{"id":9106,"date":"2024-01-18T22:25:58","date_gmt":"2024-01-18T14:25:58","guid":{"rendered":"\/?p=9106"},"modified":"2024-01-18T22:25:58","modified_gmt":"2024-01-18T14:25:58","slug":"spacy%e5%9f%ba%e7%a1%80%e6%95%99%e7%a8%8b","status":"publish","type":"post","link":"\/?p=9106","title":{"rendered":"spacy\u57fa\u7840\u6559\u7a0b"},"content":{"rendered":"<h2>1 \u80cc\u666f\u4ecb\u7ecd\u4e0espaCy\u5b89\u88c5<\/h2>\n<p>spaCy\u662f\u4e00\u4e2a\u57fa\u4e8ePython\u7f16\u5199\u7684\u5f00\u6e90<a href=\"https:\/\/so.csdn.net\/so\/search?q=\u81ea\u7136\u8bed\u8a00\u5904\u7406&amp;spm=1001.2101.3001.7020\">\u81ea\u7136\u8bed\u8a00\u5904\u7406<\/a>\u5e93\u3002\u57fa\u4e8e\u81ea\u7136\u5904\u7406\u9886\u57df\u7684\u6700\u65b0\u7814\u7a76\uff0cspaCy\u63d0\u4f9b\u4e86\u4e00\u7cfb\u5217\u9ad8\u6548\u4e14\u6613\u7528\u7684\u5de5\u5177\uff0c\u7528\u4e8e\u6587\u672c\u9884\u5904\u7406\u3001\u6587\u672c\u89e3\u6790\u3001\u547d\u540d\u5b9e\u4f53\u8bc6\u522b\u3001\u8bcd\u6027\u6807\u6ce8\u3001\u53e5\u6cd5\u5206\u6790\u548c\u6587\u672c\u5206\u7c7b\u7b49\u4efb\u52a1\u3002<br \/>\nspaCy\u7684\u5b98\u65b9\u4ed3\u5e93\u5730\u5740\u4e3a\uff1a<a href=\"https:\/\/github.com\/explosion\/spaCy\">spaCy-github<\/a>\u3002\u672c\u6587\u4e3b\u8981\u53c2\u8003\u5176\u5b98\u65b9\u7f51\u7ad9\u7684\u6587\u6863\uff0cspaCy\u7684\u5b98\u65b9\u7f51\u7ad9\u5730\u5740\u4e3a\uff1a<a href=\"https:\/\/spacy.io\/\">spaCy<\/a>\u3002<\/p>\n<h3>1.1 \u81ea\u7136\u8bed\u8a00\u5904\u7406\u7b80\u4ecb<\/h3>\n<p>\u81ea\u7136\u8bed\u8a00\u5904\u7406\uff08Natural Language Processing\uff0c\u7b80\u79f0NLP\uff09\u662f\u4e00\u95e8\u7814\u7a76\u4eba\u7c7b\u8bed\u8a00\u4e0e\u8ba1\u7b97\u673a\u4e4b\u95f4\u4ea4\u4e92\u7684\u9886\u57df\uff0c\u65e8\u5728\u4f7f\u8ba1\u7b97\u673a\u80fd\u591f\u7406\u89e3\u3001\u89e3\u6790\u3001\u751f\u6210\u548c\u5904\u7406\u4eba\u7c7b\u8bed\u8a00\u3002NLP\u7ed3\u5408\u4e86\u8ba1\u7b97\u673a\u79d1\u5b66\u3001\u4eba\u5de5\u667a\u80fd\u548c\u8bed\u8a00\u5b66\u7684\u77e5\u8bc6\uff0c\u901a\u8fc7\u5404\u79cd\u7b97\u6cd5\u548c\u6280\u672f\u6765\u5904\u7406\u548c\u5206\u6790\u6587\u672c\u6570\u636e\u3002\u8fd1\u5e74\u6765\uff0c\u968f\u7740\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u7684\u53d1\u5c55\uff0c\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\u5728\u81ea\u7136\u8bed\u8a00\u5904\u7406\uff08NLP\uff09\u9886\u57df\u53d6\u5f97\u4e86\u91cd\u5927\u7684\u7a81\u7834\u3002\u5176\u4e2d\uff0c\u5faa\u73af\u795e\u7ecf\u7f51\u7edc\uff08RNN\uff09\u3001\u957f\u77ed\u65f6\u8bb0\u5fc6\u7f51\u7edc\uff08LSTM\uff09\u548cTransformer\u7b49\u6a21\u578b\u90fd\u53d1\u6325\u4e86\u5173\u952e\u4f5c\u7528\u3002\u8fd9\u4e9b\u6a21\u578b\u4e3aNLP\u4efb\u52a1\u5e26\u6765\u4e86\u66f4\u597d\u7684\u6027\u80fd\u548c\u6548\u679c\uff0c\u63a8\u52a8\u4e86NLP\u7684\u53d1\u5c55\u548c\u5e94\u7528\u3002<\/p>\n<p><img src=\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/68fb0152042e9029cedb92a4e24a3510-1705586496152-21.png\" alt=\"img\" \/><\/p>\n<p>NLP\u7684\u5e94\u7528\u975e\u5e38\u5e7f\u6cdb\uff0c\u6db5\u76d6\u4e86\u591a\u4e2a\u9886\u57df\uff0c\u5982\u673a\u5668\u7ffb\u8bd1\u3001\u4fe1\u606f\u63d0\u53d6\u3001\u6587\u672c\u5206\u7c7b\u3001\u60c5\u611f\u5206\u6790\u3001\u81ea\u52a8\u6458\u8981\u3001\u95ee\u7b54\u7cfb\u7edf\u3001\u8bed\u97f3\u8bc6\u522b\u548c\u8bed\u97f3\u5408\u6210\u7b49\u3002\u4ee5\u4e0b\u662fNLP\u5e38\u7528\u7684\u6280\u672f\u548c\u65b9\u6cd5\uff1a<\/p>\n<ul>\n<li>\u5206\u8bcd\uff1a\u5c06\u8fde\u7eed\u7684\u6587\u672c\u5206\u5272\u6210\u6709\u610f\u4e49\u7684\u8bcd\u8bed\u6216\u6807\u8bb0\uff0c\u662f\u8bb8\u591aNLP\u4efb\u52a1\u7684\u57fa\u7840\u3002<\/li>\n<li>\u8bcd\u6027\u6807\u6ce8\uff1a\u4e3a\u6587\u672c\u4e2d\u7684\u6bcf\u4e2a\u8bcd\u8bed\u8d4b\u4e88\u5176\u76f8\u5e94\u7684\u8bcd\u6027\uff0c\u5982\u540d\u8bcd\u3001\u52a8\u8bcd\u3001\u5f62\u5bb9\u8bcd\u7b49\u3002<\/li>\n<li>\u53e5\u6cd5\u5206\u6790\uff1a\u5206\u6790\u53e5\u5b50\u7684\u8bed\u6cd5\u7ed3\u6784\uff0c\u8bc6\u522b\u51fa\u53e5\u5b50\u4e2d\u7684\u77ed\u8bed\u3001\u4fee\u9970\u8bed\u548c\u4f9d\u5b58\u5173\u7cfb\u7b49\u3002<\/li>\n<li>\u8bed\u4e49\u5206\u6790\uff1a\u7406\u89e3\u6587\u672c\u7684\u610f\u4e49\u548c\u8bed\u4e49\u5173\u7cfb\uff0c\u5305\u62ec\u547d\u540d\u5b9e\u4f53\u8bc6\u522b\u3001\u8bed\u4e49\u89d2\u8272\u6807\u6ce8\u548c\u8bed\u4e49\u89e3\u6790\u7b49\u3002<\/li>\n<li>\u673a\u5668\u7ffb\u8bd1\uff1a\u5c06\u4e00\u79cd\u8bed\u8a00\u7684\u6587\u672c\u81ea\u52a8\u7ffb\u8bd1\u6210\u53e6\u4e00\u79cd\u8bed\u8a00\u3002<\/li>\n<li>\u6587\u672c\u5206\u7c7b\uff1a\u5c06\u6587\u672c\u6309\u7167\u9884\u5b9a\u4e49\u7684\u7c7b\u522b\u8fdb\u884c\u5206\u7c7b\uff0c\u5982\u5783\u573e\u90ae\u4ef6\u5206\u7c7b\u3001\u60c5\u611f\u5206\u7c7b\u7b49\u3002<\/li>\n<li>\u4fe1\u606f\u63d0\u53d6\uff1a\u4ece\u7ed3\u6784\u5316\u548c\u975e\u7ed3\u6784\u5316\u6587\u672c\u4e2d\u63d0\u53d6\u51fa\u7279\u5b9a\u7684\u4fe1\u606f\uff0c\u5982\u5b9e\u4f53\u5173\u7cfb\u62bd\u53d6\u3001\u4e8b\u4ef6\u62bd\u53d6\u7b49\u3002<\/li>\n<li>\u95ee\u7b54\u7cfb\u7edf\uff1a\u901a\u8fc7\u5bf9\u7528\u6237\u63d0\u51fa\u7684\u95ee\u9898\u8fdb\u884c\u7406\u89e3\u548c\u56de\u7b54\uff0c\u63d0\u4f9b\u51c6\u786e\u7684\u7b54\u6848\u6216\u76f8\u5173\u4fe1\u606f\u3002<\/li>\n<li>\u60c5\u611f\u5206\u6790\uff1a\u8bc6\u522b\u548c\u5206\u6790\u6587\u672c\u4e2d\u7684\u60c5\u611f\u503e\u5411\uff0c\u5982\u6b63\u9762\u3001\u8d1f\u9762\u6216\u4e2d\u6027\u60c5\u611f\u3002<\/li>\n<li>\u6587\u672c\u751f\u6210\uff1a\u4f7f\u7528NLP\u6280\u672f\u751f\u6210\u81ea\u7136\u8bed\u8a00\u6587\u672c\uff0c\u5982\u81ea\u52a8\u6458\u8981\u751f\u6210\u3001\u5bf9\u8bdd\u7cfb\u7edf\u548c\u673a\u5668\u4f5c\u6587\u7b49\u3002<\/li>\n<\/ul>\n<p>\u5728\u4f17\u591a\u81ea\u7136\u8bed\u8a00\u5904\u7406\u5e93\u4e2d\uff0cspaCy\u5e93\u63d0\u4f9b\u4e86\u8d85\u8fc773\u79cd\u8bed\u8a00\u7684\u652f\u6301\uff0c\u5e76\u4e3a25\u79cd\u8bed\u8a00\u63d0\u4f9b\u4e86\u8bad\u7ec3\u4ee3\u7801\u3002\u8be5\u5e93\u63d0\u4f9b\u4e86\u4e00\u7cfb\u5217\u7b80\u5355\u6613\u7528\u7684\u6a21\u578b\u548c\u51fd\u6570\u63a5\u53e3\uff0c\u5305\u62ec\u5206\u8bcd\u3001\u8bcd\u6027\u6807\u6ce8\u7b49\u529f\u80fd\u3002\u7528\u6237\u8fd8\u53ef\u4ee5\u4f7f\u7528PyTorch\u3001TensorFlow\u7b49\u6846\u67b6\u5728spaCy\u521b\u5efa\u81ea\u5b9a\u4e49\u6a21\u578b\uff0c\u4ee5\u6ee1\u8db3\u7279\u5b9a\u9700\u6c42\u3002spaCy\u652f\u6301\u7684\u8bed\u8a00\u6a21\u578b\u89c1<a