{"id":1955,"date":"2024-05-30T14:57:01","date_gmt":"2024-05-30T06:57:01","guid":{"rendered":"http:\/\/blog.xtaa.cn\/?p=1955"},"modified":"2024-05-30T14:57:01","modified_gmt":"2024-05-30T06:57:01","slug":"%e4%ba%ba%e5%b7%a5%e6%99%ba%e8%83%bd%e6%96%87%e6%9c%ac%e5%a4%84%e7%90%86%e7%aa%81%e7%a0%b4-%e4%bb%8eone-hot%e5%88%b0embedding%e6%a8%a1%e5%9e%8b","status":"publish","type":"post","link":"http:\/\/blog.xtaa.cn\/index.php\/2024\/05\/30\/%e4%ba%ba%e5%b7%a5%e6%99%ba%e8%83%bd%e6%96%87%e6%9c%ac%e5%a4%84%e7%90%86%e7%aa%81%e7%a0%b4-%e4%bb%8eone-hot%e5%88%b0embedding%e6%a8%a1%e5%9e%8b\/","title":{"rendered":"\u4eba\u5de5\u667a\u80fd\u6587\u672c\u5904\u7406\u7a81\u7834-\u4eceone-hot\u5230embedding\u6a21\u578b"},"content":{"rendered":"\n<p>\u6587\u672c\u5411\u91cf\u5316\u662f\u5c06\u6587\u672c\u6570\u636e\u8f6c\u6362\u4e3a\u6570\u503c\u5411\u91cf\u7684\u8fc7\u7a0b\u3002\u8fd9\u79cd\u8f6c\u6362\u4f7f\u5f97\u8ba1\u7b97\u673a\u80fd\u591f\u7406\u89e3\u548c\u5904\u7406\u6587\u672c\u6570\u636e\uff0c\u4ece\u800c\u53ef\u4ee5\u5e94\u7528\u5404\u79cd\u673a\u5668\u5b66\u4e60\u548c\u81ea\u7136\u8bed\u8a00\u5904\u7406\u6280\u672f\u3002\u6587\u672c\u5411\u91cf\u5316\u7684\u4e3b\u8981\u95ee\u9898\u662f\u5c06\u975e\u7ed3\u6784\u5316\u7684\u6587\u672c\u6570\u636e\u8f6c\u6362\u4e3a\u7ed3\u6784\u5316\u7684\u5411\u91cf\u8868\u793a\uff0c\u4ee5\u4fbf\u8ba1\u7b97\u673a\u80fd\u591f\u5bf9\u5176\u8fdb\u884c\u5904\u7406\u548c\u5206\u6790\u3002\u901a\u8fc7\u6587\u672c\u5411\u91cf\u5316\uff0c\u53ef\u4ee5\u5b9e\u73b0\u6587\u672c\u5206\u7c7b\u3001\u6587\u672c\u805a\u7c7b\u3001\u4fe1\u606f\u68c0\u7d22\u7b49\u4efb\u52a1\u3002<\/p>\n\n\n\n<p>\u5728\u5927\u6a21\u578b\u65f6\u4ee3\u4e0b\uff0c\u6587\u672c\u5411\u91cf\u5316\u7684\u6280\u672f\u53d1\u5c55\u5230\u4e86embedding\u9636\u6bb5\u3002Embedding\u662f\u4e00\u79cd\u5c06\u9ad8\u7ef4\u7684\u6587\u672c\u6570\u636e\u6620\u5c04\u5230\u4f4e\u7ef4\u7a7a\u95f4\u7684\u6280\u672f\uff0c\u5b83\u80fd\u591f\u6355\u6349\u5230\u6587\u672c\u6570\u636e\u4e2d\u7684\u8bed\u4e49\u4fe1\u606f\uff0c\u5e76\u5c06\u5176\u8868\u793a\u4e3a\u7a20\u5bc6\u5411\u91cf\u3002\u8fd9\u79cd\u8868\u793a\u66f4\u5177\u6709\u8bed\u4e49\u4fe1\u606f\uff0c\u76f8\u6bd4\u4e8e\u4f20\u7edf\u7684\u8bcd\u888b\u6a21\u578b\u6216\u8005TF-IDF\u7b49\u8868\u793a\u65b9\u6cd5\uff0cembedding\u80fd\u591f\u66f4\u597d\u5730\u6355\u6349\u5230\u5355\u8bcd\u4e4b\u95f4\u7684\u8bed\u4e49\u5173\u7cfb\u3002<\/p>\n\n\n\n<p><strong>\u6587\u672c\u5411\u91cf\u5316\u6280\u672f\u53d1\u5c55\u53f2<\/strong><\/p>\n\n\n\n<p>\u6574\u4e2a\u6587\u672c\u5411\u91cf\u5316\u7684\u53d1\u5c55\u8fc7\u7a0b\u662f\u4e00\u4e2a\u7531\u7b80\u5355\u5230\u590d\u6742\uff0c\u7531\u4f4e\u7ef4\u5230\u9ad8\u7ef4\u7684\u6f14\u53d8\u8fc7\u7a0b\u3002\u5728\u65e9\u671f\uff0c\u4eba\u4eec\u4f7f\u7528\u57fa\u4e8e\u7edf\u8ba1\u7684\u65b9\u6cd5\uff0c\u5982\u8bcd\u888b\u6a21\u578b\u3001TF-IDF\u7b49\u6765\u8868\u793a\u6587\u672c\u3002\u8fd9\u4e9b\u4f20\u7edf\u7684\u57fa\u4e8e\u7edf\u8ba1\u7684\u65b9\u6cd5\u867d\u7136\u80fd\u591f\u5b9e\u73b0\u4e00\u4e9b\u7b80\u5355\u7684\u6587\u672c\u5904\u7406\u548c\u81ea\u7136\u8bed\u8a00\u7406\u89e3\u4efb\u52a1\uff0c\u4f46\u662f\u5bf9\u4e8e\u8bed\u4e49\u4fe1\u606f\u7684\u628a\u63e1\u80fd\u529b\u6709\u9650\uff0c\u800c\u4e14\u5f88\u96be\u5904\u7406\u590d\u6742\u7684\u8bed\u8a00\u73b0\u8c61\u3002\u968f\u7740\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u7684\u6f14\u53d8\uff0c\u5c24\u5176\u662fWord2Vec\u3001GloVe\u7b49\u8bcd\u5411\u91cf\u6a21\u578b\u7684\u63d0\u51fa\uff0c\u6587\u672c\u5904\u7406\u7684\u6548\u679c\u548c\u6027\u80fd\u5927\u5e45\u63d0\u5347\uff0c\u6587\u672c\u5411\u91cf\u5316\u6280\u672f\u5f00\u59cb\u5411embedding\u65b9\u5411\u53d1\u5c55\u3002\u800c\u540e\u6765\u7684BERT\u3001GPT\u7b49\u5927\u578b\u9884\u8bad\u7ec3\u6a21\u578b\u7684\u51fa\u73b0\uff0c\u5219\u8fdb\u4e00\u6b65\u63a8\u52a8\u4e86\u6587\u672c\u5411\u91cf\u5316\u6280\u672f\u7684\u8fdb\u6b65\uff0c\u4f7f\u5f97\u6587\u672c\u5411\u91cf\u5316\u4e0d\u4ec5\u80fd\u591f\u8868\u793a\u5355\u8bcd\u6216\u77ed\u8bed\uff0c\u8fd8\u80fd\u591f\u8868\u793a\u6574\u4e2a\u53e5\u5b50\u6216\u6587\u6863\u7684\u8bed\u