpp.analysis(doc['question'][k:k+MAX_TXT_LENGTH])
texts = [mrph.midasi for mrph in result.mrph_list()]
pre_embedding_docs.append(" ".join(texts))
embeddings = bc.encode(pre_embedding_docs,is_tokenized=True)
yield emb
def seg2word(seg):
len_split = 1000
# seg = seg_in.replace(' ', '\u3000')
# seg = seg_in.replace(' ', ' ')
len_seg = len(seg)
seg_splits = [seg[i:i + len_split] for i in range(0, len_seg, len_split)]
juman_def = Juman(command="/mnt/gold/users/s18153/bin/jumanpp")
return ' '.join([
" ".join(
[mrph.midasi for mrph in juman_def.analysis(seg_part).mrph_list()])
for seg_part in seg_splits
])
def juman_wakati(text, hinshi=(), DEBUG=False, STEM_FLAG=False):
juman = Juman()
output = ""
# wakati
result = juman.analysis(text)
for mrph in result.mrph_list():
if STEM_FLAG and mrph.hinsi in hinshi:
output += mrph.repname.split("/")[0] + " "
if DEBUG:
print("stem:", mrph.repname)
print("midashi:", mrph.repname)
print("hinsi:", mrph.hinsi)
print("yomi:", mrph.yomi)
return output.strip()
def __init__(
self,
command='knp',
server=None,
port=31000,
timeout=60,
option='-tab',
rcfile='',
pattern=r'EOS',
jumancommand='jumanpp',
jumanrcfile='',
jumanoption='',
jumanpp=True,
multithreading=False,
):
self.command = command
self.server = server
self.port = port
self.timeout = timeout
self.options = option.split()
self.rcfile = rcfile
self.pattern = pattern
if server is not None:
self.analyzer = Analyzer(backend='socket',
timeout=timeout,
server=server,
port=port,
socket_option='RUN -tab -normal\n')
else:
cmds = [self.command] + self.options
if self.rcfile:
cmds += ['-r', self.rcfile]
self.analyzer = Analyzer(backend='subprocess',
multithreading=multithreading,
timeout=timeout,
command=cmds)
self.jumanpp = jumanpp
if self.rcfile and not os.path.isfile(os.path.expanduser(self.rcfile)):
raise Exception("Can't read rcfile (%s)!" % self.rcfile)
if distutils.spawn.find_executable(self.command) is None:
raise Exception("Can't find KNP command: %s" % self.command)
self.juman = Juman(command=jumancommand,
rcfile=jumanrcfile,
option=jumanoption,
jumanpp=self.jumanpp,
multithreading=multithreading)
def read_home_timeline( session ):
print( '[kazuha] - read timeline.' )
juman = Juman()
req = session.get( twitter.API_home_timeline, params = {} )
if req.status_code == 200:
timeline = json.loads( req.text )
for tweet in timeline:
u_tweet_text = unicode( "".join(tweet["text"].split()) )
print( u'[kazuha] - read timeline: '+ u_tweet_text )
juman_result = juman.analysis( u_tweet_text )
for mrph in juman_result.mrph_list():
print u"%s - (%s, %s)" % (mrph.genkei, mrph.hinsi, mrph.bunrui)
#end for
#end for
else:
print( '[kazuha] - read timeline: failure.' )
class JumanTokenizer:
def __init__(self, command, options):
self.juman = Juman(command, options)
def tokenize(self, text):
result = self.juman.analysis(text)
return [mrph.midasi for mrph in result.mrph_list()]
def main(bert_vocab_filepath, example_filepath, context_filepath,
cache_save_dir):
#Juman++
juman = Juman(jumanpp=True)
logger.info("Cache files will be saved in {}.".format(cache_save_dir))
#Tokenizer
logger.info("Create a tokenizer from {}.".format(bert_vocab_filepath))
tokenizer = BertTokenizer.from_pretrained(bert_vocab_filepath,
do_lower_case=False)
logger.info("Start loading examples from {}.".format(example_filepath))
examples = load_examples(example_filepath)
logger.info("Finished loading examples.")
logger.info("Number of examples: {}".format(len(examples)))
logger.info("Start loading contexts from {}.".format(context_filepath))
contexts = load_contexts(context_filepath)
logger.info("Finished loading contexts.")
logger.info("Start encoding examples.")
encoding = encode_examples(juman, tokenizer, examples, contexts, 512)
logger.info("Finished encoding examples.")
