{"id":366,"date":"2026-06-09T22:42:15","date_gmt":"2026-06-09T14:42:15","guid":{"rendered":"https:\/\/mitongxue.cn\/?p=366"},"modified":"2026-06-12T22:44:47","modified_gmt":"2026-06-12T14:44:47","slug":"%e8%af%8d%e8%a2%8b%e6%a8%a1%e5%9e%8b","status":"publish","type":"post","link":"https:\/\/mitongxue.cn\/index.php\/2026\/06\/09\/%e8%af%8d%e8%a2%8b%e6%a8%a1%e5%9e%8b\/","title":{"rendered":"\u8bcd\u888b\u6a21\u578b"},"content":{"rendered":"\n

\u8bcd\u888b\u6a21\u578b\uff08Bag-of-Words\uff09\u662f\u4e00\u79cd\u7b80\u5355\u7684\u6587\u672c\u8868\u793a\u65b9\u6cd5\uff0c\u4e5f\u662f\u81ea\u7136\u8bed\u8a00\u5904\u7406\u7684\u4e00\u4e2a \u7ecf\u5178\u6a21\u578b<\/strong>\u3002\u5b83\u5c06\u6587\u672c\u4e2d\u7684\u8bcd\u770b\u4f5c\u4e00\u4e2a\u4e2a\u72ec\u7acb\u7684\u4e2a\u4f53\uff0c\u4e0d\u8003\u8651\u5b83\u4eec\u5728\u53e5\u5b50\u4e2d\u7684\u987a\u5e8f\uff0c\u53ea\u5173\u5fc3\u6bcf\u4e2a\u8bcd\u51fa\u73b0\u7684\u9891\u6b21<\/strong>\u3002<\/p>\n\n\n\n

\u5de5\u4f5c\u539f\u7406<\/strong><\/p>\n\n\n\n

\u8bcd\u888b\u6a21\u578b\u7684\u6838\u5fc3\u601d\u60f3\u662f\u628a\uff1a\u6587\u672c\u8868\u793a\u6210\u4e00\u4e2a\u201c\u8bcd\u9891\u5411\u91cf\u201d\uff0c\u4e0d\u770b\u8bcd\u7684\u987a\u5e8f\u3001\u53e5\u6cd5\u7ed3\u6784\uff0c\u53ea\u5173\u6ce8\u8bcd\u662f\u5426\u51fa\u73b0<\/strong>\u548c\u51fa\u73b0\u7684\u6b21\u6570<\/strong>\u3002<\/p>\n\n\n\n

\u53ef\u4ee5\u770b\u4f5c\u662f\u4e00\u4e2a\u88c5\u8bcd\u7684\u53e3\u888b\uff0c\u91cc\u9762\u6709\u5927\u91cf\u7684\u8bcd\uff0c\u96f6\u96f6\u6563\u6563\u7684\u88c5\u5728\u91cc\u9762\u7684\u3002<\/p>\n\n\n\n

\u5047\u8bbe\u6709\u4e09\u53e5\u8bdd\u4f5c\u4e3a\u8bed\u6599\u5e93\uff08Corpus\uff09<\/p>\n\n\n\n

\u53e5\u5b501\uff1a\u6211\u559c\u6b22\u5403\u82f9\u679c  
\u53e5\u5b502\uff1a\u6211\u4e0d\u559c\u6b22\u5403\u9999\u8549  
\u53e5\u5b503\uff1a\u5979\u559c\u6b22\u5403\u8461\u8404<\/pre>\n\n\n\n
\u7b2c 1 \u6b65\u5206\u8bcd<\/strong>\uff0c\u4e5f\u5c31\u662f\u5c06\u6bcf\u4e2a\u53e5\u5b50\u5207\u5206\u6210\u8bcd\uff1a<\/p>\n\n\n\n
[\"\u6211\", \"\u559c\u6b22\", \"\u5403\", \"\u82f9\u679c\"]
[\"\u6211\", \"\u4e0d\", \"\u559c\u6b22\", \"\u5403\", \"\u9999\u8549\"]
[\"\u5979\", \"\u559c\u6b22\", \"\u5403\", \"\u8461\u8404\"]<\/pre>\n\n\n\n
\u7b2c 2 \u6b65\u6784\u5efa\u8bcd\u8868\uff0c\u63d0\u53d6\u6240\u6709\u51fa\u73b0\u8fc7\u7684\u552f\u4e00\u8bcd\u6c47<\/strong>\uff0c\u6309\u987a\u5e8f\u7f16\u53f7\uff1a<\/p>\n\n\n\n