href=\"https:\/\/spacy.io\/usage\/models\">spaCy-models<\/a>\u3002<\/p>\n<p>\u4e8b\u5b9e\u4e0a\uff0c\u6709\u4e00\u4e9b\u81ea\u7136\u8bed\u8a00\u5904\u7406\u5f00\u6e90\u5e93\uff0c\u4f8b\u5982HuggingFace\u7684<a href=\"https:\/\/github.com\/huggingface\/transformers\">Transformers<\/a>\u3001<a href=\"https:\/\/github.com\/PaddlePaddle\/PaddleNLP\">PaddleNLP<\/a>\u548c<a href=\"https:\/\/github.com\/nltk\/nltk\">NLTK<\/a>\uff0c\u76f8\u8f83\u4e8espaCy\u6765\u8bf4\u66f4\u4e3a\u4e13\u4e1a\u4e14\u6027\u80fd\u66f4\u597d\u3002\u7136\u800c\uff0c\u5bf9\u4e8e\u7b80\u5355\u7684\u5e94\u7528\u800c\u8a00\uff0cspaCy\u66f4\u4e3a\u9002\u5408\uff0c\u56e0\u4e3a\u5b83\u5177\u6709\u7b80\u5355\u6613\u7528\u3001\u529f\u80fd\u5168\u9762\uff0c\u540c\u65f6\u4e5f\u63d0\u4f9b\u4e86\u5927\u91cf\u9762\u5411\u591a\u8bed\u8a00\u9884\u8bad\u7ec3\u6a21\u578b\u7684\u4f18\u70b9\u3002\u6b64\u5916\uff0c\u968f\u7740\u4ee5GPT-3\u4e3a\u4ee3\u8868\u7684\u8bed\u8a00\u5927\u6a21\u578b\u5728\u81ea\u7136\u8bed\u8a00\u5904\u7406\u9886\u57df\u53d6\u5f97\u4e86\u5de8\u5927\u7684\u7a81\u7834\u548c\u6210\u529f\uff0c\u539f\u672c\u4e00\u4e9b\u81ea\u7136\u8bed\u8a00\u5904\u7406\u5e93\u5728\u7cbe\u5ea6\u4e0a\u5b9e\u9645\u4e0d\u5982\u8bed\u8a00\u5927\u6a21\u578b\u3002\u7136\u800c\uff0c\u4f7f\u7528\u8bed\u8a00\u5927\u6a21\u578b\u9700\u8981\u5e9e\u5927\u7684\u63a8\u7406\u8d44\u6e90\uff0c\u800c\u5728\u5bf9\u7cbe\u5ea6\u8981\u6c42\u4e0d\u9ad8\u7684\u573a\u666f\u4e2d\uff0c\u4f7f\u7528spaCy\u8fd9\u7c7b\u5c0f\u5de7\u7684\u81ea\u7136\u8bed\u8a00\u5904\u7406\u5e93\u4f9d\u7136\u662f\u5f88\u5408\u9002\u7684\u9009\u62e9\u3002<\/p>\n<h3>1.2 spaCy\u5b89\u88c5<\/h3>\n<p>spaCy\u91c7\u7528\u91c7\u7528\u6a21\u5757\u548c\u8bed\u8a00\u6a21\u5757\u4e00\u8d77\u5b89\u88c5\u7684\u6a21\u5f0f\u3002spaCy\u7684\u8bbe\u8ba1\u76ee\u6807\u4e4b\u4e00\u662f\u6a21\u5757\u5316\u548c\u53ef\u5b9a\u5236\u6027\u3002\u5b83\u5141\u8bb8\u7528\u6237\u4ec5\u5b89\u88c5\u5fc5\u9700\u7684\u6a21\u5757\u548c\u8bed\u8a00\u6570\u636e\uff0c\u4ee5\u51cf\u5c11\u5b89\u88c5\u7684\u6574\u4f53\u5927\u5c0f\u548c\u51cf\u8f7b\u8d44\u6e90\u8d1f\u62c5\u3002\u5982\u679c\u4f7f\u7528spaCy\u7684\u6a21\u578b\uff0c\u9700\u8981\u901a\u8fc7pip\u5b89\u88c5\u6a21\u578b\u6240\u9700\u7684\u6a21\u578b\u5305\u6765\u4f7f\u7528\u9884\u8bad\u7ec3\u6a21\u578b\u3002\u8fd9\u662f\u56e0\u4e3aspaCy\u7684\u6a21\u578b\u5305\u542b\u4e86\u8bad\u7ec3\u540e\u7684\u6743\u91cd\u53c2\u6570\u548c\u5176\u4ed6\u5fc5\u8981\u7684\u6587\u4ef6\uff0c\u8fd9\u4e9b\u6587\u4ef6\u5728\u5b89\u88c5\u65f6\u88ab\u5b58\u50a8\u5728\u7279\u5b9a\u4f4d\u7f6e\uff0c\u800c\u4e0d\u662f\u4ee5\u5355\u4e2a\u6587\u4ef6\u7684\u5f62\u5f0f\u5b58\u5728\u3002\u5982\u679c\u9700\u8981\u8fdb\u884c\u6a21\u578b\u8bad\u7ec3\u548cgpu\u8fd0\u884c\u5219\u9700\u8981\u9009\u5b9a\u5bf9\u5e94\u7684\u5b89\u88c5\u5305\u3002\u5c06\u6a21\u5757\u548c\u8bed\u8a00\u6a21\u5757\u4e00\u8d77\u5b89\u88c5\uff0c\u53ef\u4ee5\u7b80\u5316spaCy\u7684\u914d\u7f6e\u8fc7\u7a0b\u3002\u7528\u6237\u65e0\u9700\u5355\u72ec\u4e0b\u8f7d\u548c\u914d\u7f6e\u8bed\u8a00\u6570\u636e\uff0c\u4e5f\u4e0d\u9700\u8981\u624b\u52a8\u6307\u5b9a\u8981\u4f7f\u7528\u7684\u8bed\u8a00\u6a21\u578b\u3002\u8fd9\u6837\u53ef\u4ee5\u51cf\u5c11\u7528\u6237\u7684\u5de5\u4f5c\u91cf\u548c\u5b89\u88c5\u8fc7\u7a0b\u4e2d\u7684\u6f5c\u5728\u9519\u8bef\u3002\u4f46\u662f\u53ef\u5b9a\u5236\u6027\u5c31\u5f88\u5f31\u4e86\uff0c\u6240\u4ee5spaCy\u9002\u5408\u7cbe\u5ea6\u8981\u6c42\u4e0d\u9ad8\u7684\u7b80\u5355\u4f7f\u7528\uff0c\u5de5\u7a0b\u5e94\u7528\u9009\u62e9\u5176\u4ed6\u5927\u578b\u81ea\u7136\u8bed\u8a00\u5904\u7406\u5e93\u66f4\u52a0\u5408\u9002\u3002<\/p>\n<p>\u4e3a\u4e86\u5b9e\u73b0\u8fd9\u4e00\u76ee\u6807\uff0cspaCy\u63d0\u4f9b\u4e86\u914d\u7f6e\u5316\u7684\u5b89\u88c5\u6307\u4ee4\u9009\u62e9\u9875\u9762\u4f9b\u7528\u6237\u4f7f\u7528\u3002\u5b89\u88c5\u6307\u4ee4\u9009\u62e9\u9875\u9762\u5730\u5740\u4e3a<a href=\"https:\/\/spacy.io\/usage\">spaCy-usage<\/a>\u3002\u4e0b\u56fe\u5c55\u793a\u4e86\u672c\u6587\u7684\u5b89\u88c5\u914d\u7f6e\u9879\uff0c\u672c\u6587\u91c7\u7528\u4e86\u6700\u7b80\u5355\u7684cpu\u63a8\u7406\u6a21\u5f0f\u3002<\/p>\n<p><img src=\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/c8280322bf3f8d9473f7664ab2ad2f59.png\" alt=\"img\" \/><\/p>\n<p>spaCy\u5b89\u88c5\u5b8c\u6bd5\u540e\uff0c\u8fd0\u884c\u4ee5\u4e0b\u4ee3\u7801\u5373\u53ef\u5224\u65adspaCy\u53ca\u76f8\u5bf9\u5e94\u7684\u8bed\u8a00\u6a21\u578b\u662f\u5426\u5b89\u88c5\u6210\u529f\u3002<\/p>\n<pre><code class=\"language-python\"># jupyter notebook\u73af\u5883\u53bb\u9664warning\nimport warnings\nwarnings.filterwarnings(&quot;ignore&quot;)\nimport spacy\nspacy.__version__\n\n# =&gt;&#039;3.6.0&#039;\nimport spacy\n# \u52a0\u8f7d\u5df2\u5b89\u88c5\u7684\u4e2d\u6587\u6a21\u578b\nnlp = spacy.load(&#039;zh_core_web_sm&#039;)\n# \u6267\u884c\u4e00\u4e9b\u7b80\u5355\u7684NLP\u4efb\u52a1\ndoc = nlp(&quot;\u65e9\u4e0a\u597d!