os.makedirs(cache_save_dir, exist_ok=True)
torch.save(encoding["input_ids"],
os.path.join(cache_save_dir, "input_ids.pt"))
torch.save(encoding["attention_mask"],
os.path.join(cache_save_dir, "attention_mask.pt"))
torch.save(encoding["token_type_ids"],
os.path.join(cache_save_dir, "token_type_ids.pt"))
torch.save(encoding["labels"], os.path.join(cache_save_dir, "labels.pt"))
logger.info("Saved cache files in {}.".format(cache_save_dir))
def title_clean(title_ls):
tmp_ls = copy.deepcopy([title_ls])
for i in range(len(tmp_ls) - 1):
if tmp_ls[i] is None:
del tmp_ls[i]
for i in range(len(tmp_ls)):
tmp_ls[i] = normalize('NFKC', tmp_ls[i])
tmp_ls[i] = tmp_ls[i].replace(' ', '')
tmp_ls[i] = re.sub(r'−.+?$', '', tmp_ls[i])
tmp_ls[i] = re.sub(r'ーY.+?$', '', tmp_ls[i])
tmp_ls[i] = re.sub(r'\|.+?$', '', tmp_ls[i])
jumanpp = Juman()
sep_ls = []
for tmp in tmp_ls:
sep_ls.append(' '.join([mrph.midasi for mrph in jumanpp.analysis(tmp)]))
return sep_ls[0]
class JumanTokenizer():
def __init__(self):
self.juman = Juman(jumanpp=True)
def tokenize(self, text):
result = self.juman.analysis(text)
return [mrph.midasi for mrph in result.mrph_list()]
class JumanService(object):
def __init__(self):
self.__juman = Juman()
def analysis(self, string):
formattedString = JumanKnpUtil.format_input_string(string)
return self.__juman.analysis(formattedString)
class JumanTokenizer():
_trans_tables = str.maketrans({"\"": "", "@": "＠", "#": "＃"})
def __init__(self, ):
self.juman = Juman()
def _preprocess(self, sentences):
return sentences.replace(" ",
"").replace("\n",
"").translate(self._trans_tables)
def tokenize(self, text):
result = self.juman.analysis(text)
return [mrph.midasi for mrph in result.mrph_list()]
def _preprocess_list(self, datas):
return [[x, self._preprocess(x)] for x in datas]
def _tokenize_for_multi(self, datas):
try:
return [datas[0], self.tokenize(datas[1])]
except:
return []
def tokenize_multi(self, datas, thread=cpu_count()):
datas = self._preprocess_list(datas)
num_of_datas = len(datas)
with Pool(thread) as pool:
imap = pool.imap_unordered(self._tokenize_for_multi, datas)
result = list(tqdm(imap, total=num_of_datas))
return result
def word_distance(s1, s2):
juman = Juman()
r = len(s1 + s2) - len(re.sub("[a-zA-Z0-9]", "", s1 + s2))
if r > len((s1 + s2).replace(" ", "")) // 2:
return word_distance_en(s1, s2)
sss = [
set(
[item.midasi for item in juman.analysis(ss).mrph_list() \
if item.hinsi in {'名詞', '動詞', '形容詞', '指示詞'}\
or '内容語' in item.imis
]
) for ss in [s1, s2]
]
if min(len(sss[0]), len(sss[1])) == 0:
return 0
return float(len(sss[0] & sss[1])) / min(len(sss[0]), len(sss[1]))
def __init__(self, command='knp', server=None, port=31000, timeout=60,
option='-tab', rcfile='', pattern=r'EOS',
jumancommand='jumanpp', jumanrcfile='', jumanpp=True):
self.command = command
self.server = server
self.port = port
self.timeout = timeout
self.option = option.split()
self.rcfile = rcfile
self.pattern = pattern
self.socket = None
self.subprocess = None
self.jumanpp = jumanpp
if self.rcfile and not os.path.isfile(os.path.expanduser(self.rcfile)):
raise Exception("Can't read rcfile (%s)!" % self.rcfile)
if distutils.spawn.find_executable(self.command) is None:
raise Exception("Can't find KNP command: %s" % self.command)
self.juman = Juman(command=jumancommand, rcfile=jumanrcfile, jumanpp=self.jumanpp)
def __init__(self):
self.juman = Juman()
self.knp = KNP()
# coding: utf-8
from pyknp import Juman
import sys
import codecs
juman = Juman()
input_file = "../data/sample.txt"
f = codecs.open(input_file, 'r', 'utf-8')