\u53ef\u4ee5\u7406\u89e3\u4e3a\u9488\u5bf9\u4e0a\u9762\u7684\u591a\u4e2a\u8bcd\u8868\u8fdb\u884c\u53bb\u91cd\uff0c\u6784\u5efa\u4e00\u4e2a\u8bcd\u8868<\/p>\n\n\n\n
\u8bcd\u8868 = [\"\u6211\", \"\u559c\u6b22\", \"\u5403\", \"\u82f9\u679c\", \"\u4e0d\", \"\u9999\u8549\", \"\u5979\", \"\u8461\u8404\"]<\/pre>\n\n\n\n
\u8fd9\u4e2a\u8bcd\u8868\u786e\u5b9a\u4e86\u5411\u91cf\u7684\u201c\u7ef4\u5ea6\u201d\uff0c\u5171 8 \u4e2a\u8bcd\uff0c\u6240\u4ee5\u5411\u91cf\u957f\u5ea6\u5c31\u662f 8<\/strong>\u3002<\/p>\n\n\n\n
\u7b2c 3 \u6b65\u6784\u5efa\u8bcd\u9891\u5411\u91cf\uff0c\u4e5f\u5c31\u662f\u6bcf\u53e5\u8bdd\u5bf9\u5e94\u4e00\u4e2a\u5411\u91cf<\/p>\n\n\n\n
\u8bcd\u6c47\u4f4d\u7f6e<\/th>\u6211<\/th>\u559c\u6b22<\/th>\u5403<\/th>\u82f9\u679c<\/th>\u4e0d<\/th>\u9999\u8549<\/th>\u5979<\/th>\u8461\u8404<\/th><\/tr><\/thead>
\u53e5\u5b501 \u5411\u91cf<\/strong><\/td>1<\/td>1<\/td>1<\/td>1<\/td>0<\/td>0<\/td>0<\/td>0<\/td><\/tr>
\u53e5\u5b502 \u5411\u91cf<\/strong><\/td>1<\/td>1<\/td>1<\/td>0<\/td>1<\/td>1<\/td>0<\/td>0<\/td><\/tr>
\u53e5\u5b503 \u5411\u91cf<\/strong><\/td>0<\/td>1<\/td>1<\/td>0<\/td>0<\/td>0<\/td>1<\/td>1<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
\u8fd9\u4e2a\u8868\u5c31\u662f BoW \u7279\u5f81\u5411\u91cf\u5316\u540e\u7684\u7ed3\u679c\uff0c\u6bcf\u53e5\u8bdd\u53d8\u6210\u4e00\u4e2a\u201c\u5411\u91cf\u201d<\/strong>\uff0c\u53ef\u4ee5\u7528\u4e8e\u8ba1\u7b97\u673a\u5904\u7406\u3002<\/p>\n\n\n\n
\u8fd9\u4e00\u70b9\u975e\u5e38\u91cd\u8981\uff0c\u4e5f\u662f\u4e00\u4e2a\u91cc\u7a0b\u7891\u5f0f\u7684\u8fdb\u6b65\uff0c\u5c06\u53e5\u5b50\u8fdb\u884c\u5411\u91cf\u5316\u4e4b\u540e\uff0c\u610f\u5473\u7740\u53ef\u4ee5\u8fdb\u884c\u6570\u5b66\u8ba1\u7b97\u4e86\u3002<\/p>\n\n\n\n
BoW \u76f8\u5bf9\u4e8e N-Gram \u7684\u6539\u8fdb\uff1a<\/p>\n\n\n\n
\u95ee\u9898<\/th>N-Gram \u7684\u7f3a\u9677<\/th>BoW \u7684\u89e3\u51b3<\/th><\/tr><\/thead>
\u7ef4\u5ea6\u81a8\u80c0\u4e25\u91cd<\/strong><\/td>N-Gram \u7ec4\u5408\u6570\u975e\u5e38\u591a\uff08\u7279\u522b\u662f Trigram \u4ee5\u4e0a\uff09<\/td>BoW \u53ea\u7edf\u8ba1\u8bcd\uff0c\u4e0d\u8003\u8651\u7ec4\u5408\uff0c\u5411\u91cf\u7ef4\u5ea6\u66f4\u5c0f<\/strong><\/td><\/tr>
