def vectorize(docs, model):
try:
mecab = MeCab.Tagger("-Ochasen")
data = mecab.parse(docs)
data = StringIO(data.replace("\"", ""))
data = pd.read_csv(data, sep='\t', header=None)
data = data.loc[(data[3].str.find("名詞") >= 0)
& (data[3].str.find("接頭詞") < 0) &
(data[3].str.find("サ変接続") < 0)]
data = data[data[0].isin(
pd.DataFrame.from_dict(model.wv.vocab, orient='index').index)]
data = data[[0]]
vec = pd.DataFrame(
data.applymap(model.wv.word_vec)[0].values.mean(axis=0))
return vec
except:
return []