4e49\u4fe1\u606f\u3002\u65b0\u6280\u672f\u751f\u6210\u7684\u6587\u672c\u5411\u91cf\u4e0d\u4ec5\u53ef\u4ee5\u9ad8\u6548\u5b8c\u6210\u4f20\u7edf\u7684\u81ea\u7136\u8bed\u8a00\u5904\u7406\u4efb\u52a1\uff0c\u8fd8\u5728\u65b0\u5174\u7684\u751f\u6210\u5f0f\u4eba\u5de5\u667a\u80fd\u6280\u672f\uff08\u5982\u68c0\u7d22\u589e\u5f3a\u751f\u6210\u7b49\u6280\u672f\uff09\u4e2d\u53d1\u6325\u4e86\u4f5c\u7528\u3002<\/p>\n\n\n\n<p><strong>One-Hot\u7f16\u7801\uff08\u51fa\u73b0\u4e8e2000\u5e74\u524d\uff09<\/strong><\/p>\n\n\n\n<p><strong>One-Hot\u7f16\u7801\u662f\u4e00\u79cd\u5c06\u5206\u7c7b\u6570\u636e\u8f6c\u6362\u4e3a\u4e8c\u8fdb\u5236\u8868\u793a\u7684\u65b9\u6cd5\uff0c\u5e38\u7528\u4e8e\u6570\u636e\u5904\u7406\u548c\u673a\u5668\u5b66\u4e60\u3002One-Hot\u7f16\u7801\u7684\u6982\u5ff5\u4e00\u822c\u88ab\u8ba4\u4e3a\u662f\u968f\u7740\u8ba1\u7b97\u673a\u79d1\u5b66\u7684\u53d1\u5c55\u800c\u81ea\u7136\u51fa\u73b0\u3002<\/strong>\u5728\u6570\u5b57\u8ba1\u7b97\u548c\u6570\u636e\u5904\u7406\u7684\u65e9\u671f\u9636\u6bb5\uff0c\u4e8c\u8fdb\u5236\u8868\u793a\u6781\u4e3a\u76db\u884c\u3002\u5f53\u9700\u8981\u8ba1\u7b97\u673a\u5904\u7406\u975e\u6570\u503c\u7684\u6570\u636e\uff08\u5982\u989c\u8272\uff0c\u4ea7\u54c1\u7c7b\u578b\u7b49\uff09\u65f6\uff0cOne-Hot\u7f16\u7801\u5c31\u662f\u4e00\u79cd\u5c06\u5404\u7c7b\u6570\u636e\u6570\u503c\u5316\u7684\u6734\u7d20\u65b9\u6cd5\u3002<\/p>\n\n\n\n<p>\u5728\u6587\u672c\u5411\u91cf\u5316\u4e2d\uff0cOne-Hot\u7f16\u7801\u7684\u5206\u7c7b\u6570\u636e\u4e00\u822c\u4e3a\u5355\u8bcd\u6216\u5b57\u7b26\uff0c\u4ee5\u4e0b\u4ee5\u5355\u8bcd\u4e3a\u4f8b\u3002\u901a\u8fc7\u8fd9\u79cd\u7f16\u7801\u65b9\u5f0f\uff0c\u6bcf\u4e2a\u552f\u4e00\u7684\u5355\u8bcd\u90fd\u7528\u4e00\u4e2a\u5411\u91cf\u8868\u793a\u3002\u5411\u91cf\u7684\u957f\u5ea6\u7b49\u4e8e\u8bcd\u6c47\u8868\u7684\u5355\u8bcd\u6570\u91cf\uff0c\u5176\u4e2d\u53ea\u6709\u4e00\u4e2a\u4f4d\u7f6e\u4e3a1\uff0c\u5176\u4f59\u4f4d\u7f6e\u90fd\u4e3a0\u3002\u8fd9\u79cd\u8868\u8fbe\u65b9\u5f0f\u751f\u6210\u7684\u662f\u7a00\u758f\u5411\u91cf\uff0c\u5176\u4e2d\u6bcf\u4e2a\u5355\u8bcd\u90fd\u7531\u4e00\u4e2a\u552f\u4e00\u7684\u4e8c\u8fdb\u5236\u5411\u91cf\u8868\u793a\uff0c\u8be5\u5411\u91cf\u4e2d\u53ea\u6709\u4e00\u4e2a\u9ad8\u4f4d\uff081\uff09\uff0c\u5176\u4ed6\u90fd\u662f\u4f4e\u4f4d\uff080\uff09\u3002<\/p>\n\n\n\n<p>\u4ee5\u201c\u732b\u201d\u3001\u201c\u72d7\u201d\u3001\u201c\u9c7c\u201d\u4e3a\u4f8b\uff0cOne-Hot\u7f16\u7801\u5728\u6587\u672c\u5411\u91cf\u5316\u4e2d\u7684\u5b9e\u73b0\u6b65\u9aa4\u5927\u81f4\u5982\u4e0b\uff1a<\/p>\n\n\n\n<p>\u9996\u5148\u8bc6\u522b\u6240\u6709\u7684\u552f\u4e00\u5355\u8bcd\uff0c\u5e76\u5efa\u7acb\u957f\u5ea6\u4e3a3\u7684\u8bcd\u6c47\u8868{\u201c\u732b\u201d, \u201c\u72d7\u201d, \u201c\u9c7c\u201d}\u3002\u5176\u6b21\u4e3a\u6bcf\u4e2a\u5355\u8bcd\u5206\u914d\u7d22\u5f15\uff0c\u6bd4\u5982\u201c\u732b\u201d\u3001\u201c\u72d7\u201d\u3001\u201c\u9c7c\u201d\u5206\u522b\u5206\u914d\u7d22\u5f15\u4f4d\u7f6e0\uff0c1\uff0c2\u3002\u7136\u540e\u4e3a\u6bcf\u4e2a\u5355\u8bcd\u521b\u5efa\u5411\u91cf\uff0c\u4f8b\u5982\uff0c\u201c\u732b\u201d\u7684\u5411\u91cf\u662f[1, 0, 0]\u3002\u6700\u540e\u5047\u4f7f\u6709\u4e00\u6bb5\u6587\u672c\u5185\u5bb9\u4e3a\u201c\u732b\u732b\u72d7\u72d7\u201d\uff0c\u5b83\u7ecf\u8fc7\u5411\u91cf\u5316\u540e\u5c31\u662f<\/p>\n\n\n\n<p>[[1, 0, 0],<\/p>\n\n\n\n<p>[1, 0, 0],<\/p>\n\n\n\n<p>[0, 1, 0],<\/p>\n\n\n\n<p>[0, 1, 0]]\u3002<\/p>\n\n\n\n<p>\u4f18\u70b9\uff1a<\/p>\n\n\n\n<p>\u25cf \u7b80\u5355\u6613\u5b9e\u73b0\u3002<\/p>\n\n\n\n<p>\u25cf \u80fd\u63d0\u4f9b\u6e05\u6670\u660e\u786e\u7684\u5355\u8bcd\u8868\u793a\u3002<\/p>\n\n\n\n<p>\u7f3a\u70b9\uff1a<\/p>\n\n\n\n<p>\u25cf \u8bcd\u6c47\u8868\u504f\u5927\u65f6\u4f1a\u5bfc\u81f4\u5411\u91cf\u7ef4\u5ea6\u8fc7\u9ad8\uff0c\u4ece\u800c\u9020\u6210\u5185\u5b58\u4f7f\u7528\u91cf\u504f\u5927\u3002<\/p>\n\n\n\n<p>\u25cf \u65e0\u6cd5\u5b58\u50a8\u5355\u8bcd\u4e4b\u95f4\u7684\u8bed\u4e49\u76f8\u4f3c\u6027\uff08\u4f8b\u5982\uff0c\u201d\u732b\u201c\u548c\u201d\u72d7\u201c\u540c\u4e3a\u5ba0\u7269\u52a8\u7269\u5e94\u66f4\u52a0\u76f8\u4f3c\uff0c\u7136\u800c\u5728One-Hot\u7f16\u7801\u4e2d\u201c\u732b\u201d\u3001\u201c\u72d7\u201d\u3001\u201c\u9c7c\u201d\u4e09\u8005\u7684\u5411\u91cf\u8ddd\u79bb\u76f8\u540c\uff0c\u6240\u4ee5\u4e0d\u80fd\u4f53\u73b0\u8fd9\u4e00\u7279\u6027\uff09\u3002<\/p>\n\n\n\n<p>\u25cf \u7a00\u758f\u5411\u91cf\u5728\u67d0\u4e9b\u4efb\u52a1\u4e2d\u8ba1\u7b97\u6548\u7387\u4f4e\u3002<\/p>\n\n\n\n<p><strong>Bag of words \uff08\u51fa\u73b0\u4e8e2000\u5e74\u524d\uff09<\/strong><\/p>\n\n\n\n<p><strong>Bag of Words\uff08BoW\uff09\u53c8\u79f0\u8bcd\u888b\u6a21\u578b\uff0c\u662f\u4e00\u79cd\u7b80\u5355\u800c\u6709\u6548\u7684\u6587\u672c\u5411\u91cf\u5316\u65b9\u6cd5\u3002\u8bcd\u888b\u6a21\u578b\u76ee\u524d\u6ca1\u6709\u5f52\u5c5e\u4e00\u4e2a\u5177\u4f53\u7684\u53d1\u660e\u8005\uff0c\u800c\u662f\u5728\u4fe1\u606f\u68c0\u7d22\u3001\u6587\u672c\u5206\u7c7b\u548c\u81ea\u7136\u8bed\u8a00\u5904\u7406\u7b49\u9886\u57df\u4e2d\u9010\u6e10\u53d1\u5c55\u6f14\u53d8\u51fa\u6765\u7684\u4e00\u79cd\u5e38\u7528\u7684\u6587\u672c\u8868\u793a\u65b9\u6cd5\u3002<\/strong>One-Hot\u7f16\u7801\u867d\u7136\u5b8c\u6210\u4e86\u6570\u636e\u4ece\u975e\u6570\u503c\u5230\u6570\u503c\u7684\u8f6c\u53d8\uff0c\u4f46\u5728\u6587\u672c\u5206\u7c7b\uff0c\u7279\u5f81\u63d0\u53d6\u7b49\u65b9\u9762\u4f5c\u7528\u5341\u5206\u6709\u9650\uff0c\u4e14\u6bcf\u4e2a\u8bcd\u90fd\u5bf9\u5e94\u4e00\u4e2a\u7a00\u758f\u5411\u91cf\uff0c\u5b58\u50a8\u6548\u7387\u8f83\u4f4e\u3002\u8bcd\u888b\u6a21\u578b\u5c06\u6587\u672c\u770b\u4f5c\u662f\u5355\u8bcd\u7684\u65e0\u5e8f\u96c6\u5408\uff0c\u5ffd\u7565\u8bcd\u7684\u987a\u5e8f\u548c\u8bed\u6cd5\uff0c\u53ea\u5173\u6ce8\u8bcd\u7684\u51fa\u73b0\u9891\u7387\u3002\u56e0\u6b64\u8bcd\u888b\u6a21\u578b\u5728One-Hot\u7f16\u7801\u7684\u57fa\u7840\u4e0a\u63d0\u5347\u4e86\u6587\u672c\u5411\u91cf\u5316\u7684\u6548\u679c\uff0c\u5e76\u51cf\u5c11\u4e86\u5b58\u50a8\u8d1f\u62c5\u3002<\/p>\n\n\n\n<p>\u4ecd\u4ee5\u201c\u732b\u201d\u3001\u201c\u72d7\u201d\u3001\u201c\u9c7c\u201d\u4e3a\u4f8b\uff0c\u8bcd\u888b\u6a21\u578b\u5728\u6587\u672c\u5411\u91cf\u5316\u4e2d\u7684\u5b9e\u73b0\u6b65\u9aa4\u5927\u81f4\u5982\u4e0b\uff1a<\/p>\n\n\n\n<p>\u4e0eOne-Hot\u7f16\u7801\u76f8\u540c\uff0c\u9996\u5148\u8bc6\u522b\u6240\u6709\u7684\u552f\u4e00\u5355\u8bcd\uff0c\u5e76\u5efa\u7acb\u957f\u5ea6\u4e3a3\u7684\u8bcd\u6c47\u8868{\u201c\u732b\u201d, \u201c\u72d7\u201d, \u201c\u9c7c\u201d}\u3002\u5176\u6b21\u5bf9\u4e8e\u6bcf\u4e00\u6587\u672c\uff0c\u521b\u5efa\u4e00\u4e2a\u4e0e\u8bcd\u6c47\u8868\u7b49\u957f\u7684\u5411\u91cf\u3002\u5411\u91cf\u7684\u6bcf\u4e2a\u4f4d\u7f6e\u5bf9\u5e94\u8bcd\u6c47\u8868\u4e2d\u7684\u4e00\u4e2a\u8bcd\uff0c\u8be5\u4f4d\u7f6e\u7684\u503c\u8868\u793a\u8be5\u8bcd\u5728\u6587\u6863\u4e2d\u51fa\u73b0\u7684\u6b21\u6570\u3002\u5047\u4f7f\u201c\u732b\u201d\u3001\u201c\u72d7\u201d\u3001\u201c\u9c7c\u201d\u7684\u7d22\u5f15\u4f4d\u7f6e\u4ecd\u7136\u5206\u522b\u4e3a0\uff0c1\uff0c2\uff0c\u800c\u6587\u672c\u4f9d\u7136\u662f\u201c\u732b\u732b\u72d7\u72d7\u201d\u3002\u4e0eone-hot\u7f16\u7801\u4f1a\u751f\u6210\u4e00\u7ec4\u5411\u91cf\u4e0d\u540c\uff0c\u8bcd\u888b\u6a21\u578b\u53ea\u4f1a\u4ea7\u751f\u4e00\u4e2a\u5411\u91cf[2, 2, 0]\u3002\u5411\u91cf\u4e2d\u7b2c0\u4e2a\u5143\u7d202\u4ee3\u8868\u201c\u732b\u201d\u51fa\u73b0\u4e86\u4e24\u6b21\uff0c\u7b2c1\u4e2a\u5143\u7d202\u4ee3\u8868\u201c\u72d7\u201d\u51fa\u73b0\u4e86\u4e24\u6b21\uff0c\u7b2c2\u4e2a\u5143\u7d200\u4ee3\u8868\u201c\u9c7c\u201d\u51fa\u73b0\u4e86\u96f6\u6b21\u3002<\/p>\n\n\n\n<p>\u4f18\u70b9\uff1a<\/p>\n\n\n\n<p>\u25cf \u7b80\u5355\u6613\u5b9e\u73b0\u3002<\/p>\n\n\n\n<p>\u25cf \u5bf9\u4e8e\u5c0f\u89c4\u6a21\u6570\u636e\u96c6\uff0c\u8bcd\u888b\u6a21\u578b\u7684\u8ba1\u7b97\u6548\u7387\u5f88\u9ad8\u3002<\/p>\n\n\n\n<p>\u7f3a\u70b9\uff1a<\/p>\n\n\n\n<p>\u25cf \u5bf9\u4e8e\u5927\u578b\u8bcd\u6c47\u8868\uff0c\u8bcd\u888b\u6a21\u578b\u751f\u6210\u7684\u5411\u91cf\u7ef4\u5ea6\u5f88\u9ad8\uff0c\u4ece\u800c\u5bfc\u81f4\u5185\u5b58\u4f7f\u7528\u91cf\u5927\u3002<\/p>\n\n\n\n<p>\u25cf \u5b8c\u5168\u5ffd\u7565\u4e86\u8bcd\u7684\u987a\u5e8f\uff0c\u65e0\u6cd5\u6355\u6349\u53e5\u5b50\u7684\u8bed\u6cd5\u548c\u7ed3\u6784\u4fe1\u606f\u3002<\/p>\n\n\n\n<p>\u25cf \u4ec5\u4ec5\u7edf\u8ba1\u8bcd\u9891\uff0c\u800c\u65e0\u6cd5\u6355\u6349\u8bcd\u4e0e\u8bcd\u4e4b\u95f4\u7684\u8bed\u4e49\u5173\u7cfb\u3002<\/p>\n\n\n\n<p><strong>TF-IDF\uff08\u51fa\u73b0\u4e8e2000\u5e74\u524d\uff09<\/strong><\/p>\n\n\n\n<p>TF-IDF\uff08Term Frequency-Inverse Document Frequency\uff09\uff0c\u7ffb\u8bd1\u4e3a\u8bcd\u9891-\u9006\u6587\u6863\u9891\u7387\uff0c\u662f\u4e00\u79cd\u5e38\u7528\u4e8e\u6587\u672c\u6316\u6398\u548c\u4fe1\u606f\u68c0\u7d22\u7684\u7edf\u8ba1\u65b9\u6cd5\uff0c\u7528\u6765\u8bc4\u4f30\u4e00\u4e2a\u8bcd\u5728\u4e00\u4e2a\u6587\u6863\u96c6\u6216\u8bed\u6599\u5e93\u4e2d\u7684\u91cd\u8981\u7a0b\u5ea6\u3002\u73b0\u4eca\u7684TF-IDF\u65b9\u6cd5\u88ab\u8ba4\u4e3a\u662f\u6587\u672c\u6316\u6398\u548c\u4fe1\u606f\u68c0\u7d22\u9886\u57df\u9010\u6e10\u53d1\u5c55\u548c\u5b8c\u5584\u7684\u7ed3\u679c\uff0c\u4f46\u9006\u6587\u6863\u9891\u7387\u7684\u57fa\u672c\u6982\u5ff5\u7531\u8ba1\u7b97\u673a\u79d1\u5b66\u5bb6Karen Sp\u00e4rck Jones\u57281957\u5e74\u63d0\u51fa\uff0c\u7528\u6765\u8bc4\u4f30\u5728\u4e00\u4e2a\u6587\u6863\u96c6\u4e2d\uff0c\u4e00\u4e2a\u8bcd\u5bf9\u5b83\u6240\u5728\u7684\u6587\u6863\u7684\u91cd\u8981\u6027\u3002TF-IDF\u53ef\u4ee5\u5e2e\u52a9\u7b5b\u9009\u51fa\u5728\u6587\u6863\u4e2d\u5177\u6709\u91cd\u8981\u610f\u4e49\u7684\u8bcd\u6c47\uff0c\u5e38\u7528\u4e8e\u5173\u952e\u8bcd\u63d0\u53d6\u3001\u6587\u6863\u76f8\u4f3c\u5ea6\u8ba1\u7b97\u548c\u6587\u672c\u5206\u7c7b\u7b49\u4efb\u52a1\u3002<\/p>\n\n\n\n<p>TF-IDF\u7531\u4e24\u4e2a\u90e8\u5206\u7ec4\u6210\uff1a\u8bcd\u9891\uff08TF\uff09\u548c\u9006\u6587\u6863\u9891\u7387\uff08IDF\uff09\u3002\u8bcd\u9891\u662f\u4e00\u4e2a\u8bcd\u5728\u4e00\u4e2a\u6587\u6863\u4e2d\u51fa\u73b0\u7684\u9891\u7387\u3002\u8bcd\u9891\u8d8a\u9ad8\uff0c\u8868\u660e\u8fd9\u4e2a\u8bcd\u5728\u8be5\u6587\u6863\u4e2d\u8d8a\u91cd\u8981\u3002\u8bcd\u9891\u7684\u516c\u5f0f\u4e3a\uff1a<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/p3-sign.toutiaoimg.com\/tos-cn-i-6w9my0ksvp\/7d79b85867474537afe2634134d2303a~noop.image?_iz=58558&amp;from=article.pc_detail&amp;lk3s=953192f4&amp;x-expires=1717655407&amp;x-signature=Wkz%2BLJgqstBs5QW11p8Qyk9Bzmk%3D\" alt=\"\"\/><\/figure>\n\n\n\n<p>\u9006\u6587\u6863\u9891\u7387\u8861\u91cf\u4e00\u4e2a\u8bcd\u5728\u6574\u4e2a\u6587\u6863\u96c6\uff08\u53ef\u80fd\u5305\u542b\u591a\u4e2a\u6587\u6863\uff09\u4e2d\u51fa\u73b0\u7684\u7a00\u6709\u7a0b\u5ea6\u3002\u4e00\u4e2a\u8bcd\u5728\u8d8a\u591a\u6587\u6863\u4e2d\u51fa\u73b0\uff0c\u5176\u533a\u5206\u80fd\u529b\u8d8a\u4f4e\uff0cIDF\u503c\u5c31\u8d8a\u5c0f\u3002\u9006\u6587\u6863\u9891\u7387\u7684\u8ba1\u7b97\u516c\u5f0f\u4e3a\uff1a<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/p3-sign.toutiaoimg.com\/tos-cn-i-6w9my0ksvp\/4002d5f87b104da9b3feb5dda18fed58~noop.image?_iz=58558&amp;from=article.pc_detail&amp;lk3s=953192f4&amp;x-expires=1717655407&amp;x-signature=Sy%2FXiHefDvkCvnrdZn3hJlH3PjM%3D\" alt=\"\"\/><\/figure>\n\n\n\n<p>\u4e24\u4e2a\u516c\u5f0f\u4e2d\uff1a<\/p>\n\n\n\n<p>ft,d\u662f\u8bcdt\u5728\u6587\u6863d\u4e2d\u51fa\u73b0\u7684\u6b21\u6570\u3002<\/p>\n\n\n\n<p>nd\u662f\u6587\u6863d\u4e2d\u6240\u6709\u8bcd\u7684\u603b\u6570\u3002<\/p>\n\n\n\n<p>N\u662f\u6587\u6863\u96c6D\u4e2d\u6587\u6863\u7684\u603b\u6570\u3002<\/p>\n\n\n\n<p>|{d\u2208D:t\u2208d}|\u662f\u5305\u542b\u8bcdt\u7684\u6587\u6863\u603b\u6570\u3002<\/p>\n\n\n\n<p>\u6700\u540e\uff0cTF-IDF \u662f\u8bcd\u9891\u548c\u9006\u6587\u6863\u9891\u7387\u7684\u4e58\u79ef\uff0c\u7528\u6765\u8861\u91cf\u4e00\u4e2a\u8bcd\u5728\u4e00\u4e2a\u6587\u6863\u4e2d\u7684\u91cd\u8981\u6027\uff0c\u5176\u8ba1\u7b97\u516c\u5f0f\u4e3a\uff1a<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/p3-sign.toutiaoimg.com\/tos-cn-i-6w9my0ksvp\/964ae5dbbdb24f068ccbbfc631ecf545~noop.image?