f_out = codecs.open(input_file + '_juman_result.txt','w', 'utf-8')
for line in f:
result = juman.analysis(line[:-1].replace(" ", ""))
#print ' '.join(mrph.midasi for mrph in result)
f_out.write(' '.join(mrph.midasi for mrph in result) + '\n')
class Solver(object):
def __init__(self):
self.juman = Juman()
self.knp = KNP()
def Q61(self):
u"""61. 文を標準入力から読み込み、それを単語単位に分かち書きせよ (形態素間にスペースを挿入)
"""
input_sentence = raw_input()
result = self.juman.analysis(input_sentence.decode("utf8"))
for mrph in result.mrph_list():
sys.stdout.write("{} ".format(mrph.midasi.encode("utf8")))
sys.stdout.write("\n")
return
def Q62(self):
u"""62. 形態素解析結果を読み込み、名詞だけを抽出してプリントせよ
ヒント: mrph.hinsi が u"名詞" という文字列と一致するかどうかを判定
"""
data = u""
for line in iter(sys.stdin.readline, ""): # 入力文を1行ずつ読む
data += line.decode("utf8")
if line.strip() == "EOS": # 1文が終わったら解析
result = self.juman.result(data)
s = ",".join(mrph.midasi for mrph in result.mrph_list() if mrph.hinsi == u"名詞") # 名詞だけ表示
if len(s) > 0:
print(s)
data = u""
def Q63(self):
u"""62. 形態素解析結果を読み込み、名詞だけを抽出してプリントせよ
ヒント: mrph.hinsi が u"名詞" という文字列と一致するかどうかを判定
"""
data = u""
for line in iter(sys.stdin.readline, ""): # 入力文を1行ずつ読む
data += line.decode("utf8")
if line.strip() == "EOS": # 1文が終わったら解析
result = self.juman.result(data)
s = ",".join(mrph.genkei for mrph in result.mrph_list() if mrph.hinsi == u"動詞") # 動詞だけ表示
if len(s) > 0:
print(s)
data = u""
def Q64(self):
u"""64. 形態素解析結果を読み込み、形態素の原形を頻度順に並べよ
ヒント: ディクショナリ、sorted 関数を使う
"""
data = u""
hist = {}
for line in iter(sys.stdin.readline, ""): # 入力文を1行ずつ読む
data += line.decode("utf8")
if line.strip() == "EOS": # 1文が終わったら解析
result = self.juman.result(data)
for mrph in result.mrph_list():
try:
hist[mrph.genkei] += 1
except KeyError:
hist[mrph.genkei] = 1
data = u""
for key, val in sorted(hist.items(), key=lambda t: t[1], reverse=True):
print("{},{}".format(key.encode("utf8"), val))
def Q65(self):
u"""65. 形態素解析結果を読み込み、全形態素数 (総数) に対する述語の割合を計算せよ
ここで、述語とは、動詞、イ形容詞 (形容詞)、ナ形容詞 (形容動詞) とする
"""
data = u""
num = 0
denom = 0
for line in iter(sys.stdin.readline, ""): # 入力文を1行ずつ読む
data += line.decode("utf8")
if line.strip() == "EOS": # 1文が終わったら解析
result = self.juman.result(data)
if verbose:
logger.info("denom: {}".format(denom))
for mrph in result.mrph_list():
denom += 1
if mrph.hinsi == u"動詞":
num += 1
continue
if mrph.hinsi == u"形容詞" and mrph.bunrui.startswith(u"イ形容詞"):
num += 1
continue
if mrph.hinsi == u"形容動詞" and mrph.bunrui.startswith(u"ナ形容詞"):
num += 1
continue
data = u""
print("{}/{}={}".format(num, denom, float(num) / denom))
def Q66(self):
u"""66. 形態素解析結果を読み込み、「サ変名詞+する/できる」というパターンを抽出しプリントせよ
"""
data = u""
extract = set()
for line in iter(sys.stdin.readline, ""): # 入力文を1行ずつ読む
data += line.decode("utf8")
if line.strip() == "EOS": # 1文が終わったら解析
result = self.juman.result(data)
buff = None
for mrph in result.mrph_list():
if mrph.genkei == u"できる" or mrph.genkei == u"する":
if buff is not None:
extract.add((buff.genkei.encode("utf8"), mrph.genkei.encode("utf8")))
if mrph.bunrui == u"サ変名詞":
buff = mrph
else:
buff = None
data = u""
for t in extract:
print("{}+{}".format(t[0], t[1]))
def Q67(self):
u"""67. 形態素解析結果を読み込み、「AのB」という表現 (A と B は名詞の1形態素) をすべてプリントせよ
"""
data = u""
extract = set()
for line in iter(sys.stdin.readline, ""): # 入力文を1行ずつ読む
data += line.decode("utf8")
if line.strip() == "EOS": # 1文が終わったら解析
result = self.juman.result(data)
buff = []
for mrph in result.mrph_list():
if mrph.genkei == u"の" and len(buff) == 1:
buff.append(u"の")
continue
if mrph.hinsi == u"名詞":
if len(buff) == 0:
buff.append(mrph.genkei)
continue
if len(buff) == 2:
extract.add((buff[0], mrph.genkei))
buff = []
data = u""
