言語処理100本ノック 2020「43. 名詞を含む文節が動詞を含む文節に係るものを抽出」

問題文

nlp100.github.io

問題の概要

「42. 係り元と係り先の文節の表示」に「名詞を含む文節が，動詞を含む文節に係るとき」の条件を付与します。

class Morph:
    def __init__(self, dc):
        self.surface = dc['surface']
        self.base = dc['base']
        self.pos = dc['pos']
        self.pos1 = dc['pos1']


class Chunk:
    def __init__(self, morphs, dst):
        self.morphs = morphs    # 形態素（Morphオブジェクト）のリスト
        self.dst = dst          # 係り先文節インデックス番号
        self.srcs = []          # 係り元文節インデックス番号のリスト


def parse_cabocha(block):
    def check_create_chunk(tmp):
        if len(tmp) > 0:
            c = Chunk(tmp, dst)
            res.append(c)
            tmp = []
        return tmp

    res = []
    tmp = []
    dst = None
    for line in block.split('\n'):
        if line == '':
            tmp = check_create_chunk(tmp)
        elif line[0] == '*':
            dst = line.split(' ')[2].rstrip('D')
            tmp = check_create_chunk(tmp)
        else:
            (surface, attr) = line.split('\t')
            attr = attr.split(',')
            lineDict = {
                'surface': surface,
                'base': attr[6],
                'pos': attr[0],
                'pos1': attr[1]
            }
            tmp.append(Morph(lineDict))

    for i, r in enumerate(res):
        res[int(r.dst)].srcs.append(i)
    return res


filename = 'ch05/ai.ja.txt.cabocha'
with open(filename, mode='rt', encoding='utf-8') as f:
    blocks = f.read().split('EOS\n')
blocks = list(filter(lambda x: x != '', blocks))
blocks = [parse_cabocha(block) for block in blocks]

for b in blocks:
    for m in b:
        if int(m.dst) > -1:
            pre_text = ''.join([mo.surface if mo.pos != '記号' else '' for mo in m.morphs])
            pre_pos = [mo.pos for mo in m.morphs]
            post_text = ''.join([mo.surface if mo.pos != '記号' else '' for mo in b[int(m.dst)].morphs])
            post_pos = [mo.pos for mo in b[int(m.dst)].morphs]
            if '名詞' in pre_pos and '動詞' in post_pos:
                print(pre_text, post_text, sep='\t')