\u6570\u636e\u7a00\u758f\u4e25\u91cd<\/strong><\/td>N-Gram \u4e2d\u5f88\u591a\u8bcd\u7ec4\u7ec4\u5408\u6ca1\u51fa\u73b0<\/td>BoW \u7528\u8bcd\u9891\u7edf\u8ba1\uff0c\u4e0d\u4f1a\u56e0\u672a\u51fa\u73b0\u7684\u7ec4\u5408\u800c\u5f52 0<\/strong>\uff0c\u66f4\u7a33\u5b9a<\/td><\/tr>
\u4e0a\u4e0b\u6587\u5efa\u6a21\u590d\u6742<\/strong><\/td>N-Gram \u6709\u4e0a\u4e0b\u6587\uff0c\u4f46\u5efa\u6a21\u4ee3\u4ef7\u5927<\/td>BoW \u5b8c\u5168\u4e0d\u5efa\u4e0a\u4e0b\u6587\uff0c\u9002\u5408\u8ba1\u7b97\u6548\u7387\u4f18\u5148\u7684\u573a\u666f<\/strong><\/td><\/tr>
\u64c5\u957f\u7684\u9886\u57df<\/strong><\/td>N-Gram \u901a\u5e38\u7528\u4e8e\u8bed\u8a00\u5efa\u6a21\u3001\u751f\u6210<\/td>BoW \u66f4\u9002\u5408\u5206\u7c7b\u3001\u805a\u7c7b\u3001\u68c0\u7d22\u7b49\u4efb\u52a1<\/strong>\uff0c\u4e0d\u662f\u7528\u6765\u201c\u751f\u6210\u6587\u672c\u201d\u7684\u6a21\u578b<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
\u7531\u6b64\u53ef\u4ee5\u770b\u51fa\uff0c\u8bcd\u888b\u6a21\u578b\u66f4\u64c5\u957f\u7684\u4efb\u52a1\u662f\uff1a<\/p>\n\n\n\n
    \n
  • \u6587\u672c\u5206\u7c7b<\/li>\n\n\n\n
  • \u6587\u672c\u76f8\u4f3c\u5ea6\u5206\u6790<\/li>\n\n\n\n
  • \u4fe1\u606f\u68c0\u7d22 \/ \u6587\u6863\u6392\u5e8f<\/li>\n<\/ul>\n\n\n\n
    [1, 1, 1, 1, 0 , 0, 0, 0]
    \u200b
    [1, 1, 1, 0, 0 ,0, 0, 0]
    [0, 0, 0, 0, 0, 0, 0, 1]<\/code><\/pre>\n\n\n\n
    \u4ee3\u7801\u5b9e\u8df5<\/strong><\/p>\n\n\n\n
    # \u5bfc\u5165\u5fc5\u8981\u7684\u5e93
    from sklearn.feature_extraction.text import CountVectorizer  # \u7528\u4e8e\u6784\u5efa\u8bcd\u888b\u6a21\u578b
    import jieba  # \u4e2d\u6587\u5206\u8bcd\u5e93
    \u200b
    # 1. \u5b9a\u4e49\u539f\u59cb\u7684\u4e2d\u6587\u8bed\u6599\uff08\u6bcf\u4e2a\u5143\u7d20\u662f\u4e00\u53e5\u8bdd\uff09
    corpus = [
        \"\u6211\u559c\u6b22\u5403\u82f9\u679c\",
        \"\u6211\u4e0d\u559c\u6b22\u5403\u9999\u8549\",
        \"\u5979\u559c\u6b22\u5403\u8461\u8404\"
    ]
    \u200b
    # 2. \u5bf9\u6bcf\u4e2a\u53e5\u5b50\u8fdb\u884c jieba \u5206\u8bcd\uff0c\u5e76\u7528\u7a7a\u683c\u8fde\u63a5\uff08\u56e0\u4e3a CountVectorizer \u9ed8\u8ba4\u6309\u7a7a\u683c\u5206\u8bcd\uff09
    corpus_cut = [\" \".join(jieba.cut(sentence)) for sentence in corpus]
    \u200b
    # \u8f93\u51fa\u5206\u8bcd\u540e\u7684\u8bed\u6599\uff0c\u770b\u4e00\u4e0b jieba \u662f\u5982\u4f55\u5207\u8bcd\u7684
    print(\"\u3010\u5206\u8bcd\u7ed3\u679c\u3011\uff1a\")
    for i, sent in enumerate(corpus_cut):
