python如何提取文本信息？

def ie_preprocess(document):

    # nltk 默认的句子分段器

    sentences = nltk.sent_tokenize(document)

    # nltk默认分词器

    sentences = [nltk.word_tokenize(sent) for sent in sentences]

    # nltk默认词性标记

    sentences = [nltk.pos_tag(sent) for sent in sentences]

# 使用一元标注器建立一个词块划分器。根据每个词的词性标记，尝试确定正确的词块标记。

class UnigramChunker(nltk.ChunkParserI):

    # constructor

    def __init__(self, train_sents):

        # 将训练数据转换成适合训练标注器的形式。tree2conlltags()方法将每个词块树映射到一个三元组（word，tag，chunk）的列表

        train_data = [[(t, c) for w, t, c in nltk.chunk.tree2conlltags(sent)]

                      for sent in train_sents]

        # 训练一元分块器

        # self.tagger = nltk.UnigramTagger(train_data)

        # 训练二元分块器

        self.tagger = nltk.BigramTagger(train_data)

    # sentence为一个已标注的句子

    def parse(self, sentence):

        # 提取词性标记

        pos_tags = [pos for (word, pos) in sentence]

        # 使用标注器为词性标记 标注IOB词块

        tagged_pos_tags = self.tagger.tag(pos_tags)

        # 提取词块标记

        chunktags = [chunktag for (pos, chunktag) in tagged_pos_tags]

        # 将词块标记与原句组合

        conlltags = [(word, pos, chunktag) for ((word, pos), chunktag)

                     in zip(sentence, chunktags)]

        # 转换成词块树

        return nltk.chunk.conlltags2tree(conlltags)