2021-05-01

def print_topic(texts, n_topics, n_words):
    '''
    输出主题模型结果
    :param n_topics: LDA主题数量
    :param n_words: 主题的词范围（数量）
    :return: 输出公共主题，单日主题
    '''

    # 准备语料库/Vocab
    vocabs = corpora.Dictionary(texts)  # generate vocabulary dictionary
    corpus = [vocabs.doc2bow(word) for word in texts]  #为训练模型生成语料库
    corpora.MmCorpus.serialize('weibo.mm', corpus)

    # 训练LDA模型
    # 设定主题个数：
    lda = models.ldamodel.LdaModel(corpus=corpus, id2word=vocabs, num_topics=n_topics, random_state=1)
    # 展示训练结果
    # lda.print_topics(num_topics=n_topics, num_words=n_words)

    corpus_lda = lda[corpus]
    main_topics = []
    for i in range(len(texts)):
        # print(corpus_lda[i])
        # 选取单日"权重最大"的主题
        main_topics.append(sorted(corpus_lda[i], key=lambda x: x[1], reverse=True)[0])

    # 梳理主题结果：排除在每个主题列表中都出现的词；公共主题，单日主题
    # 整理单日的主题
    for t in range(0, n_topics):
        name = 't' + str(t)
        locals()['t' + str(t)] = []
        for i in range(0, n_words):
            locals()['t' + str(t)].append((lda.show_topic(t, n_words)[i])[0])

    # 提取公共主题
    p_topic = locals()['t' + str(0)]
    for t in range(1, n_topics):
        p_topic = list(set(p_topic).intersection(set(locals()['t' + str(t)])))
    print("public topic:", p_topic)#打印公共主题

    # 提取单日"独有的"主题
    day_topics = []
    for t in range(0, n_topics):
        locals()["t" + str(t)] = list(set(locals()["t" + str(t)]).difference(set(p_topic)))
        day_topics.append(locals()["t" + str(t)])
        print(t, ":", locals()["t" + str(t)])#打印每日的主题

    # 返回变量：  每一天权重最高的topic的数据
    days_topic = []
    for i in range(len(texts)):
        main_topic = main_topics[i]
        days_topic.append(day_topics[main_topic[0]])

    return main_topics, p_topic, days_topic