        print(f\"\u53e5\u5b50{i + 1}\uff1a{sent}\")
    \u200b
    # 3. \u521b\u5efa CountVectorizer \u5b9e\u4f8b\uff08\u8bcd\u888b\u6a21\u578b\u5de5\u5177\uff09
    vectorizer = CountVectorizer()
    \u200b
    # 4. \u8bad\u7ec3\u8bcd\u888b\u6a21\u578b\uff0c\u5e76\u5c06\u6587\u672c\u8f6c\u6362\u4e3a\u8bcd\u9891\u5411\u91cf
    X = vectorizer.fit_transform(corpus_cut)
    \u200b
    # 5. \u8f93\u51fa\u8bcd\u8868\uff08\u5373 BoW \u6a21\u578b\u6240\u6784\u5efa\u7684\u6240\u6709\u8bcd\u6c47\uff09
    print(\"\\n\u3010\u8bcd\u8868\u3011\uff1a\")
    print(vectorizer.get_feature_names_out())
    \u200b
    # 6. \u8f93\u51fa\u8bcd\u9891\u5411\u91cf\uff08\u7a00\u758f\u77e9\u9635 \u2192 \u8f6c\u6362\u6210\u6570\u7ec4\u5f62\u5f0f\u5c55\u793a\uff09
    print(\"\\n\u3010\u5411\u91cf\u8868\u793a\u3011\uff1a\")
    print(X.toarray())<\/code><\/pre>\n\n\n\n
    \ud83e\udd14 \u5c06\u53e5\u5b50\u8f6c\u4e3a\u4e86\u5411\u91cf\u8868\u793a\uff0c\u6709\u4ec0\u4e48\u7528\u5462\uff1f<\/p>\n\n\n\n
    \u6839\u636e\u5411\u91cf\u5c31\u53ef\u4ee5\u53bb\u8bc4\u5224\u6587\u672c\u7684\u76f8\u4f3c\u5ea6\uff0c\u4ee5\u53ca\u5bf9\u6587\u672c\u8fdb\u884c\u5206\u7c7b\u3002<\/p>\n\n\n\n
    # 1. \u5bfc\u5165\u5fc5\u8981\u5e93
    from sklearn.feature_extraction.text import CountVectorizer  # \u7528\u4e8e\u6784\u5efa\u8bcd\u888b\u6a21\u578b
    from sklearn.naive_bayes import MultinomialNB  # \u6734\u7d20\u8d1d\u53f6\u65af\u5206\u7c7b\u5668
    from sklearn.pipeline import make_pipeline  # \u6784\u5efa\u8bad\u7ec3\u6d41\u7a0b
    import jieba  # \u4e2d\u6587\u5206\u8bcd\u5e93
    import numpy as np
    \u200b
    # 2. \u51c6\u5907\u8bed\u6599\u548c\u6807\u7b7e\uff081 \u8868\u793a\u6b63\u9762\uff0c0 \u8868\u793a\u8d1f\u9762\uff09
    texts = [
        \"\u8fd9\u5bb6\u5e97\u7684\u83dc\u592a\u96be\u5403\u4e86\",       # 0
        \"\u670d\u52a1\u6001\u5ea6\u5f88\u5dee\uff0c\u518d\u4e5f\u4e0d\u6765\u4e86\",  # 0
        \"\u83dc\u54c1\u5f88\u4e30\u5bcc\uff0c\u5473\u9053\u4e0d\u9519\",     # 1
        \"\u771f\u662f\u4e00\u6b21\u6109\u5feb\u7684\u7528\u9910\u4f53\u9a8c\",     # 1
        \"\u4ef7\u683c\u592a\u8d35\uff0c\u5403\u4e0d\u9971\",         # 0
        \"\u73af\u5883\u5f88\u597d\uff0c\u670d\u52a1\u4e5f\u5468\u5230\"       # 1
    ]
    labels = [0, 0, 1, 1, 0, 1]
    \u200b
    # 3. \u4f7f\u7528 jieba \u8fdb\u884c\u5206\u8bcd\uff0c\u5e76\u4ee5\u7a7a\u683c\u8fde\u63a5\u9002\u914d CountVectorizer