&quot;)\nfor token in doc:\n    # token.text\u8868\u793a\u6807\u8bb0\u7684\u539f\u59cb\u6587\u672c\uff0ctoken.pos_\u8868\u793a\u6807\u8bb0\u7684\u8bcd\u6027\uff08part-of-speech\uff09\uff0ctoken.dep_\u8868\u793a\u6807\u8bb0\u4e0e\u5176\u4ed6\u6807\u8bb0\u4e4b\u95f4\u7684\u53e5\u6cd5\u4f9d\u5b58\u5173\u7cfb\n    print(token.text, token.pos_, token.dep_)\n12345678910\n&quot;&quot;&quot;\n\u65e9\u4e0a NOUN nmod:tmod\n\u597d VERB ROOT\n! PUNCT punct\n&quot;&quot;&quot;<\/code><\/pre>\n<h2>2 spaCy\u5feb\u901f\u5165\u95e8<\/h2>\n<p>\u8be5\u90e8\u5206\u5185\u5bb9\u548c\u56fe\u7247\u4e3b\u8981\u6765\u81ea\u4e8e<a href=\"https:\/\/spacy.io\/usage\/spacy-101\">spacy-101<\/a>\u7684\u603b\u7ed3\u3002spaCy\u63d0\u4f9b\u7684\u4e3b\u8981\u51fd\u6570\u6a21\u5757\u5206\u4e3a\u4ee5\u4e0b\u6a21\u5757\uff0c\u63a5\u4e0b\u6765\u5206\u522b\u5bf9\u8fd9\u4e9b\u6a21\u5757\u8fdb\u884c\u4ecb\u7ecd\u3002<\/p>\n<table>\n<thead>\n<tr>\n<th>\u540d\u79f0<\/th>\n<th>\u63cf\u8ff0<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Tokenization<\/td>\n<td>\u5c06\u6587\u672c\u5206\u5272\u6210\u5355\u8bcd\u3001\u6807\u70b9\u7b26\u53f7\u7b49\u3002<\/td>\n<\/tr>\n<tr>\n<td>Part-of-speech (POS) Tagging<\/td>\n<td>\u7ed9\u6807\u8bb0\u5206\u914d\u8bcd\u6027\uff0c\u5982\u52a8\u8bcd\u6216\u540d\u8bcd\u3002<\/td>\n<\/tr>\n<tr>\n<td>Dependency Parsing<\/td>\n<td>\u5206\u914d\u53e5\u6cd5\u4f9d\u5b58\u6807\u7b7e\uff0c\u63cf\u8ff0\u4e2a\u522b\u6807\u8bb0\u4e4b\u95f4\u7684\u5173\u7cfb\uff0c\u5982\u4e3b\u8bed\u6216\u5bbe\u8bed\u3002<\/td>\n<\/tr>\n<tr>\n<td>Lemmatization<\/td>\n<td>\u5206\u914d\u5355\u8bcd\u7684\u57fa\u672c\u5f62\u5f0f\u3002\u4f8b\u5982\uff0c\u201cwas\u201d\u7684\u57fa\u672c\u5f62\u5f0f\u662f\u201cbe\u201d\uff0c\u201crats\u201d\u7684\u57fa\u672c\u5f62\u5f0f\u662f\u201crat\u201d\u3002<\/td>\n<\/tr>\n<tr>\n<td>Sentence Boundary Detection (SBD)<\/td>\n<td>\u67e5\u627e\u548c\u5206\u5272\u5355\u4e2a\u53e5\u5b50\u3002<\/td>\n<\/tr>\n<tr>\n<td>Named Entity Recognition (NER)<\/td>\n<td>\u5bf9\u547d\u540d\u7684\u201c\u73b0\u5b9e\u4e16\u754c\u201d\u5bf9\u8c61\u8fdb\u884c\u6807\u8bb0\uff0c\u5982\u4eba\u7269\u3001\u516c\u53f8\u6216\u5730\u70b9\u3002<\/td>\n<\/tr>\n<tr>\n<td>Entity Linking (EL)<\/td>\n<td>\u5c06\u6587\u672c\u5b9e\u4f53\u4e0e\u77e5\u8bc6\u5e93\u4e2d\u7684\u552f\u4e00\u6807\u8bc6\u7b26\u8fdb\u884c\u6d88\u5c90\u3002<\/td>\n<\/tr>\n<tr>\n<td>Similarity<\/td>\n<td>\u6bd4\u8f83\u5355\u8bcd\u3001\u6587\u672c\u7247\u6bb5\u548c\u6587\u6863\u4e4b\u95f4\u7684\u76f8\u4f3c\u7a0b\u5ea6\u3002<\/td>\n<\/tr>\n<tr>\n<td>Text Classification<\/td>\n<td>\u4e3a\u6574\u4e2a\u6587\u6863\u6216\u6587\u6863\u7684\u90e8\u5206\u5206\u914d\u7c7b\u522b\u6216\u6807\u7b7e\u3002<\/td>\n<\/tr>\n<tr>\n<td>Rule-based Matching<\/td>\n<td>\u6839\u636e\u5176\u6587\u672c\u548c\u8bed\u8a00\u6ce8\u91ca\u67e5\u627e\u6807\u8bb0\u5e8f\u5217\uff0c\u7c7b\u4f3c\u4e8e\u6b63\u5219\u8868\u8fbe\u5f0f\u3002<\/td>\n<\/tr>\n<tr>\n<td>Training<\/td>\n<td>\u66f4\u65b0\u548c\u6539\u8fdb\u7edf\u8ba1\u6a21\u578b\u7684\u9884\u6d4b\u80fd\u529b\u3002<\/td>\n<\/tr>\n<tr>\n<td>Serialization<\/td>\n<td>\u5c06\u5bf9\u8c61\u4fdd\u5b58\u5230\u6587\u4ef6\u6216\u5b57\u8282\u5b57\u7b26\u4e32\u4e2d\u3002<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h3>2.1 \u5206\u8bcd<\/h3>\n<p>\u5728\u5904\u7406\u8fc7\u7a0b\u4e2d\uff0cspaCy\u9996\u5148\u5bf9\u6587\u672c\u8fdb\u884c\u6807\u8bb0\uff0c\u5373\u5c06\u5176\u5206\u6bb5\u4e3a\u5355\u8bcd\u3001\u6807\u70b9\u7b26\u53f7\u7b49Token\u3002\u8fd9\u662f\u901a\u8fc7\u5e94\u7528\u6bcf\u79cd\u8bed\u8a00\u7279\u6709\u7684\u89c4\u5219\u6765\u5b9e\u73b0\u7684\u3002<strong>Token\u8868\u793a\u81ea\u7136\u8bed\u8a00\u6587\u672c\u7684\u6700\u5c0f\u5355\u4f4d\u3002\u6bcf\u4e2aToken\u90fd\u4ee3\u8868\u7740\u6587\u672c\u4e2d\u7684\u4e00\u4e2a\u539f\u5b50\u5143\u7d20\uff0c\u901a\u5e38\u662f\u5355\u8bcd\u6216\u6807\u70b9\u7b26\u53f7\u3002<\/strong><\/p>\n<pre><code class=\"language-python\">import spacy\n\nnlp = spacy.load(&quot;zh_core_web_sm&quot;)\n# \u4f7f\u5bf9\u6587\u672c\u8fdb\u884c\u4e00\u952e\u5904\u7406\ndoc = nlp(&quot;\u5357\u4eac\u957f\u6c5f\u5927\u6865\u662f\u91d1\u9675\u56db\u5341\u666f\u4e4b\u4e00\uff01&quot;)\n# \u904d\u5386doc\u4e2d\u7684\u6bcf\u4e2aToken\nfor token in doc:\n    print(token.text)\n&quot;&quot;&quot;\n\u5357\u4eac\n\u957f\u6c5f\n\u5927\u6865\n\u662f\n\u91d1\u9675\n\u56db\u5341\n\u666f\n\u4e4b\u4e00\n\uff01\n&quot;&quot;&quot;\n\nimport spacy\n\nnlp = spacy.load(&quot;en_core_web_sm&quot;)\ndoc = nlp(&quot;Apple is looking at buying U.K. startup for $1 billion&quot;)\nfor token in doc:\n    print(token.text)\n\n&quot;&quot;&quot;\nApple\nis\nlooking\nat\nbuying\nU.K.\nstartup\nfor\n$\n1\nbillion\n&quot;&quot;&quot;<\/code><\/pre>\n<p>\u5bf9\u4e8e\u4e2d\u6587\u5206\u8bcd\u6709\u65f6\u4f1a\u51fa\u73b0\u4e13\u6709\u540d\u8bcd\u88ab\u62c6\u5206\uff0c\u6bd4\u5982\u5357\u4eac\u957f\u6c5f\u5927\u6865\u88ab\u62c6\u5206\u4e3a\u5357\u4eac\u3001\u957f\u6c5f\u3001\u5927\u6865\u3002\u6211\u4eec\u53ef\u4ee5\u6dfb\u52a0\u81ea\u5b9a\u4e49\u8bcd\u5178\u6765\u89e3\u51b3\u8be5\u95ee\u9898\uff0c\u4f46\u662f\u8981\u6ce8\u610f\u7684\u662f\u81ea\u5b9a\u4e49\u8bcd\u5178\u6dfb\u52a0\u53ea\u9488\u5bf9\u67d0\u4e9b\u8bed\u8a00\u6a21\u578b\u3002<\/p>\n<pre><code class=\"language-python\">import spacy\n\nnlp = spacy.load(&quot;zh_core_web_sm&quot;)\n# \u6dfb\u52a0\u81ea\u5b9a\u4e49\u8bcd\u6c47\nnlp.tokenizer.pkuseg_update_user_dict([&quot;\u5357\u4eac\u957f\u6c5f\u5927\u6865&quot;,&quot;\u91d1\u9675\u56db\u5341\u666f&quot;])\n\ndoc = nlp(&quot;\u5357\u4eac\u957f\u6c5f\u5927\u6865\u662f\u91d1\u9675\u56db\u5341\u666f\u4e4b\u4e00\uff01&quot;)\nfor token in doc:\n    print(token.text)\n&quot;&quot;&quot;\n\u5357\u4eac\u957f\u6c5f\u5927\u6865\n\u662f\n\u91d1\u9675\u56db\u5341\u666f\n\u4e4b\u4e00\n\uff01\n&quot;&quot;&quot;<\/code><\/pre>\n<h3>2.2 \u8bcd\u6027\u6807\u6ce8\u4e0e\u4f9d\u5b58\u5173\u7cfb<\/h3>\n<p>spaCy\u5728\u5206\u8bcd\u540e\uff0c\u4f1a\u5bf9\u53e5\u5b50\u4e2d\u6bcf\u4e2a\u8bcd\u8fdb\u884c\u8bcd\u6027\u6807\u6ce8\u4ee5\u53ca\u786e\u5b9a\u53e5\u5b50\u4e2d\u5355\u8bcd\u4e4b\u95f4\u7684\u8bed\u6cd5\u5173\u7cfb\uff0c\u8981\u6ce8\u610f\u8fd9\u4e9b\u5173\u7cfb\u5177\u4f53\u8303\u56f4\u53d6\u51b3\u4e8e\u6240\u4f7f\u7528\u7684\u6a21\u578b\u3002\u793a\u4f8b\u4ee3\u7801\u5982\u4e0b\u6240\u793a\uff1a<\/p>\n<pre><code class=\"language-python\">import spacy\n\nnlp = spacy.load(&quot;en_core_web_sm&quot;)\ndoc = nlp(&quot;Apple is looking at buying U.K. startup for $1 billion&quot;)\n\n# token.text: \u5355\u8bcd\u7684\u539f\u59cb\u5f62\u5f0f\u3002\n# token.lemma_: \u5355\u8bcd\u7684\u57fa\u672c\u5f62\u5f0f\uff08\u6216\u8bcd\u5e72\uff09\u3002\u4f8b\u5982\uff0c\u201crunning\u201d\u7684\u8bcd\u5e72\u662f\u201crun\u201d\u3002\n# token.pos_: \u5355\u8bcd\u7684\u7c97\u7c92\u5ea6\u7684\u8bcd\u6027\u6807\u6ce8\uff0c\u5982\u540d\u8bcd\u3001\u52a8\u8bcd\u3001\u5f62\u5bb9\u8bcd\u7b49\u3002\n# token.tag_: \u5355\u8bcd\u7684\u7ec6\u7c92\u5ea6\u7684\u8bcd\u6027\u6807\u6ce8\uff0c\u63d0\u4f9b\u66f4\u591a\u7684\u8bed\u6cd5\u4fe1\u606f\u3002\n# token.dep_: \u5355\u8bcd\u5728\u53e5\u5b50\u4e2d\u7684\u4f9d\u5b58\u5173\u7cfb\u89d2\u8272\uff0c\u4f8b\u5982\u4e3b\u8bed\u3001\u5bbe\u8bed\u7b49\u3002\n# token.shape_: \u5355\u8bcd\u7684\u5f62\u72b6\u4fe1\u606f\uff0c\u4f8b\u5982\uff0c\u5355\u8bcd\u7684\u5927\u5c0f\u5199\uff0c\u662f\u5426\u6709\u6807\u70b9\u7b26\u53f7\u7b49\u3002\n# token.is_alpha: \u8fd9\u662f\u4e00\u4e2a\u5e03\u5c14\u503c\uff0c\u7528\u4e8e\u68c0\u67e5token\u662f\u5426\u5168\u90e8\u7531\u5b57\u6bcd\u7ec4\u6210\u3002\n# token.is_stop: \u8fd9\u662f\u4e00\u4e2a\u5e03\u5c14\u503c\uff0c\u7528\u4e8e\u68c0\u67e5token\u662f\u5426\u4e3a\u505c\u7528\u8bcd\uff08\u5982\u201cthe\u201d\u3001\u201cis\u201d\u7b49\u5728\u82f1\u8bed\u4e2d\u975e\u5e38\u5e38\u89c1\u4f46\u901a\u5e38\u4e0d\u5305\u542b\u592a\u591a\u4fe1\u606f\u7684\u8bcd\uff09\u3002\nfor token in doc:\n    print(token.text, token.lemma_, token.pos_, token.tag_, token.dep_,\n            token.shape_, token.is_alpha, token.is_stop)\n\n&quot;&quot;&quot;\nApple Apple PROPN NNP nsubj Xxxxx True False\nis be AUX VBZ aux xx True True\nlooking look VERB VBG ROOT xxxx True False\nat at ADP IN prep xx True True\nbuying buy VERB VBG pcomp xxxx True False\nU.K. U.K. PROPN NNP dobj X.X. False False\nstartup startup NOUN NN advcl xxxx True False\nfor for ADP IN prep xxx True True\n$ $ SYM $ quantmod $ False False\n1 1 NUM CD compound d False False\nbillion billion NUM CD pobj xxxx True False\n&quot;&quot;&quot;<\/code><\/pre>\n<p>\u5728\u4e0a\u8ff0\u4ee3\u7801\u4e2dpos_\u6240\u7528\u662f\u5e38\u89c1\u7684\u5355\u8bcd\u8bcd\u6027\u6807\u6ce8\u3002tag_\u6240\u652f\u6301\u7684\u8bcd\u6027\u6807\u6ce8\u53ca\u89e3\u91ca\u5982\u4e0b\uff1a<\/p>\n<pre><code class=\"language-python\"># \u83b7\u53d6\u6240\u6709\u8bcd\u6027\u6807\u6ce8\uff08tag_\uff09\u548c\u5b83\u4eec\u7684\u63cf\u8ff0\ntag_descriptions = {tag: spacy.explain(tag) for tag in nlp.get_pipe(&#039;tagger&#039;).labels}\n# \u6253\u5370\u8bcd\u6027\u6807\u6ce8\uff08tag_\uff09\u53ca\u5176\u63cf\u8ff0\n# print(&quot;\u8bcd\u6027\u6807\u6ce8 (TAG) \u53ca\u5176\u63cf\u8ff0\uff1a&quot;)\n# for tag, description in tag_descriptions.items():\n#     print(f&quot;{tag}: {description}&quot;)<\/code><\/pre>\n<p>dep__\u6240\u652f\u6301\u7684\u4f9d\u5b58\u5173\u7cfb\u53ca\u89e3\u91ca\u5982\u4e0b\uff1a<\/p>\n<pre><code class=\"language-python\"># \u83b7\u53d6\u6240\u6709\u4f9d\u5b58\u5173\u7cfb\u6807\u6ce8\u548c\u5b83\u4eec\u7684\u63cf\u8ff0\npar_descriptions = {par: spacy.explain(par) for par in nlp.get_pipe(&#039;parser&#039;).labels}\n# print(&quot;\u4f9d\u5b58\u5173\u7cfb\u53ca\u5176\u63cf\u8ff0\uff1a&quot;)\n# for par, description in par_descriptions.items():\n#     print(f&quot;{par}: {description}&quot;)<\/code><\/pre>\n<h3>2.3 <a href=\"https:\/\/so.csdn.net\/so\/search?q=\u547d\u540d\u5b9e\u4f53\u8bc6\u522b&amp;spm=1001.2101.3001.7020\">\u547d\u540d\u5b9e\u4f53\u8bc6\u522b<\/a><\/h3>\n<p>\u547d\u540d\u5b9e\u4f53\u8bc6\u522b(Named Entity Recognition, \u7b80\u79f0NER)\u662f\u81ea\u7136\u8bed\u8a00\u5904\u7406\u4e2d\u7684\u4e00\u9879\u57fa\u7840\u4efb\u52a1\uff0c\u5e94\u7528\u8303\u56f4\u975e\u5e38\u5e7f\u6cdb\u3002 NER\u662f\u6307\u8bc6\u522b\u6587\u672c\u4e2d\u5177\u6709\u7279\u5b9a\u610f\u4e49\u6216\u8005\u6307\u4ee3\u6027\u5f3a\u7684\u5b9e\u4f53\uff0c\u901a\u5e38\u5305\u62ec\u4eba\u540d\u3001\u5730\u540d\u3001\u673a\u6784\u540d\u3001\u65e5\u671f\u65f6\u95f4\u3001\u4e13\u6709\u540d\u8bcd\u7b49\u3002spaCy\u4f7f\u7528\u5982\u4e0b\uff1a<\/p>\n<pre><code class=\"language-python\">import spacy\n\nnlp = spacy.load(&quot;zh_core_web_sm&quot;)\n# \u6dfb\u52a0\u81ea\u5b9a\u4e49\u8bcd\u6c47\nnlp.tokenizer.pkuseg_update_user_dict([&quot;\u4e1c\u65b9\u660e\u73e0&quot;])\n\n# \u81ea\u5b9a\u4e49\u8bcd\u6c47\u53ef\u80fd\u4e0d\u4f1a\u8fdb\u5165\u5b9e\u4f53\u8bc6\u522b\u3002\ndoc = nlp(&quot;\u4e1c\u65b9\u660e\u73e0\u662f\u4e00\u5ea7\u4f4d\u4e8e\u4e2d\u56fd\u4e0a\u6d77\u5e02\u7684\u6807\u5fd7\u6027\u5efa\u7b51\uff0c\u5efa\u9020\u4e8e1991\u5e74\uff0c\u662f\u4e00\u5ea7\u9ad8\u5ea6\u4e3a468\u7c73\u7684\u7535\u89c6\u5854\u3002&quot;)\nfor ent in doc.ents:\n    # \u5b9e\u4f53\u6587\u672c\uff0c\u5f00\u59cb\u4f4d\u7f6e\uff0c\u7ed3\u675f\u4f4d\u7f6e\uff0c\u5b9e\u4f53\u6807\u7b7e\n    print(ent.text, ent.start_char, ent.end_char, ent.label_)\n\n&quot;&quot;&quot;\n\u4e2d\u56fd\u4e0a\u6d77\u5e02 9 14 GPE\n1991\u5e74 24 29 DATE\n468\u7c73 36 40 QUANTITY\n&quot;&quot;&quot;<\/code><\/pre>\n<p>\u5982\u679c\u60f3\u77e5\u9053\u6240\u6709\u7684\u5b9e\u4f53\u4ee5\u53ca\u5176\u5bf9\u5e94\u7684\u542b\u4e49\uff0c\u53ef\u4ee5\u6267\u884c\u4ee5\u4e0b\u4ee3\u7801\uff1a<\/p>\n<pre><code class=\"language-python\"># \u83b7\u53d6\u547d\u540d\u5b9e\u4f53\u6807\u7b7e\u53ca\u5176\u542b\u4e49\nentity_labels = nlp.get_pipe(&#039;ner&#039;).labels\n\n# \u6253\u5370\u8f93\u51fa\u6240\u6709\u547d\u540d\u5b9e\u4f53\u53ca\u5176\u542b\u4e49\nfor label in entity_labels:\n    print(&quot;{}: {}&quot;.format(label,spacy.explain(label)))\n123456\nCARDINAL: Numerals that do not fall under another type\nDATE: Absolute or relative dates or periods\nEVENT: Named hurricanes, battles, wars, sports events, etc.