_iz=58558&amp;from=article.pc_detail&amp;lk3s=953192f4&amp;x-expires=1717655407&amp;x-signature=gpqeMIm%2B0WWrNneIngDSeSHOtds%3D\" alt=\"\"\/><\/figure>\n\n\n\n<p>\u4e3e\u4f8b\u6765\u8bf4\uff0c\u5047\u8bbe\u6709\u4ee5\u4e0b\u4e09\u4e2a\u6587\u6863\uff1a<\/p>\n\n\n\n<p>\u6587\u68631\uff1a \u201d\u72d7\u8ffd\u732b\u3002\u201c<\/p>\n\n\n\n<p>\u6587\u68632\uff1a &#8220;\u732b\u722c\u5899\u3002&#8221;<\/p>\n\n\n\n<p>\u6587\u68633\uff1a &#8220;\u72d7\u8db4\u5730\u3002&#8221;<\/p>\n\n\n\n<p>TF-IDF(&#8220;\u732b&#8221;,\u6587\u68631, \u6587\u6863\u96c6)\u7684\u8ba1\u7b97\u5982\u4e0b\uff1a<\/p>\n\n\n\n<p>TF(&#8220;\u732b&#8221;,\u6587\u68631) = 1\/3\u3002<\/p>\n\n\n\n<p>IDF(&#8220;\u732b&#8221;,\u6587\u6863\u96c6) = log(3 \/ (1 + 2)) = 0\u3002<\/p>\n\n\n\n<p>TF-IDF(&#8220;\u732b&#8221;,\u6587\u68631,\u6587\u6863\u96c6) = 1\/3 * 0 = 0\u3002<\/p>\n\n\n\n<p>\u53ef\u4ee5\u770b\u51fa\uff0c\u7531\u4e8e\u201c\u732b\u201d \u5728\u591a\u4e2a\u6587\u6863\u4e2d\u51fa\u73b0\uff0c\u56e0\u6b64\u5176\u533a\u5206\u80fd\u529b\u8f83\u4f4e\uff0cIDF\u503c\u4e3a0\uff0c\u5bfc\u81f4TF-IDF\u503c\u4e5f\u4e3a0\u3002<\/p>\n\n\n\n<p>\u4f18\u70b9\uff1a<\/p>\n\n\n\n<p>\u25cf \u7b80\u5355\u6613\u5b9e\u73b0\u3002<\/p>\n\n\n\n<p>\u25cf \u8003\u8651\u4e86\u8bcd\u7684\u91cd\u8981\u6027\uff0c\u51cf\u5c11\u4e86\u5e38\u89c1\u4f46\u4e0d\u91cd\u8981\u7684\u8bcd\uff08\u5982\u201c\u7684\u201d\u3001\u201c\u4e0e\u201d\u7b49\uff09\u7684\u6743\u91cd\u3002<\/p>\n\n\n\n<p>\u25cf \u9002\u7528\u4e8e\u81ea\u7136\u8bed\u8a00\u5904\u7406\u4e2d\u7684\u591a\u79cd\u4efb\u52a1\uff0c\u5982\u6587\u672c\u5206\u7c7b\u3001\u6587\u672c\u805a\u7c7b\u3001\u4fe1\u606f\u68c0\u7d22\u7b49\u3002<\/p>\n\n\n\n<p>\u7f3a\u70b9\uff1a<\/p>\n\n\n\n<p>\u25cf \u5bf9\u4e8e\u5927\u578b\u6587\u6863\u96c6\uff0c\u7531\u4e8e\u8bcd\u6c47\u91cf\u5e9e\u5927\uff0c\u751f\u6210\u7684 TF-IDF \u77e9\u9635\u5f80\u5f80\u975e\u5e38\u7a00\u758f\uff0c\u6240\u4ee5\u8ba1\u7b97\u6548\u7387\u4e0d\u9ad8\u3002<\/p>\n\n\n\n<p>\u25cf \u53ea\u8003\u8651\u4e86\u8bcd\u6c47\u7684\u51fa\u73b0\u9891\u7387\uff0c\u800c\u5ffd\u7565\u4e86\u8bcd\u6c47\u5728\u6587\u672c\u4e2d\u7684\u987a\u5e8f\u4fe1\u606f\u3002<\/p>\n\n\n\n<p>\u25cf \u6ca1\u6709\u8003\u8651\u8bcd\u6c47\u4e4b\u95f4\u7684\u8bed\u4e49\u5173\u7cfb\u3002<\/p>\n\n\n\n<p>\u25cf TF-IDF \u7684\u8ba1\u7b97\u65b9\u5f0f\u5bfc\u81f4\u5b83\u7684\u6548\u679c\u4f9d\u8d56\u4e8e\u8db3\u591f\u5927\u7684\u6587\u6863\u96c6\uff0c\u6570\u636e\u91cf\u4e0d\u8db3\u53ef\u80fd\u5bfc\u81f4\u6548\u679c\u4e0d\u4f73\u3002<\/p>\n\n\n\n<p>\u5e94\u7528\u573a\u666f\uff1a<\/p>\n\n\n\n<p>\u25cf \u5173\u952e\u8bcd\u63d0\u53d6<\/p>\n\n\n\n<p>\u25cf \u6587\u6863\u76f8\u4f3c\u5ea6\u8ba1\u7b97<\/p>\n\n\n\n<p>\u25cf \u4fe1\u606f\u68c0\u7d22<\/p>\n\n\n\n<p>\u25cf \u6587\u672c\u5206\u7c7b<\/p>\n\n\n\n<p><strong>Word2Vec\uff08\u51fa\u73b0\u4e8e2013\u5e74\uff09<\/strong><\/p>\n\n\n\n<p>Word2Vec\u662f\u4e00\u79cd\u57fa\u4e8e\u795e\u7ecf\u7f51\u7edc\u7684embedding\u6a21\u578b\uff0c\u7531\u8c37\u6b4c\u7684\u7814\u7a76\u4eba\u5458Tomas Mikolov\u7b49\u57282013\u5e74\u63d0\u51fa\u3002Word2Vec\u7684\u4f5c\u8005\u8ba4\u4e3a\u4e4b\u524d\u7684\u6587\u672c\u5411\u91cf\u5316\u65b9\u6cd5\u5f88\u5c11\u8003\u8651\u8bcd\u4e0e\u8bcd\u4e4b\u95f4\u7684\u610f\u4e49\u5173\u8054\uff0c\u6240\u4ee5\u6548\u679c\u4e0d\u4f73\u3002\u56e0\u6b64\uff0cWord2Vec\u5c06\u5355\u8bcd\u6620\u5c04\u5230\u4f4e\u7ef4\u5411\u91cf\u7a7a\u95f4\u4e2d\uff0c\u4f7f\u5f97\u76f8\u4f3c\u8bcd\u5728\u5411\u91cf\u7a7a\u95f4\u4e2d\u4e5f\u4fdd\u6301\u76f8\u8fd1\u7684\u8ddd\u79bb\u3002<\/p>\n\n\n\n<p>Word2Vec\u4e3b\u8981\u7531\u4e24\u79cd\u6a21\u578b\u67b6\u6784\u7ec4\u6210\uff1aSkip-Gram\u548cContinuous Bag of Words\uff08CBOW\uff09\u3002Skip-Gram\u901a\u8fc7\u5f53\u524d\u8bcd\u9884\u6d4b\u5176\u4e0a\u4e0b\u6587\u8bcd\u3002\u7ed9\u5b9a\u4e00\u4e2a\u4e2d\u5fc3\u8bcd\uff0c\u6a21\u578b\u4f1a\u9884\u6d4b\u4e00\u4e2a\u56fa\u5b9a\u957f\u5ea6\u7684\u4e0a\u4e0b\u6587\u7a97\u53e3\uff08\u524d\u540e\u82e5\u5e72\u8bcd\uff09\u4e2d\u7684\u6240\u6709\u8bcd\u3002\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\uff0c\u6a21\u578b\u4f1a\u5b66\u4e60\u5230\u6bcf\u4e2a\u8bcd\u7684\u5411\u91cf\u8868\u793a\uff0c\u4f7f\u5f97\u80fd\u591f\u66f4\u597d\u5730\u9884\u6d4b\u8fd9\u4e9b\u4e0a\u4e0b\u6587\u8bcd\u3002\u4e0eSkip-Gram\u76f8\u53cd\uff0cCBOW\u7684\u76ee\u6807\u662f\u901a\u8fc7\u4e0a\u4e0b\u6587\u8bcd\u9884\u6d4b\u5f53\u524d\u8bcd\u3002\u7ed9\u5b9a\u4e00\u4e2a\u4e0a\u4e0b\u6587\u7a97\u53e3\u4e2d\u7684\u6240\u6709\u8bcd\uff0c\u6a21\u578b\u4f1a\u9884\u6d4b\u8fd9\u4e2a\u7a97\u53e3\u7684\u4e2d\u5fc3\u8bcd\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/p3-sign.toutiaoimg.com\/tos-cn-i-6w9my0ksvp\/1b54a9ddfa074d3facc1b6012af826c3~noop.image?