def predict_line_types(self, sc_text):
'''台本テキストから行の種類を予測して改行で繋げて返す
'''
# 形態素解析器
juman = JumanPsc(command=settings.JUMAN_COMMAND,
option=settings.JUMAN_OPTION)
# 予測モデル
with open(settings.PSC_PARSE_MODEL_PATH, 'rb') as f:
tree = pickle.load(f)
# 台本を行に分けて、各行の種類を予測する
lines = sc_text.splitlines()
classes = predict(juman, tree, lines)
# 行の種類を改行文字で繋げて返す
line_types = (str(c) for c in classes)
return '\n'.join(line_types)
def main():
in_file = sys.argv[1]
out_file = sys.argv[2]
normalize = False
if len(sys.argv) > 3 and sys.argv[3] == '--normalize':
normalize = True
# 引数のファイル名を絶対パスに変換
in_path = pathlib.Path(in_file)
out_path = pathlib.Path(out_file)
in_file = str(in_path.resolve())
out_file = str(out_path.resolve())
juman = JumanPsc(command=JUMAN_COMMAND, option=JUMAN_OPTION)
model_file = os.path.join(os.path.dirname(__file__), 'model.pkl')
with open(model_file, 'rb') as f:
tree = pickle.load(f)
model.give_labels(juman, tree, in_file, out_file, normalize=normalize)
'''形態素列マッチングを単品で実行して確認する
'''
from psc_parse import JumanPsc, MrphMatch, MRPH_MTCH_PTN
from juman_settings import *
juman = JumanPsc(command=JUMAN_COMMAND, option=JUMAN_OPTION)
# line = ' 男「ようこそ皆さん、私の名前はジョニー。今日は私のリサイタルショーにおいで下さってありがとうございます。今宵、しばし皆様の時間を拝借して、私の歌をお聞きください。では、まずはこの歌から」'
line = '女「男 イヌ'
mrphs = juman.analysis(line).mrph_list()
mrph_match = MrphMatch(mrphs)
# kwargs = {
# 'ptn': (
# MrphMatch.match_spaces,
# MrphMatch.match_noun,
# MrphMatch.match_left_bracket,
# ),
# 'idx': 0
# }
kwargs = MRPH_MTCH_PTN['0002']
result = mrph_match.match(**kwargs)
print(f'マッチ判定: {result.matched}')
print(f'マッチ範囲文字列: "{result.matched_str}"')
print(f'マッチ範囲内形態素数: {result.matched_count}')
from juman_settings import JUMAN_COMMAND, JUMAN_OPTION
# 入力ディレクトリ
input_dir = 'script_samples'
# 出力ディレクトリ
output_dir = 'script_features'
# 教師ラベルファイルのディレクトリ
targets_dir = 'script_targets'
# 最初に出力ディレクトリを空にするか
empty_output_dir = True
# 特徴量を正規化するか
normalize = False
# 相対パスを絶対パスに
input_dir = os.path.join(os.path.dirname(__file__), input_dir)
output_dir = os.path.join(os.path.dirname(__file__), output_dir)
targets_dir = os.path.join(os.path.dirname(__file__), targets_dir)
# 特徴量を作る
juman = JumanPsc(command=JUMAN_COMMAND, option=JUMAN_OPTION)
make_features(juman,
input_dir,
output_dir,
targets_dir,
empty_output_dir=empty_output_dir,
normalize=normalize)
'''JumanPsc の形態素解析を単品の文字列で試す
'''
from psc_parse import JumanPsc
from juman_settings import JUMAN_COMMAND, JUMAN_OPTION
juman = JumanPsc(command=JUMAN_COMMAND, option=JUMAN_OPTION)
s = '今日もいい天気'
try:
mrphs = juman.analysis(s)
for mrph in mrphs:
print(f'"{mrph.midasi}"') # 見出し
print(f' genkei: {mrph.genkei}') # 原形
print(f' hinsi: {mrph.hinsi}') # 品詞
print(f' bunrui: {mrph.bunrui}') # 品詞細分類
print(f' katuyou2: {mrph.katuyou2}') # 活用形
except Exception as e:
print(e)