\nFAC: Buildings, airports, highways, bridges, etc.\nGPE: Countries, cities, states\nLANGUAGE: Any named language\nLAW: Named documents made into laws.\nLOC: Non-GPE locations, mountain ranges, bodies of water\nMONEY: Monetary values, including unit\nNORP: Nationalities or religious or political groups\nORDINAL: &quot;first&quot;, &quot;second&quot;, etc.\nORG: Companies, agencies, institutions, etc.\nPERCENT: Percentage, including &quot;%&quot;\nPERSON: People, including fictional\nPRODUCT: Objects, vehicles, foods, etc. (not services)\nQUANTITY: Measurements, as of weight or distance\nTIME: Times smaller than a day\nWORK_OF_ART: Titles of books, songs, etc.\n123456789101112131415161718<\/code><\/pre>\n<h3>2.4 \u8bcd\u5411\u91cf\u4e0e\u76f8\u4f3c\u6027<\/h3>\n<p>\u8bcd\u5411\u91cf\u662f\u81ea\u7136\u8bed\u8a00\u5904\u7406\u4e2d\u4e00\u79cd\u91cd\u8981\u7684\u8868\u793a\u65b9\u5f0f\uff0c\u5b83\u5c06\u5355\u8bcd\u6620\u5c04\u4e3a\u5b9e\u6570\u5411\u91cf\u3002\u8fd9\u79cd\u8868\u793a\u65b9\u5f0f\u80fd\u591f\u6355\u6349\u5355\u8bcd\u4e4b\u95f4\u7684\u8bed\u4e49\u5173\u7cfb\uff0c\u5e76\u5c06\u8bed\u4e49\u4fe1\u606f\u8f6c\u5316\u4e3a\u8ba1\u7b97\u673a\u80fd\u591f\u5904\u7406\u7684\u6570\u503c\u5f62\u5f0f\u3002<\/p>\n<p>\u4f20\u7edf\u7684\u81ea\u7136\u8bed\u8a00\u5904\u7406\u65b9\u6cd5\u5f80\u5f80\u5c06\u6587\u672c\u8868\u793a\u4e3a\u79bb\u6563\u7684\u7b26\u53f7\uff0c\u4f8b\u5982\u72ec\u70ed\u7f16\u7801\u6216\u8005\u8bcd\u888b\u6a21\u578b\u3002\u7136\u800c\uff0c\u8fd9\u79cd\u65b9\u6cd5\u5ffd\u7565\u4e86\u5355\u8bcd\u4e4b\u95f4\u7684\u8bed\u4e49\u76f8\u4f3c\u6027\uff0c\u800c\u4e14\u7ef4\u5ea6\u8fc7\u9ad8\uff0c\u9020\u6210\u7a00\u758f\u6027\u95ee\u9898\u3002\u76f8\u6bd4\u4e4b\u4e0b\uff0c\u8bcd\u5411\u91cf\u901a\u8fc7\u5c06\u6bcf\u4e2a\u5355\u8bcd\u6620\u5c04\u5230\u8fde\u7eed\u7684\u5411\u91cf\u7a7a\u95f4\u4e2d\uff0c\u53ef\u4ee5\u66f4\u597d\u5730\u6355\u6349\u5355\u8bcd\u4e4b\u95f4\u7684\u8bed\u4e49\u5173\u7cfb\uff0c\u5e76\u4e14\u964d\u4f4e\u4e86\u7279\u5f81\u7a7a\u95f4\u7684\u7ef4\u5ea6\uff0c\u4f7f\u5f97\u6587\u672c\u5904\u7406\u66f4\u52a0\u9ad8\u6548\u3002\u901a\u8fc7\u8ba1\u7b97\u4e24\u4e2a\u8bcd\u5411\u91cf\u4e4b\u95f4\u7684\u8ddd\u79bb\u6216\u5939\u89d2\u53ef\u4ee5\u8861\u91cf\u8bcd\u5411\u91cf\u7684\u76f8\u4f3c\u6027\u3002<\/p>\n<p>\u63d0\u53d6\u53e5\u5b50\u4e2d\u6bcf\u4e00\u4e2a\u8bcd\u7684\u8bcd\u5411\u91cf\u4ee3\u7801\u5982\u4e0b\uff1a<\/p>\n<pre><code class=\"language-python\">import spacy\n\n# \u52a0\u8f7d\u4e2d\u6587\u6a21\u578b&quot;zh_core_web_sm&quot;\nnlp = spacy.load(&quot;zh_core_web_sm&quot;)\n\n# \u5bf9\u7ed9\u5b9a\u6587\u672c\u8fdb\u884c\u5206\u8bcd\u548c\u8bcd\u6027\u6807\u6ce8\ntokens = nlp(&quot;\u4e1c\u65b9\u660e\u73e0\u662f\u4e00\u5ea7\u4f4d\u4e8e\u4e2d\u56fd\u4e0a\u6d77\u5e02\u7684\u6807\u5fd7\u6027\u5efa\u7b51\uff01&quot;)\n\n# \u904d\u5386\u5206\u8bcd\u540e\u7684\u6bcf\u4e2a\u8bcd\u8bed\nfor token in tokens:\n    # \u8f93\u51fa\u8bcd\u8bed\u7684\u6587\u672c\u5185\u5bb9\u3001\u662f\u5426\u6709\u5bf9\u5e94\u7684\u5411\u91cf\u8868\u793a\u3001\u5411\u91cf\u8303\u6570\u548c\u662f\u5426\u4e3a\u672a\u767b\u5f55\u8bcd\uff08Out-of-vocabulary\uff0c\u5373\u4e0d\u5728\u8bcd\u5411\u91cf\u8bcd\u5178\u4e2d\u7684\u8bcd\uff09\n    print(token.text, token.has_vector, token.vector_norm, token.is_oov)\n\n&quot;&quot;&quot;\n\u4e1c\u65b9 True 11.572288 True\n\u660e\u73e0 True 10.620552 True\n\u662f True 12.337883 True\n\u4e00 True 12.998204 True\n\u5ea7\u4f4d True 10.186406 True\n\u4e8e True 13.540245 True\n\u4e2d\u56fd True 12.459145 True\n\u4e0a\u6d77\u5e02 True 12.004954 True\n\u7684 True 12.90457 True\n\u6807\u5fd7\u6027 True 13.601862 True\n\u5efa\u7b51 True 10.46621 True\n\uff01 True 12.811246 True\n&quot;&quot;&quot;<\/code><\/pre>\n<p>\u5982\u679c\u60f3\u5f97\u5230\u67d0\u4e2a\u53e5\u5b50\u6216\u8005\u67d0\u4e2a\u8bcd\u7684\u8bcd\u5411\u91cf\uff0c\u4ee3\u7801\u5982\u4e0b\uff1a<\/p>\n<pre><code class=\"language-python\"># \u8be5\u8bcd\u5411\u91cf\u6ca1\u6709\u5f52\u4e00\u5316\ntokens.vector.shape\n&quot;&quot;&quot;\n(96,)\n&quot;&quot;&quot;<\/code><\/pre>\n<p>spaCy\u63d0\u4f9b\u4e86similarity\u51fd\u6570\u4ee5\u8ba1\u7b97\u4e24\u4e2a\u6587\u672c\u5411\u91cf\u7684\u76f8\u4f3c\u5ea6\u3002<br \/>\n\u793a\u4f8b\u4ee3\u7801\u5982\u4e0b\uff1a<\/p>\n<pre><code class=\"language-python\">import spacy\n\n# \u4e3a\u4e86\u65b9\u4fbf\u8fd9\u91cc\u4f7f\u7528\u5c0f\u6a21\u578b\uff0c\u63a8\u8350\u4f7f\u7528\u66f4\u5927\u7684\u6a21\u578b\nnlp = spacy.load(&quot;zh_core_web_sm&quot;)  \ndoc1 = nlp(&quot;\u4e1c\u65b9\u660e\u73e0\u662f\u4e00\u5ea7\u4f4d\u4e8e\u4e2d\u56fd\u4e0a\u6d77\u5e02\u7684\u6807\u5fd7\u6027\u5efa\u7b51&quot;)\ndoc2 = nlp(&quot;\u5357\u4eac\u957f\u6c5f\u5927\u6865\u662f\u91d1\u9675\u56db\u5341\u666f\u4e4b\u4e00\uff01&quot;)\n\n# \u8ba1\u7b97\u4e24\u4e2a\u6587\u672c\u7684\u76f8\u4f3c\u5ea6\nprint(doc1, &quot;&lt;-&gt;&quot;, doc2, doc1.similarity(doc2))\n\n&quot;&quot;&quot;\n\u4e1c\u65b9\u660e\u73e0\u662f\u4e00\u5ea7\u4f4d\u4e8e\u4e2d\u56fd\u4e0a\u6d77\u5e02\u7684\u6807\u5fd7\u6027\u5efa\u7b51 &lt;-&gt; \u5357\u4eac\u957f\u6c5f\u5927\u6865\u662f\u91d1\u9675\u56db\u5341\u666f\u4e4b\u4e00\uff01 