_iz=58558&amp;from=article.pc_detail&amp;lk3s=953192f4&amp;x-expires=1717655407&amp;x-signature=C8HC3z09UzDQuyo3L3s8fOEXglY%3D\" alt=\"\"\/><\/figure>\n\n\n\n<p>Word2Vec\u6a21\u578b\u7684\u6838\u5fc3\u601d\u60f3\u662f\u901a\u8fc7\u8bad\u7ec3\u795e\u7ecf\u7f51\u7edc\uff0c\u4f7f\u5f97\u5355\u8bcd\u4e0e\u5176\u4e0a\u4e0b\u6587\u4e4b\u95f4\u7684\u5173\u7cfb\u53ef\u4ee5\u5728\u5411\u91cf\u7a7a\u95f4\u4e2d\u88ab\u6709\u6548\u5730\u8868\u793a\u3002Word2Vec\u7684\u8f93\u5165\u5c42\u662f\u4e00\u4e2aone-hot\u5411\u91cf\uff08one-hot vector\uff09\uff0c\u957f\u5ea6\u4e3a\u8bcd\u6c47\u8868\u5927\u5c0f\uff08V\uff09\u3002\u7d27\u63a5\u7740\u662f\u4e00\u4e2a\u6295\u5f71\u5c42\uff0c\u7531\u8f93\u5165\u5c42\u7ecf\u8fc7\u4e00\u4e2a\u6743\u91cd\u77e9\u9635W\uff08\u7ef4\u5ea6\u4e3aV x N\uff0cN\u4e3a\u5d4c\u5165\u5411\u91cf\u7684\u7ef4\u5ea6\uff09\uff0c\u6295\u5f71\u5230N\u7ef4\u5411\u91cf\u7a7a\u95f4\u4e2d\u3002\u6295\u5f71\u5c42\u7684\u8f93\u51fa\u901a\u8fc7\u53e6\u4e00\u4e2a\u6743\u91cd\u77e9\u9635W&#8217;\uff08\u7ef4\u5ea6\u4e3aN x V\uff09\uff0c\u6620\u5c04\u56de\u4e00\u4e2a\u8bcd\u6c47\u8868\u5927\u5c0f\u7684\u5411\u91cf\uff0c\u6b64\u4e3a\u8f93\u51fa\u5c42\u3002\u6700\u540e\u7ecf\u8fc7\u4e00\u4e2aSoftmax\u5c42\u5f97\u5230\u6bcf\u4e2a\u8bcd\u7684\u6982\u7387\u3002\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\uff0cWord2Vec\u6a21\u578b\u901a\u8fc7\u6700\u5927\u5316\u76ee\u6807\u51fd\u6570\uff08\u53d6\u51b3\u4e8e\u4f7f\u7528Skip-Gram\u8fd8\u662fCBOW\uff0c\u4ee5\u53ca\u91c7\u7528\u7684\u4f18\u5316\u51fd\u6570\u7684\u65b9\u6cd5\uff09\u6765\u66f4\u65b0\u6743\u91cd\u77e9\u9635\uff0c\u4ece\u800c\u4f7f\u5f97embedding\u80fd\u591f\u6355\u6349\u5230\u8bcd\u6c47\u7684\u8bed\u4e49\u4fe1\u606f\u3002<\/p>\n\n\n\n<p>\u4f18\u70b9\uff1a<\/p>\n\n\n\n<p>\u25cf \u7531\u4e8e\u91c7\u7528\u4e86\u6d45\u5c42\u795e\u7ecf\u7f51\u7edc\uff0c\u8bad\u7ec3\u901f\u5ea6\u8f83\u5feb\uff0c\u9002\u5408\u5904\u7406\u5927\u89c4\u6a21\u6570\u636e\u3002<\/p>\n\n\n\n<p>\u25cf \u80fd\u591f\u6355\u6349\u5230\u8bcd\u4e0e\u8bcd\u4e4b\u95f4\u7684\u8bed\u4e49\u5173\u7cfb\uff0c\u5982\u8bcd\u6027\u3001\u540c\u4e49\u8bcd\u7b49\u3002<\/p>\n\n\n\n<p>\u25cf \u53ef\u4ee5\u7528\u4e8e\u591a\u79cd\u81ea\u7136\u8bed\u8a00\u5904\u7406\u4efb\u52a1\uff0c\u5982\u6587\u672c\u5206\u7c7b\u3001\u60c5\u611f\u5206\u6790\u3001\u673a\u5668\u7ffb\u8bd1\u7b49\u3002<\/p>\n\n\n\n<p>\u7f3a\u70b9\uff1a<\/p>\n\n\n\n<p>\u25cf \u6bcf\u4e2a\u8bcd\u53ea\u6709\u4e00\u4e2a\u56fa\u5b9a\u7684\u5411\u91cf\u8868\u793a\uff0c\u65e0\u6cd5\u5904\u7406\u591a\u4e49\u8bcd\u7684\u4e0d\u540c\u8bed\u4e49\u3002<\/p>\n\n\n\n<p>\u25cf \u65e0\u6cd5\u8003\u8651\u8bcd\u5728\u4e0d\u540c\u4e0a\u4e0b\u6587\u4e2d\u7684\u8bed\u4e49\u53d8\u5316\u3002<\/p>\n\n\n\n<p>\u5e94\u7528\u573a\u666f\uff1a<\/p>\n\n\n\n<p>\u25cf \u4fe1\u606f\u68c0\u7d22<\/p>\n\n\n\n<p>\u25cf \u6587\u672c\u5206\u7c7b<\/p>\n\n\n\n<p>\u25cf \u63a8\u8350\u7cfb\u7edf\uff0c\u5982\u6839\u636e\u7528\u6237\u7684\u8bc4\u8bba\u8c03\u6574\u63a8\u8350\u7684\u4ea7\u54c1<\/p>\n\n\n\n<p><strong>GloVe\uff08\u51fa\u73b0\u4e8e2014\u5e74\uff09<\/strong><\/p>\n\n\n\n<p>GloVe (Global Vectors for Word Representation)\u662f\u4e00\u79cd\u65e0\u76d1\u7763\u7684\u6587\u672c\u5411\u91cf\u5316\u7684\u5b66\u4e60\u7b97\u6cd5\uff0c\u7531\u65af\u5766\u798f\u5927\u5b66\u7684Jeffrey Pennington\u7b49\u57282014\u5e74\u63d0\u51fa\u3002\u5b83\u901a\u8fc7\u6355\u6349\u5355\u8bcd\u5728\u5927\u89c4\u6a21\u8bed\u6599\u5e93\u4e2d\u7684\u5171\u73b0\u7edf\u8ba1\u4fe1\u606f\uff0c\u751f\u6210\u9ad8\u8d28\u91cf\u7684\u8bcd\u5411\u91cf\u3002GloVe\u7684\u4f5c\u8005\u8ba4\u4e3a\u4f7f\u7528\u5171\u73b0\u8bcd\u7684\u6982\u7387\u7684\u6bd4\u53ef\u4ee5\u66f4\u597d\u5730\u533a\u5206\u76f8\u5173\u548c\u65e0\u5173\u7684\u8bcd\uff0c\u56e0\u6b64GloVe\u7684\u6838\u5fc3\u662f\u5171\u73b0\u77e9\u9635\u548c\u8bcd\u5bf9\u5171\u73b0\u6982\u7387\u3002\u5171\u73b0\u77e9\u9635\u4e2d\u7684\u6bcf\u4e2a\u5143\u7d20Xi,j\u8868\u793a\u8bcdj\u51fa\u73b0\u5728\u8bcdi\u7684\u4e0a\u4e0b\u6587\u4e2d\u7684\u6b21\u6570\u3002\u8bcd\u5bf9\u5171\u73b0\u6982\u7387\u5b9a\u4e49\u4e3a\uff1a<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/p6-sign.toutiaoimg.com\/tos-cn-i-6w9my0ksvp\/e0eba41923594e9a811d006f6225e1b1~noop.image?