    texts_cut = [\" \".join(jieba.cut(text)) for text in texts]
    \u200b
    # \u8f93\u51fa\u5206\u8bcd\u7ed3\u679c
    print(\"\u3010\u5206\u8bcd\u7ed3\u679c\u3011\uff1a\")
    for i, text in enumerate(texts_cut):
        print(f\"\u53e5\u5b50{i+1}\uff1a{text}\")
    \u200b
    # 4. \u521d\u59cb\u5316 CountVectorizer\uff0c\u5e76\u62df\u5408\u8bed\u6599\u5e93
    vectorizer = CountVectorizer()
    X = vectorizer.fit_transform(texts_cut)
    \u200b
    # \u8f93\u51fa\u8bcd\u8868\uff08\u8bcd\u888b\uff09
    print(\"\\n\u3010\u8bcd\u8868\uff08Vocabulary\uff09\u3011\uff1a\")
    print(vectorizer.get_feature_names_out())
    \u200b
    # \u8f93\u51fa\u5411\u91cf\u8868\u793a\uff08\u8bcd\u9891\u77e9\u9635\uff09
    print(\"\\n\u3010\u8bcd\u9891\u5411\u91cf\u8868\u793a\u3011\uff1a\")
    print(X.toarray())
    \u200b
    # 5. \u4f7f\u7528\u6734\u7d20\u8d1d\u53f6\u65af\u5206\u7c7b\u5668\u8fdb\u884c\u8bad\u7ec3
    model = MultinomialNB()
    model.fit(X, labels)
    \u200b
    # 6. \u6d4b\u8bd5\u65b0\u6837\u672c
    test_comments = [
        \"\u670d\u52a1\u4e0d\u597d\uff0c\u4e1c\u897f\u4e5f\u96be\u5403\",   # 0
        \"\u996d\u5f88\u597d\u5403\uff0c\u73af\u5883\u4e5f\u4e0d\u9519\",   # 1
        \"\u592a\u8d35\u4e86\uff0c\u4e0d\u63a8\u8350\",       # 0
        \"\u4e0b\u6b21\u8fd8\u4f1a\u6765\uff0c\u771f\u7684\u5f88\u6ee1\u610f\"  # 1
    ]
    test_cut = [\" \".join(jieba.cut(comment)) for comment in test_comments]
    \u200b
    # \u8f93\u51fa\u6d4b\u8bd5\u5206\u8bcd
    print(\"\\n\u3010\u6d4b\u8bd5\u96c6\u5206\u8bcd\u3011\uff1a\")
    for i, sent in enumerate(test_cut):
        print(f\"\u6d4b\u8bd5{i+1}\uff1a{sent}\")
    \u200b
    # \u5c06\u6d4b\u8bd5\u8bed\u53e5\u8f6c\u6362\u4e3a\u8bcd\u888b\u5411\u91cf
    X_test = vectorizer.transform(test_cut)
    \u200b
    # \u8f93\u51fa\u6d4b\u8bd5\u5411\u91cf
    print(\"\\n\u3010\u6d4b\u8bd5\u96c6\u5411\u91cf\u3011\uff1a\")
    print(X_test.toarray())
    \u200b
    # 7. \u6267\u884c\u9884\u6d4b
    predictions = model.predict(X_test)
    \u200b
    # \u8f93\u51fa\u6700\u7ec8\u9884\u6d4b\u7ed3\u679c
    print(\"\\n\u3010\u60c5\u611f\u9884\u6d4b\u7ed3\u679c\u3011\uff1a\")
    for comment, label in zip(test_comments, predictions):
        print(f\"'{comment}' \u2192 {'\u6b63\u9762' if label == 1 else '\u8d1f\u9762'}\")<\/code><\/pre>\n\n\n\n