for t in extract:
print("{}の{}".format(t[0].encode("utf8"), t[1].encode("utf8")))
def Q68(self):
u"""68. 文を標準入力から読み込み、それを文節単位に分かち書きせよ (文節間にスペースを挿入)
"""
input_sentence = raw_input()
result = self.knp.parse(input_sentence.decode("utf8"))
for bnst in result.bnst_list():
sys.stdout.write("{} ".format("".join(mrph.midasi.encode("utf8") for mrph in bnst.mrph_list())))
sys.stdout.write("\n")
return
def Q69(self):
u"""69. 構文解析結果を読み込み、接頭辞を含む文節をプリントせよ
"""
data = u""
extract = set()
for line in iter(sys.stdin.readline, ""):
data += line.decode("utf8")
if line.strip() == "EOS":
result = self.knp.result(data)
for bnst in result.bnst_list():
if len(filter(lambda x: x.hinsi == u"接頭辞", bnst.mrph_list())) < 1:
continue
extract.add("{} ".format("".join(mrph.midasi.encode("utf8") for mrph in bnst.mrph_list())))
data = u""
for bnst in extract:
if len(bnst) > 0:
print(bnst)
return
def Q70(self):
u"""70. 構文解析結果を読み込み、名詞を2つ以上含む文節をプリントせよ
"""
data = u""
extract = set()
for line in iter(sys.stdin.readline, ""):
data += line.decode("utf8")
if line.strip() == "EOS":
result = self.knp.result(data)
for bnst in result.bnst_list():
if len(filter(lambda x: x.hinsi == u"名詞", bnst.mrph_list())) < 2:
continue
extract.add("{} ".format("".join(mrph.midasi.encode("utf8") for mrph in bnst.mrph_list())))
data = u""
for bnst in extract:
if len(bnst) > 0:
print(bnst)
return
class KNP(object):
""" KNPを用いて構文解析を行う/KNPの解析結果を読み取るモジュール
Args:
command (str): KNPコマンド
option (str): KNP解析オプション
(詳細解析結果を出力する-tabは必須。
省略・照応解析を行う -anaphora, 格解析を行わず構文解析のみを行う -dpnd など)
rcfile (str): KNP設定ファイルへのパス
pattern (str): KNP出力の終端記号
jumancommand (str): JUMANコマンド
jumanrcfile (str): JUMAN設定ファイルへのパス
jumanpp (bool): JUMAN++を用いるかJUMANを用いるか
"""
def __init__(self, command='knp', server=None, port=31000, timeout=60,
option='-tab', rcfile='', pattern=r'EOS',
jumancommand='jumanpp', jumanrcfile='', jumanpp=True):
self.command = command
self.server = server
self.port = port
self.timeout = timeout
self.option = option.split()
self.rcfile = rcfile
self.pattern = pattern
self.socket = None
self.subprocess = None
self.jumanpp = jumanpp
if self.rcfile and not os.path.isfile(os.path.expanduser(self.rcfile)):
raise Exception("Can't read rcfile (%s)!" % self.rcfile)
if distutils.spawn.find_executable(self.command) is None:
raise Exception("Can't find KNP command: %s" % self.command)
self.juman = Juman(command=jumancommand, rcfile=jumanrcfile, jumanpp=self.jumanpp)
def knp(self, sentence):
""" parse関数と同じ """
self.parse(sentence)
def parse(self, sentence, juman_format=JUMAN_FORMAT.DEFAULT):
"""
文字列を入力として構文解析を行い、文節列オブジェクトを返す
Args:
sentence (str): 文を表す文字列
juman_format (JUMAN_FORMAT): Jumanのlattice出力形式
Returns:
BList: 文節列オブジェクト
"""
assert(isinstance(sentence, six.text_type))
juman_lines = self.juman.juman_lines(sentence)
juman_str = "%s%s" % (juman_lines, self.pattern)
if not self.socket and not self.subprocess:
if self.server is not None:
self.socket = Socket(
self.server, self.port, "RUN -tab -normal\n")
else:
command = [self.command] + self.option
if self.rcfile:
command.extend(['-r', self.rcfile])
self.subprocess = Subprocess(command)
if self.socket:
knp_lines = self.socket.query(juman_str, pattern=r'^%s$'%(self.pattern))
else:
knp_lines = self.subprocess.query(juman_str, pattern=r'^%s$'%(self.pattern))
return BList(knp_lines, self.pattern, juman_format)
def result(self, input_str, juman_format=JUMAN_FORMAT.DEFAULT):
"""
ある文に関するKNP解析結果を文節列オブジェクトに変換する
Args:
input_str (str): ある文に関するKNPの出力結果
juman_format (JUMAN_FORMAT): Jumanのlattice出力形式
Returns:
BList: 文節列オブジェクト
"""
return BList(input_str, self.pattern, juman_format)