0.5743045135827821\n&quot;&quot;&quot;<\/code><\/pre>\n<p>\u5728\u4e0a\u9762\u4ee3\u7801\u4e2d\u76f8\u4f3c\u5ea6\u8ba1\u7b97\u65b9\u5f0f\u9ed8\u8ba4\u4f7f\u7528\u4f59\u5f26\u76f8\u4f3c\u5ea6\u3002\u4f59\u5f26\u76f8\u4f3c\u5ea6\u8303\u56f4\u4e3a0\u52301\uff0c\u6570\u503c\u8d8a\u9ad8\u8868\u660e\u8bcd\u5411\u91cf\u8d8a\u76f8\u4f3c\u3002\u4e00\u822c\u6765\u8bf4\u8bcd\u5411\u91cf\u76f8\u4f3c\u5ea6\u4f7f\u7528spacy\u901a\u7528\u6a21\u578b\u51c6\u786e\u5ea6\u53ef\u80fd\u5f88\u4f4e\uff0c\u53ef\u4ee5\u5c1d\u8bd5\u4f7f\u7528\u4e13\u7528\u6a21\u578b\u6216\u8005\u81ea\u884c\u8bad\u7ec3\u6a21\u578b\uff0cspacy\u63a8\u8350\u4f7f\u7528<a href=\"https:\/\/github.com\/explosion\/sense2vec\">sense2vec<\/a>\u6765\u8ba1\u7b97\u6a21\u578b\u76f8\u4f3c\u5ea6\u3002<\/p>\n<p>\u6b64\u5916\u5982\u679c\u4ec5\u4ec5\u4f7f\u7528spacy\u63d0\u53d6\u6587\u672c\u5411\u91cf\uff0c\u53ef\u4ee5\u7528numpy\u624b\u52a8\u8ba1\u7b97\u6587\u672c\u76f8\u4f3c\u5ea6\uff0c\u4ee3\u7801\u5982\u4e0b\uff1a<\/p>\n<pre><code class=\"language-python\">import numpy as np\nimport spacy\nnlp = spacy.load(&quot;zh_core_web_sm&quot;)  \ndoc1 = nlp(&quot;\u4e1c\u65b9\u660e\u73e0\u662f\u4e00\u5ea7\u4f4d\u4e8e\u4e2d\u56fd\u4e0a\u6d77\u5e02\u7684\u6807\u5fd7\u6027\u5efa\u7b51&quot;)\ndoc2 = nlp(&quot;\u5357\u4eac\u957f\u6c5f\u5927\u6865\u662f\u91d1\u9675\u56db\u5341\u666f\u4e4b\u4e00\uff01&quot;)\n# \u83b7\u53d6doc1\u548cdoc2\u7684\u8bcd\u5411\u91cf\nvec1 = doc1.vector\nvec2 = doc2.vector\n\n# \u4f7f\u7528NumPy\u8ba1\u7b97\u76f8\u4f3c\u5ea6\u5f97\u5206\uff0cnp.linalg.norm(vec1)\u5c31\u662fdoc1.vector_norm\nsimilarity_score = np.dot(vec1, vec2) \/ (np.linalg.norm(vec1) * np.linalg.norm(vec2))\n\nprint(doc1, &quot;&lt;-&gt;&quot;, doc2,similarity_score)\n\n&quot;&quot;&quot;\n\u4e1c\u65b9\u660e\u73e0\u662f\u4e00\u5ea7\u4f4d\u4e8e\u4e2d\u56fd\u4e0a\u6d77\u5e02\u7684\u6807\u5fd7\u6027\u5efa\u7b51 &lt;-&gt; \u5357\u4eac\u957f\u6c5f\u5927\u6865\u662f\u91d1\u9675\u56db\u5341\u666f\u4e4b\u4e00\uff01 0.5743046\n&quot;&quot;&quot;<\/code><\/pre>\n<h2>3 spaCy\u7ed3\u6784\u4f53\u7cfb<\/h2>\n<h3>3.1 spaCy\u5904\u7406\u6d41\u7a0b<\/h3>\n<p>\u5f53\u5728\u4e00\u4e2a\u6587\u672c\u4e0a\u8c03\u7528nlp\u6a21\u578b\u65f6\uff0cspaCy\u9996\u5148\u5bf9\u6587\u672c\u8fdb\u884c\u5206\u8bcd\u5904\u7406\uff0c\u751f\u6210\u4e00\u4e2aDoc\u5bf9\u8c61\u3002\u63a5\u7740\uff0cDoc\u5bf9\u8c61\u5c06\u5728\u51e0\u4e2a\u4e0d\u540c\u7684\u6b65\u9aa4\u4e2d\u8fdb\u884c\u5904\u7406\u3002\u8bad\u7ec3\u597d\u7684\u5904\u7406\u6d41\u7a0b\u901a\u5e38\u5305\u62ec\u8bcd\u6027\u6807\u6ce8\u5668\u3001\u4f9d\u5b58\u53e5\u6cd5\u89e3\u6790\u5668\u548c\u5b9e\u4f53\u8bc6\u522b\u5668\u7b49\u5904\u7406\u7ec4\u4ef6\u3002\u8fd9\u4e9b\u7ec4\u4ef6\u76f8\u4e92\u72ec\u7acb\uff0c\u6bcf\u4e2a\u5904\u7406\u6d41\u7a0b\u7ec4\u4ef6\u90fd\u4f1a\u8fd4\u56de\u5904\u7406\u540e\u7684Doc\u5bf9\u8c61\uff0c\u7136\u540e\u5c06\u5176\u4f20\u9012\u7ed9\u4e0b\u4e00\u4e2a\u7ec4\u4ef6\u3002<br \/>\n\u6700\u7ec8\u751f\u6210\u7684Doc\u5bf9\u8c61\u662f\u4e00\u4e2a\u5305\u542b\u4e86\u6240\u6709\u5355\u8bcd\u548c\u6807\u70b9\u7b26\u53f7\u7684\u5e8f\u5217\uff0c\u6bcf\u4e2a\u5355\u8bcd\u88ab\u8868\u793a\u4e3aToken\u5bf9\u8c61\u3002\u6bcf\u4e2aToken\u5bf9\u8c61\u5305\u542b\u4e86\u5355\u8bcd\u672c\u8eab\u7684\u5185\u5bb9\u3001\u8bcd\u6027\u6807\u6ce8\u3001\u8bcd\u5f62\u8fd8\u539f\u540e\u7684\u5f62\u5f0f\u7b49\u4fe1\u606f\u3002\u4ee5\u4e0b\u56fe\u7247\u89e3\u91ca\u4e86\u4f7f\u7528spaCy\u8fdb\u884c\u6587\u672c\u5904\u7406\u7684\u8fc7\u7a0b\u3002<\/p>\n<p><img src=\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/dcf2d56ed46f4eeab5015d42da88db60-1705586515523-25.png\" alt=\"\u5728\u8fd9\u91cc\u63d2\u5165\u56fe\u7247\u63cf\u8ff0\" \/><\/p>\n<p>\u5982\u4e0a\u56fe\u6240\u793a\uff0c\u6a21\u578b\u7ba1\u9053\u4e2d\u6240\u6d89\u53ca\u5230\u7684\u6a21\u5757\u4e3b\u8981\u53d6\u51b3\u4e8e\u8be5\u6a21\u578b\u7684\u7ed3\u6784\u548c\u8bad\u7ec3\u65b9\u5f0f\u3002\u5176\u4e2d\u5206\u8bcdtokenizer\u662f\u4e00\u4e2a\u7279\u6b8a\u7684\u7ec4\u4ef6\u4e14\u72ec\u7acb\u4e8e\u5176\u4ed6\u7ec4\u4ef6\u4e4b\u5916\uff0c\u8fd9\u662f\u56e0\u4e3a\u5176\u4ed6\u7ec4\u4ef6\u6a21\u5757\u5728\u8c03\u7528\u524d\u90fd\u4f1a\u5148\u8c03\u7528tokenizer\u4ee5\u5bf9\u5b57\u7b26\u4e32\u8fdb\u884c\u5206\u8bcd\u3002\u6240\u6709\u652f\u6301\u4e3b\u8981\u7684\u6a21\u5757\u5982\u4e0b\uff0c\u8fd9\u4e9b\u6a21\u5757\u7684\u4f7f\u7528\u5df2\u5728\u524d\u4e00\u7ae0\u8fdb\u884c\u4ecb\u7ecd\u3002<\/p>\n<table>\n<thead>\n<tr>\n<th>\u540d\u79f0<\/th>\n<th>\u7ec4\u4ef6<\/th>\n<th>\u521b\u5efa<\/th>\n<th>\u63cf\u8ff0<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>tokenizer<\/td>\n<td>Tokenizer<\/td>\n<td>Doc<\/td>\n<td>\u5c06\u6587\u672c\u5206\u5272\u4e3a\u6807\u8bb0\u3002<\/td>\n<\/tr>\n<tr>\n<td>tagger<\/td>\n<td>Tagger<\/td>\n<td>Token.tag<\/td>\n<td>\u4e3a\u6807\u8bb0\u5206\u914d\u8bcd\u6027\u6807\u7b7e\u3002<\/td>\n<\/tr>\n<tr>\n<td>parser<\/td>\n<td>DependencyParser<\/td>\n<td>Token.head,Token.dep,Doc.sents,Doc.noun_chunks<\/td>\n<td>\u5206\u914d\u4f9d\u8d56\u5173\u7cfb\u6807\u7b7e\u3002<\/td>\n<\/tr>\n<tr>\n<td>ner<\/td>\n<td>EntityRecognizer<\/td>\n<td>Doc.ents,Token.ent_iob,Token.ent_type<\/td>\n<td>\u68c0\u6d4b\u548c\u6807\u8bb0\u547d\u540d\u5b9e\u4f53\u3002<\/td>\n<\/tr>\n<tr>\n<td>lemmatizer<\/td>\n<td>Lemmatizer<\/td>\n<td>Token.lemma<\/td>\n<td>\u5206\u914d\u5355\u8bcd\u7684\u57fa\u672c\u5f62\u5f0f\u3002<\/td>\n<\/tr>\n<tr>\n<td>textcat<\/td>\n<td>TextCategorizer<\/td>\n<td>Doc.cats<\/td>\n<td>\u5206\u914d\u6587\u6863\u6807\u7b7e\u3002<\/td>\n<\/tr>\n<tr>\n<td>custom<\/td>\n<td>\u81ea\u5b9a\u4e49\u7ec4\u4ef6<\/td>\n<td>Doc.<em>.xxx,Token.