_iz=58558&amp;from=article.pc_detail&amp;lk3s=953192f4&amp;x-expires=1717655407&amp;x-signature=5ilnNl2F1jeF5aB8JJwvElVP4qY%3D\" alt=\"\"\/><\/figure>\n\n\n\n<p>\u610f\u4e3a\u8bcdj\u5728\u8bcdi\u7684\u4e0a\u4e0b\u6587\u4e2d\u51fa\u73b0\u7684\u6982\u7387\u3002GloVe \u7684\u4f18\u5316\u76ee\u6807\u662f\u901a\u8fc7\u6700\u5c0f\u5316\u4e00\u4e2a\u635f\u5931\u51fd\u6570\uff0c\u4f7f\u5f97embedding\u80fd\u591f\u5f88\u597d\u5730\u9884\u6d4b\u5171\u73b0\u6982\u7387\u3002\u635f\u5931\u51fd\u6570\u5982\u4e0b\uff1a<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/p3-sign.toutiaoimg.com\/tos-cn-i-6w9my0ksvp\/a3cfd999ec27424790d13786be951c75~noop.image?_iz=58558&amp;from=article.pc_detail&amp;lk3s=953192f4&amp;x-expires=1717655407&amp;x-signature=Jk5MSdyF37xGJirEtVIoJwa3GkQ%3D\" alt=\"\"\/><\/figure>\n\n\n\n<p>\u5176\u4e2d\uff0cwi\u548cwj\u5206\u522b\u662f\u8bcdi\u548c\u8bcdj\u7684\u8bcd\u5411\u91cf\uff0cbi\u548cbj\u662f\u504f\u7f6e\u9879\uff0cf(Xij)\u662f\u4e00\u4e2a\u52a0\u6743\u51fd\u6570\uff0c\u7528\u4e8e\u51cf\u5c11\u9ad8\u9891\u5171\u73b0\u5bf9\u6a21\u578b\u7684\u5f71\u54cd\u3002\u52a0\u6743\u51fd\u6570f\u901a\u5e38\u5b9a\u4e49\u4e3a\uff1a<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/p3-sign.toutiaoimg.com\/tos-cn-i-6w9my0ksvp\/4be732e838a643268dd9e36e56f2b2ff~noop.image?_iz=58558&amp;from=article.pc_detail&amp;lk3s=953192f4&amp;x-expires=1717655407&amp;x-signature=sKaHmOAJPLgGJiyX9z4dTkGDD90%3D\" alt=\"\"\/><\/figure>\n\n\n\n<p>\u5176\u4e2dXmax\u662f\u4e00\u4e2a\u9608\u503c\uff0c\u5728GloVe\u4f5c\u8005\u7684\u5b9e\u9a8c\u4e2d\u88ab\u8bbe\u7f6e\u4e3a100\uff0c\u800c3\/4\u88ab\u8ba4\u4e3a\u662f\u4e00\u4e2a\u5408\u9002\u7684\u03b1\u7684\u503c\u3002<br>\u4f18\u70b9:<br>\u25cf\u5145\u5206\u5229\u7528\u4e86\u5168\u5c40\u5171\u73b0\u7edf\u8ba1\u4fe1\u606f\uff0c\u4f7f\u5f97\u8bcd\u5411\u91cf\u53ef\u4ee5\u6355\u6349\u5230\u8bcd\u8bed\u4e4b\u95f4\u7684\u590d\u6742\u5173\u7cfb\u3002<br>\u25cf\u901a\u8fc7\u5bf9\u5171\u73b0\u77e9\u9635\u8fdb\u884c\u5904\u7406\uff0cGloVe \u7684\u8bad\u7ec3\u901f\u5ea6\u8f83\u5feb\uff0c\u9002\u5408\u5927\u89c4\u6a21\u8bed\u6599\u5e93\u3002<br>\u25cf\u751f\u6210\u7684\u8bcd\u5411\u91cf\u5728\u8bed\u4e49\u4e0a\u5177\u6709\u8f83\u5f3a\u7684\u89e3\u91ca\u529b\uff0c\u53ef\u4ee5\u5f88\u597d\u5730\u5e94\u7528\u4e8e\u5404\u79cd\u81ea\u7136\u8bed\u8a00\u5904\u7406\u4efb\u52a1\uff0c\u5982\u6587\u672c\u5206\u7c7b\u3001\u60c5\u611f\u5206\u6790\u3001\u95ee\u7b54\u7cfb\u7edf\u7b49\u3002<\/p>\n\n\n\n<p><br>\u7f3a\u70b9\uff1a<br>\u25cf\u6bcf\u4e2a\u8bcd\u53ea\u6709\u4e00\u4e2a\u56fa\u5b9a\u7684\u5411\u91cf\u8868\u793a\uff0c\u65e0\u6cd5\u5904\u7406\u591a\u4e49\u8bcd\u7684\u4e0d\u540c\u8bed\u4e49\u3002<br>\u25cf\u901a\u5e38\u9700\u8981\u5927\u89c4\u6a21\u7684\u8bed\u6599\u5e93\u6765\u83b7\u5f97\u9ad8\u8d28\u91cf\u7684\u8bcd\u5411\u91cf\u8868\u793a\u3002<br>\u25cf \u5c3d\u7ba1\u8bad\u7ec3\u8fc7\u7a0b\u53ef\u4ee5\u5e76\u884c\u5316\uff0c\u4f46\u4ecd\u9700\u8981\u8f83\u9ad8\u7684\u8ba1\u7b97\u8d44\u6e90\u548c\u65f6\u95f4\u6765\u8bad\u7ec3\u9ad8\u8d28\u91cfembeddings\u3002<br>\u25cf\u867d\u7136\u5173\u6ce8\u5355\u8bcd\u4e4b\u95f4\u7684\u5171\u73b0\u9891\u7387\uff0c\u4f46\u5ffd\u7565\u4e86\u8bed\u6cd5\u7ed3\u6784\u3002<\/p>\n\n\n\n<p><br>\u5e94\u7528\u573a\u666f\uff1a<br>\u25cf\u6587\u672c\u5206\u6790<br>\u25cf\u4fe1\u606f\u68c0\u7d22<br>\u25cf\u95ee\u7b54\u7cfb\u7edf<\/p>\n\n\n\n<p><strong>FastText\uff08\u51fa\u73b0\u4e8e2016\u5e74\uff09<\/strong><\/p>\n\n\n\n<p>FastText\u662f\u7531Facebook\u7684\u4eba\u5de5\u667a\u80fd\u7814\u7a76\u56e2\u961f\u5f00\u53d1\u7684\u4e00\u79cd\u7528\u4e8e\u6587\u672c\u5206\u7c7b\u548c\u8bcd\u5411\u91cf\u8868\u793a\u7684\u5f00\u6e90\u5e93\uff0c\u9002\u7528\u4e8e\u5927\u89c4\u6a21\u6570\u636e\u96c6\u548c\u9ad8\u6027\u80fd\u9700\u6c42\u7684\u4efb\u52a1\u3002FastText\u7684\u76f8\u5173\u8bba\u6587\u7531Piotr