    \u8bcd\u888b\u6a21\u578b\u7684\u7f3a\u9677<\/strong><\/p>\n\n\n\n
      \n
    1. \u53ea\u770b\u8bcd\u6709\u6ca1\u6709\uff0c\u4e0d\u770b\u4e0a\u4e0b\u6587<\/li>\n\n\n\n
    2. \u8bcd\u5e8f\u88ab\u5ffd\u7565\uff1a\u5728\u8bcd\u888b\u6a21\u578b\u91cc\u9762\uff0c\u201c\u6211\u6253\u4e86\u4ed6\u201d\u548c\u201c\u4ed6\u6253\u4e86\u6211\u201d\uff0c\u4e24\u53e5\u8bdd\u4f1a\u88ab\u770b\u4f5c\u662f\u4e00\u4e2a\u610f\u601d\u3002<\/li>\n\n\n\n
    3. \u6ca1\u8bed\u4e49\u6cdb\u5316\u80fd\u529b\uff1a\u6a21\u578b\u4e0d\u77e5\u9053\u201c\u6ee1\u610f\u201d\u548c\u201c\u4e0d\u9519\u201d\u610f\u601d\u76f8\u8fd1\uff0c\u201c\u6109\u5feb\u201d\u548c\u201c\u5f00\u5fc3\u201d\u662f\u540c\u4e49\u8bcd\u3002<\/li>\n<\/ol>\n","protected":false},"excerpt":{"rendered":"
      \u8bcd\u888b\u6a21\u578b\uff08Bag-of-Words\uff09\u662f\u4e00\u79cd\u7b80\u5355\u7684\u6587..<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[27],"tags":[],"class_list":["post-366","post","type-post","status-publish","format-standard","hentry","category-ai"],"featured_image_urls":{"full":"","thumbnail":"","medium":"","medium_large":"","large":"","1536x1536":"","2048x2048":""},"author_info":{"info":["\u9648 \u67d0\u4eba"]},"category_info":"AI<\/a>","tag_info":"AI","comment_count":"0","_links":{"self":[{"href":"https:\/\/mitongxue.cn\/index.php\/wp-json\/wp\/v2\/posts\/366","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/mitongxue.cn\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/mitongxue.cn\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/mitongxue.cn\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/mitongxue.cn\/index.php\/wp-json\/wp\/v2\/comments?post=366"}],"version-history":[{"count":1,"href":"https:\/\/mitongxue.cn\/index.php\/wp-json\/wp\/v2\/posts\/366\/revisions"}],"predecessor-version":[{"id":367,"href":"https:\/\/mitongxue.cn\/index.php\/wp-json\/wp\/v2\/posts\/366\/revisions\/367"}],"wp:attachment":[{"href":"https:\/\/mitongxue.cn\/index.php\/wp-json\/wp\/v2\/media?parent=366"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/mitongxue.cn\/index.php\/wp-json\/wp\/v2\/categories?post=366"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/mitongxue.cn\/index.php\/wp-json\/wp\/v2\/tags?post=366"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}