<\/em>.xxx,Span._.xxx<\/td>\n<td>\u5206\u914d\u81ea\u5b9a\u4e49\u5c5e\u6027\u3001\u65b9\u6cd5\u6216\u5c5e\u6027\u3002<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u4e00\u4e2aspacy\u7684\u6a21\u578b\u6240\u652f\u6301\u7684\u6587\u672c\u5904\u7406\u7ec4\u4ef6\u67e5\u770b\u65b9\u5f0f\u5982\u4e0b\uff1a<\/p>\n<pre><code class=\"language-python\">import spacy\n\n# \u52a0\u8f7d\u4e2d\u6587\u6a21\u578b&quot;zh_core_web_sm&quot;\nnlp = spacy.load(&quot;zh_core_web_sm&quot;)\n# \u67e5\u770b\u6240\u652f\u6301\u7684\u7ec4\u4ef6\nnlp.pipe_names\n\n&quot;&quot;&quot;\n[&#039;tok2vec&#039;, &#039;tagger&#039;, &#039;parser&#039;, &#039;attribute_ruler&#039;, &#039;ner&#039;]\n&quot;&quot;&quot;<\/code><\/pre>\n<p>\u57fa\u4e8e\u4ee5\u4e0b\u4ee3\u7801\u53ef\u4ee5\u63a7\u5236\u7ec4\u4ef6\u7684\u9009\u62e9\u548c\u4f7f\u7528\uff0c\u4ee5\u52a0\u5feb\u6267\u884c\u901f\u5ea6\uff1a<\/p>\n<pre><code class=\"language-python\"># \u52a0\u8f7d\u4e0d\u5305\u542b\u547d\u540d\u5b9e\u4f53\u8bc6\u522b\u5668\uff08NER\uff09\u7684\u7ba1\u9053\nnlp = spacy.load(&quot;zh_core_web_sm&quot;, exclude=[&quot;ner&quot;])\n# \u67e5\u770b\u6240\u652f\u6301\u7684\u7ec4\u4ef6\nnlp.pipe_names\n&quot;&quot;&quot;\n[&#039;tok2vec&#039;, &#039;tagger&#039;, &#039;parser&#039;, &#039;attribute_ruler&#039;]\n&quot;&quot;&quot;\n# \u53ea\u542f\u7528tagger\u7ba1\u9053\nnlp = spacy.load(&quot;zh_core_web_sm&quot;,enable=[ &quot;tagger&quot;])\nnlp.pipe_names\n\n&quot;&quot;&quot;\n[&#039;tagger&#039;]\n&quot;&quot;&quot;\n\n# \u52a0\u8f7d\u8bcd\u6027\u6807\u6ce8\u5668\uff08tagger\uff09\u548c\u4f9d\u5b58\u53e5\u6cd5\u89e3\u6790\u5668\uff08parser\uff09\uff0c\u4f46\u4e0d\u542f\u7528\u5b83\u4eec\nnlp = spacy.load(&quot;zh_core_web_sm&quot;, disable=[&quot;tagger&quot;, &quot;parser&quot;],)\n# \u7981\u7528\u67d0\u4e9b\u7ec4\u4ef6\nnlp.disable_pipe(&quot;ner&quot;)\nnlp.pipe_names\n\n&quot;&quot;&quot;\n[&#039;tok2vec&#039;, &#039;attribute_ruler&#039;]\n&quot;&quot;&quot;<\/code><\/pre>\n<h3>3.2 spaCy\u5de5\u7a0b\u7ed3\u6784<\/h3>\n<p>spaCy\u4e2d\u7684\u4e2d\u5fc3\u6570\u636e\u7ed3\u6784\u662fLanguage\u7c7b\u3001Vocab\u548cDoc\u5bf9\u8c61\u3002Language\u7c7b\u7528\u4e8e\u5904\u7406\u6587\u672c\u5e76\u5c06\u5176\u8f6c\u6362\u4e3aDoc\u5bf9\u8c61\u3002\u5b83\u901a\u5e38\u5b58\u50a8\u4e3a\u4e00\u4e2a\u540d\u4e3anlp\u7684\u53d8\u91cf\u3002Doc\u5bf9\u8c61\u62e5\u6709\u4ee4\u724c\u5e8f\u5217\u53ca\u5176\u6240\u6709\u6ce8\u91ca\u3002\u901a\u8fc7\u5728Vocab\u4e2d\u96c6\u4e2d\u5b57\u7b26\u4e32\u3001\u8bcd\u5411\u91cf\u548c\u8bcd\u6cd5\u5c5e\u6027\u3002\u8fd9\u4e9b\u4e3b\u8981\u7c7b\u548c\u5bf9\u8c61\u7684\u4ecb\u7ecd\u5982\u4e0b\u6240\u793a\uff1a<\/p>\n<p><img src=\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/bf6bd566d63b484ead0baf6826f6c89d.png\" alt=\"\u5728\u8fd9\u91cc\u63d2\u5165\u56fe\u7247\u63cf\u8ff0\" \/><\/p>\n<p>\u5e38\u7528\u6a21\u5757\u7684\u4ecb\u7ecd\u5982\u4e0b\uff1a<\/p>\n<p><strong>Doc<\/strong><\/p>\n<p>Doc\u662fspaCy\u4e2d\u4e00\u4e2a\u91cd\u8981\u7684\u5bf9\u8c61\uff0c\u5b83\u4ee3\u8868\u4e86\u4e00\u4e2a\u6587\u672c\u6587\u6863\uff0c\u5e76\u5305\u542b\u4e86\u6587\u672c\u4e2d\u7684\u6240\u6709\u4fe1\u606f\uff0c\u5982\u5355\u8bcd\u3001\u6807\u70b9\u3001\u8bcd\u6027\u3001\u4f9d\u8d56\u5173\u7cfb\u7b49\u3002\u53ef\u4ee5\u901a\u8fc7spaCy\u7684Language\u5bf9\u8c61\u5bf9\u6587\u672c\u8fdb\u884c\u5904\u7406\uff0c\u5f97\u5230\u4e00\u4e2aDoc\u5bf9\u8c61\u3002<\/p>\n<p><strong>DocBin<\/strong><\/p>\n<p>DocBin \u662f\u7528\u4e8e\u9ad8\u6548\u5e8f\u5217\u5316\u548c\u53cd\u5e8f\u5217\u5316Doc\u5bf9\u8c61\u7684\u6570\u636e\u7ed3\u6784\uff0c\u4ee5\u5728\u4e0d\u540c\u7684\u8fc7\u7a0b\u4e2d\u4fdd\u5b58\u548c\u52a0\u8f7dDoc\u5bf9\u8c61\u3002\u4f7f\u7528\u4ee3\u7801\u5982\u4e0b\uff1a<\/p>\n<pre><code class=\"language-python\"># \u5bfc\u5165\u6240\u9700\u7684\u5e93\nimport spacy\nfrom spacy.tokens import DocBin\n\n# \u52a0\u8f7d\u82f1\u6587\u9884\u8bad\u7ec3\u6a21\u578b\nnlp = spacy.load(&quot;en_core_web_sm&quot;)\n\n# \u5b9a\u4e49\u5f85\u5904\u7406\u6587\u672c\ntexts = [&quot;This is sentence 1.&quot;, &quot;And this is sentence 2.&quot;]\n\n# \u5c06\u6bcf\u4e2a\u6587\u672c\u8f6c\u5316\u4e3aDoc\u5bf9\u8c61\uff0c\u7528nlp\u5904\u7406\u5e76\u4fdd\u5b58\u5230docs\u5217\u8868\u4e2d\ndocs = [nlp(text) for text in texts]\n\n# \u521b\u5efa\u4e00\u4e2a\u65b0\u7684DocBin\u5bf9\u8c61\uff0c\u7528\u4e8e\u4fdd\u5b58\u6587\u6863\u6570\u636e\uff0c\u5e76\u542f\u7528\u5b58\u50a8\u7528\u6237\u6570\u636e\u7684\u529f\u80fd\ndocbin = DocBin(store_user_data=True)\n\n# \u5c06\u6bcf\u4e2aDoc\u5bf9\u8c61\u6dfb\u52a0\u5230DocBin\u4e2d\nfor doc in docs:\n    docbin.add(doc)\n\n# \u5c06DocBin\u4fdd\u5b58\u5230\u6587\u4ef6\u4e2d\nwith open(&quot;documents.spacy&quot;, &quot;wb&quot;) as f:\n    f.write(docbin.to_bytes())\n\n# \u4ece\u6587\u4ef6\u4e2d\u52a0\u8f7dDocBin\nwith open(&quot;documents.spacy&quot;, &quot;rb&quot;) as f:\n    bytes_data = f.read()\n\n# \u4ece\u5b57\u8282\u6570\u636e\u4e2d\u6062\u590d\u52a0\u8f7dDocBin\u5bf9\u8c61\nloaded_docbin = DocBin().from_bytes(bytes_data)\n\n# \u4f7f\u7528nlp.vocab\u83b7\u53d6\u8bcd\u6c47\u8868\uff0c\u5e76\u901a\u8fc7DocBin\u83b7\u53d6\u6240\u6709\u52a0\u8f7d\u7684\u6587\u6863\nloaded_docs = list(loaded_docbin.get_docs(nlp.vocab))\n\n# \u8f93\u51fa\u52a0\u8f7d\u7684\u6587\u6863\nloaded_docs\n\n&quot;&quot;&quot;\n[This is sentence 1., And this is sentence 2.]