# !wget https://dl.fbaipublicfiles.com/fasttext/vectors-crawl/cc.ja.300.vec.gzで落とせます model = gensim.models.KeyedVectors.load_word2vec_format('cc.ja.300.vec.gz', binary=False) repat = re.compile(r'^[あ-ん\u30A1-\u30F4\u4E00-\u9FD0]+$') vocab_list = [w for w in list(model.vocab.keys())[10000:50000] if len(w) > 2 and repat.fullmatch(w) and w[-1] != 'っ' and w not in list(ww_df.word) and w not in list(sw