app.py

import streamlit as st

from news_scraper import YahooNewsScraper
from tidif_calclator import JapaneseTextVectorizer
from cosine_similarity_calculator import CosineSimilarityCalculator
from summerizer import TextSummarizer

st.title("ニュース検索アプリ")

# 初期化
best_article_text = None
best_article_url = None
best_max_word = None
max_word = None
best_max_value = -1  # cos類似度は0以上なので、初期値を-1に設定
num_news = 5

# セッションステートの初期化
if 'news_fetched' not in st.session_state:
    st.session_state['news_fetched'] = False
    st.session_state['article_text_list'] = []
    st.session_state['article_url_list'] = []

if st.button('最新ニュース取得'):
    with st.spinner('ニュースを取得中...'):
        # yahooニュースをスクレイピング
        scraper = YahooNewsScraper()
        article_text_list = []
        article_url_list = []
        for i in range(num_news):
            article_text, detail_url = scraper.scrape_article(i)
            article_text_list.append(article_text)
            article_url_list.append(detail_url)
        st.session_state['news_fetched'] = True  # 処理完了フラグを設定
        st.session_state['article_text_list'] = article_text_list  # セッションステートに保存
        st.session_state['article_url_list'] = article_url_list
    st.write("取得完了しました")

if st.session_state['news_fetched']:
    search_word = st.text_input('名詞', placeholder='名詞を入力してください', max_chars=10, help='10文字以内の名詞')
    if st.button('要約作成'):
        article_text_list = st.session_state['article_text_list']
        article_url_list = st.session_state['article_url_list']
        for temp_article_text, temp_article_url in zip(article_text_list, article_url_list):
            # TD-IDF値を計算
            vectorizer = JapaneseTextVectorizer()
            tfidf_dict = vectorizer.fit_transform(temp_article_text)

            # cos類似度を計算
            word_similarity = CosineSimilarityCalculator()
            article_keyword_list = list(tfidf_dict.keys())
            result_word_similarity = word_similarity.calculate_similarity(search_word, article_keyword_list)

            # None でない値のみを抽出
            filtered_data = {k: v for k, v in result_word_similarity.items() if v is not None}

            # 最大値を持つキーとその値を取得
            if filtered_data:  # filtered_dataが空でないことを確認
                max_word = max(filtered_data, key=filtered_data.get)
                max_value = filtered_data[max_word]
                # 最大値がこれまでの最大値より大きければ更新
                if max_value > best_max_value:
                    best_max_value = max_value
                    best_max_word = max_word
                    best_article_text = temp_article_text
                    best_article_url = temp_article_url

        # テキストを要約
        summarizer = TextSummarizer()
        summary_text = summarizer.summarize(best_article_text, max_length=30, min_length=20)
        st.write(f'最も類似度が高いワードは「{best_max_word}」でした')
        st.write(f'url：{best_article_url}')
        st.text_area("要約:", summary_text, height=20)