\n&quot;&quot;&quot;<\/code><\/pre>\n<p><strong>Example<\/strong><\/p>\n<p>Example\u7528\u4e8e\u8bad\u7ec3spaCy\u6a21\u578b\uff0c\u5b83\u5305\u542b\u4e86\u4e00\u4e2a\u8f93\u5165\u6587\u672c(Doc)\u548c\u5176\u5bf9\u5e94\u7684\u6807\u6ce8\u6570\u636e\u3002\u5173\u4e8espaCy\u6a21\u578b\u7684\u8bad\u7ec3\uff0c\u89c1:<a href=\"https:\/\/spacy.io\/usage\/training\">spaCy-training<\/a><\/p>\n<p><strong>Language<\/strong><\/p>\n<p>Language\u662fspaCy\u7684\u6838\u5fc3\u5bf9\u8c61\u4e4b\u4e00\uff0c\u5b83\u8d1f\u8d23\u5904\u7406\u6587\u672c\u7684\u9884\u5904\u7406\u3001\u8bcd\u6027\u6807\u6ce8\u3001\u53e5\u6cd5\u5206\u6790\u7b49\u4efb\u52a1\u3002\u53ef\u4ee5\u901a\u8fc7spacy.load()\u6765\u52a0\u8f7d\u4e00\u4e2a\u5177\u4f53\u7684\u8bed\u8a00\u6a21\u578b\uff0c\u83b7\u53d6\u5bf9\u5e94\u7684Language\u5bf9\u8c61\u3002<\/p>\n<p><strong>Lexeme<\/strong><\/p>\n<p>Lexeme \u662f\u4e00\u4e2a\u5355\u8bcd\u5728\u8bcd\u6c47\u8868\u4e2d\u7684\u8868\u793a\uff0c\u5b83\u5305\u542b\u4e86\u5173\u4e8e\u8be5\u5355\u8bcd\u7684\u5404\u79cd\u4fe1\u606f\uff0c\u5982\u8bcd\u6027\u3001\u8bcd\u9891\u7b49\u3002\u793a\u4f8b\u4ee3\u7801\u5982\u4e0b\uff1a<\/p>\n<pre><code class=\"language-python\">nlp = spacy.load(&quot;en_core_web_sm&quot;)\n\n# \u5b9a\u4e49\u4e00\u4e2a\u5355\u8bcd\nword = &quot;hello&quot;\n\n# \u83b7\u53d6\u5355\u8bcd\u5bf9\u5e94\u7684\u8bcd\u5143\uff08lexeme\uff09\nlexeme = nlp.vocab[word]\n\n# \u6253\u5370\u8bcd\u5143\u7684\u6587\u672c\u5185\u5bb9\u3001\u662f\u5426\u4e3a\u5b57\u6bcd\uff08alphabetical\uff09\n# \u662f\u5426\u4e3a\u505c\u7528\u8bcd\uff08stopword\uff09\\\u662f\u5426\u4e3a\u5b57\u6bcd\uff08is_alpha\uff09\uff0c\u662f\u5426\u4e3a\u6570\u5b57\uff08is_digit\uff09\uff0c\u662f\u5426\u4e3a\u6807\u9898\uff08is_title\uff09\uff0c\u8bed\u8a00\uff08lang_\uff09\nprint(lexeme.text, lexeme.is_alpha, lexeme.is_stop, lexeme.is_alpha, lexeme.is_digit, lexeme.is_title, lexeme.lang_)\n\n&quot;&quot;&quot;\nhello True False True False False en\n&quot;&quot;&quot;<\/code><\/pre>\n<p>\u4e8b\u5b9e\u4e0a\u5bf9\u4e8eLexeme\uff0c\u53ea\u8981\u53ef\u80fd\uff0cspaCy\u5c31\u4f1a\u5c1d\u8bd5\u5c06\u6570\u636e\u5b58\u50a8\u5728\u4e00\u4e2a\u8bcd\u6c47\u8868Vocab\u4e2d\uff0c\u8be5\u8bcd\u6c47\u8868\u5c06\u7531\u591a\u4e2a\u6a21\u578b\u5171\u4eab\u3002\u4e3a\u4e86\u8282\u7701\u5185\u5b58\uff0cspaCy\u8fd8\u5c06\u6240\u6709\u5b57\u7b26\u4e32\u7f16\u7801\u4e3a\u54c8\u5e0c\u503c\u3002<\/p>\n<p><img src=\"https:\/\/lemon-guess.oss-cn-hangzhou.aliyuncs.com\/img\/b1a0cecb0d0a4d82acbdec0a782c60ad-1705586515523-28.png\" alt=\"\u5728\u8fd9\u91cc\u63d2\u5165\u56fe\u7247\u63cf\u8ff0\" \/><\/p>\n<p>\u5982\u4e0b\u6240\u793a\uff0c\u4e0d\u540c\u6a21\u578b\u4e0b\u201ccoffee\u201d\u7684\u54c8\u5e0c\u503c\u4e3a3197928453018144401\u3002\u4f46\u662f\u6ce8\u610f\u7684\u662f\u53ea\u662fspaCy\u8fd9\u6837\u505a\uff0c\u5176\u4ed6\u81ea\u7136\u8bed\u8a00\u5904\u7406\u5e93\u4e0d\u4e00\u5b9a\u8fd9\u6837\u505a\u3002<\/p>\n<pre><code class=\"language-python\">import spacy\n\nnlp = spacy.load(&quot;zh_core_web_sm&quot;)\ndoc = nlp(&quot;I love coffee&quot;)\nprint(doc.vocab.strings[&quot;coffee&quot;])  # 3197928453018144401\nprint(doc.vocab.strings[3197928453018144401])  # &#039;coffee&#039;\n\n&quot;&quot;&quot;\n3197928453018144401\ncoffee\n&quot;&quot;&quot;\n\nimport spacy\n\nnlp = spacy.load(&quot;en_core_web_sm&quot;)\ndoc = nlp(&quot;I love coffee&quot;)\nprint(doc.vocab.strings[&quot;coffee&quot;])  # 3197928453018144401\nprint(doc.vocab.strings[3197928453018144401])  # &#039;coffee&#039;\n\n&quot;&quot;&quot;\n3197928453018144401\ncoffee\n&quot;&quot;&quot;<\/code><\/pre>\n<p><strong>Span<\/strong><\/p>\n<p>Span \u662f\u4e00\u4e2a\u8fde\u7eed\u7684\u6587\u672c\u7247\u6bb5\uff0c\u53ef\u4ee5\u7531\u4e00\u4e2a\u6216\u591a\u4e2aToken\u7ec4\u6210\u3002\u5b83\u901a\u5e38\u7528\u4e8e\u6807\u8bb0\u5b9e\u4f53\u6216\u77ed\u8bed\u3002<\/p>\n<pre><code class=\"language-python\">nlp = spacy.load(&quot;zh_core_web_sm&quot;)\ntext = &quot;\u4e1c\u65b9\u660e\u73e0\u662f\u4e00\u5ea7\u4f4d\u4e8e\u4e2d\u56fd\u4e0a\u6d77\u5e02\u7684\u6807\u5fd7\u6027\u5efa\u7b51\uff01&quot;\ndoc = nlp(text)\n\n# \u4ecedoc\u4e2d\u9009\u62e9\u4e86\u7b2c0\u4e2a\u548c\u7b2c1\u4e2a\u8bcd\u5143\uff08token\uff09\u7ec4\u6210\u7684\u7247\u6bb5\u3002\n# \u6ce8\u610f\uff0cspaCy\u4e2d\u7684\u8bcd\u5143\u662f\u6587\u672c\u7684\u57fa\u672c\u5355\u5143\uff0c\u53ef\u80fd\u662f\u5355\u8bcd\u3001\u6807\u70b9\u7b26\u53f7\u6216\u5176\u5b83\u8bcd\u6c47\u5355\u4f4d\u3002\n# \u8fd9\u91cc\u4e1c\u65b9\u3001\u660e\u73e0\u662f\u524d\u4e24\u4e2a\u8bcd\nspan = doc[0:2]  \nprint(span.text)\n\n&quot;&quot;&quot;\n\u4e1c\u65b9\u660e\u73e0\n&quot;&quot;&quot;<\/code><\/pre>\n<h2>4 \u53c2\u8003<\/h2>\n<ul>\n<li><a href=\"https:\/\/github.com\/explosion\/spaCy\">spaCy-github<\/a><\/li>\n<li><a href=\"https:\/\/spacy.io\/\">spaCy<\/a><\/li>\n<li><a href=\"https:\/\/spacy.io\/usage\/models\">spaCy-models<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/huggingface\/transformers\">Transformers<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/PaddlePaddle\/PaddleNLP\">PaddleNLP<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/nltk\/nltk\">NLTK<\/a><\/li>\n<li><a href=\"https:\/\/spacy.io\/usage\">spaCy-usage<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/explosion\/sense2vec\">sense2vec<\/a><\/li>\n<li><a href=\"https:\/\/spacy.io\/usage\/training\">spaCy-training<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>1 \u80cc\u666f\u4ecb\u7ecd\u4e0espaCy\u5b89\u88c5 spaCy\u662f\u4e00\u4e2a\u57fa\u4e8ePython\u7f16\u5199\u7684\u5f00\u6e90\u81ea\u7136\u8bed\u8a00\u5904\u7406\u5e93\u3002\u57fa\u4e8e\u81ea\u7136\u5904\u7406\u9886\u57df\u7684\u6700\u65b0\u7814\u7a76\uff0cspaCy\u63d0\u4f9b\u4e86\u4e00\u7cfb   \u2026 &#8230;<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ngg_post_thumbnail":0},"categories":[225],"tags":[],"_links":{"self":[{"href":"\/index.php?rest_route=\/wp\/v2\/posts\/9106"}],"collection":[{"href":"\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=9106"}],"version-history":[{"count":1,"href":"\/index.php?rest_route=\/wp\/v2\/posts\/9106\/revisions"}],"predecessor-version":[{"id":9107,"href":"\/index.php?rest_route=\/wp\/v2\/posts\/9106\/revisions\/9107"}],"wp:attachment":[{"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=9106"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=9106"},{"taxonomy":"post_tag","embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=9106"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}