Bojanowski\u7b49\u4eba\u57282016\u5e74\u53d1\u8868\uff0cFastText\u7684\u4f5c\u8005\u8ba4\u4e3a\u4e4b\u524d\u7684\u6587\u672c\u5411\u91cf\u5316\u65b9\u6cd5\u6ca1\u6709\u8003\u8651\u5355\u8bcd\u5185\u90e8\u7684\u5b50\u8bcd\u4e4b\u95f4\u7684\u5173\u7cfb\uff0c\u6240\u4ee5\u6027\u80fd\u53d7\u9650\u3002\u4e3a\u4e86\u6539\u5584\u8fd9\u4e2a\u9650\u5236\uff0cFastText\u7684\u6838\u5fc3\u539f\u7406\u5305\u62ec\u4e24\u5927\u90e8\u5206\uff1a\u8bcd\u5411\u91cf\u8868\u793a\u548c\u6587\u672c\u5206\u7c7b\u3002<br><strong>\u8bcd\u5411\u91cf\u8868\u793a<\/strong><\/p>\n\n\n\n<p><br>FastText \u7684\u4e00\u4e2a\u5173\u952e\u521b\u65b0\u5728\u4e8e\u5229\u7528\u4e86\u5b50\u8bcd\u4fe1\u606f\uff08Subword Information\uff09\uff0c\u5373\u5c06\u6bcf\u4e2a\u5355\u8bcd\u8868\u793a\u4e3a\u591a\u4e2a n-gram \u7684\u7ec4\u5408\u3002<\/p>\n\n\n\n<p>\u4f8b\u5982\uff0c\u5355\u8bcd &#8220;where&#8221; \u53ef\u4ee5\u5206\u89e3\u4e3a&#8221;wh&#8221;, &#8220;whe&#8221;, &#8220;her&#8221;, &#8220;ere&#8221;, &#8220;re&#8221;\u7b49 n-gram\u3002\u8fd9\u79cd\u65b9\u6cd5\u4f7f\u5f97\u6a21\u578b\u80fd\u591f\u6355\u6349\u5230\u8bcd\u7684\u5b50\u8bcd\u4fe1\u606f\uff0c\u4ece\u800c\u66f4\u597d\u5730\u5904\u7406\u672a\u767b\u5f55\u8bcd\u548c\u62fc\u5199\u9519\u8bef\u3002FastText \u4f7f\u7528\u57fa\u4e8e\u5728\u4e0a\u6587\u4e2d\u63d0\u5230\u7684Skip-Gram\u7684\u65b9\u6cd5\u6765\u8bad\u7ec3embedding\u3002\u503c\u5f97\u6ce8\u610f\u7684\u662f\uff0cFastText\u4e0d\u4ec5\u4f7f\u7528\u5355\u8bcd\u672c\u8eab\uff0c\u8fd8\u4f7f\u7528\u5355\u8bcd\u7684\u6240\u6709n-gram\u3002\u6bcf\u4e2a\u5355\u8bcd\u88ab\u8868\u793a\u4e3a\u5b83\u6240\u6709n-gram\u5411\u91cf\u7684\u7ec4\u5408\u3002\u4e3a\u4e86\u63d0\u9ad8\u8ba1\u7b97\u6548\u7387\uff0cFastText \u4f7f\u7528\u5206\u5c42Softmax\u4ee3\u66ff\u4f20\u7edf\u7684Softmax\u3002\u5206\u5c42Softmax\u5c06\u6240\u6709\u7c7b\u522b\u7ec4\u7ec7\u6210\u4e00\u68f5\u4e8c\u53c9\u6811\uff0c\u6bcf\u4e2a\u53f6\u8282\u70b9\u5bf9\u5e94\u4e00\u4e2a\u7c7b\u522b\u3002\u901a\u8fc7\u51cf\u5c11\u7c7b\u522b\u6570\u91cf\u7684\u5bf9\u6570\u7ea7\u8ba1\u7b97\uff0c\u5206\u5c42Softmax\u6781\u5927\u5730\u52a0\u901f\u4e86\u8bad\u7ec3\u548c\u9884\u6d4b\u8fc7\u7a0b\u3002<\/p>\n\n\n\n<p><br><strong>\u6587\u672c\u5206\u7c7b\u539f\u7406<\/strong><\/p>\n\n\n\n<p><br>\u5728\u6587\u672c\u5206\u7c7b\u4efb\u52a1\u4e2d\uff0cFastText\u901a\u8fc7\u5c06\u6587\u672c\u8868\u793a\u4e3a\u4e00\u7cfb\u5217\u8bcd\u5411\u91cf\u7684\u5e73\u5747\u6765\u751f\u6210\u6587\u6863\u5411\u91cf\u3002\u5bf9\u4e8e\u4e00\u4e2a\u6587\u672cd\u4e2d\u7684\u6bcf\u4e2a\u8bcdw\uff0c\u5176\u5411\u91cf\u8868\u793a\u4e3avw\u3002\u6587\u672cd\u7684\u8868\u793a\u4e3a\u6240\u6709\u8bcd\u5411\u91cf\u7684\u5e73\u5747\uff0c\u5373\uff1a<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/p3-sign.toutiaoimg.com\/tos-cn-i-6w9my0ksvp\/46cb893611c444a683309e438c2241a8~noop.image?_iz=58558&amp;from=article.pc_detail&amp;lk3s=953192f4&amp;x-expires=1717655407&amp;x-signature=ZfbjiYD%2Fpm5NMVDh5eXhouW2WaM%3D\" alt=\"\"\/><\/figure>\n\n\n\n<p>FastText\u4f7f\u7528\u7ebf\u6027\u5206\u7c7b\u5668\u6765\u8fdb\u884c\u6587\u672c\u5206\u7c7b\u3002\u6587\u672c\u5411\u91cfvd\u901a\u8fc7\u4e00\u4e2a\u5168\u8fde\u63a5\u5c42\uff0c\u5e76\u4f7f\u7528Softmax\u51fd\u6570\u6765\u8f93\u51fa\u7c7b\u522b\u6982\u7387\u3002\u5206\u7c7b\u6a21\u578b\u7684\u516c\u5f0f\u4e3a\uff1a<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/p3-sign.toutiaoimg.com\/tos-cn-i-6w9my0ksvp\/36aca90a0f8e4e159988a7a242aeb36e~noop.image?_iz=58558&amp;from=article.pc_detail&amp;lk3s=953192f4&amp;x-expires=1717655407&amp;x-signature=y%2B6%2B%2FVIJfUZ3hfYj8zkksZVd9AM%3D\" alt=\"\"\/><\/figure>\n\n\n\n<p>\u5176\u4e2dW\u662f\u6743\u91cd\u77e9\u9635\uff0cb\u662f\u504f\u7f6e\u9879\u3002<\/p>\n\n\n\n<p>\u4f18\u70b9\uff1a<\/p>\n\n\n\n<p>\u25cf \u8bad\u7ec3\u548c\u9884\u6d4b\u901f\u5ea6\u4e5f\u5f88\u5feb\uff0c\u9002\u5408\u5b9e\u65f6\u5e94\u7528\u3002<\/p>\n\n\n\n<p>\u25cf \u80fd\u591f\u5904\u7406\u5177\u6709\u5927\u91cf\u7c7b\u522b\u7684\u4efb\u52a1\uff0c\u5e76\u4fdd\u6301\u9ad8\u6548\u6027\u3002<\/p>\n\n\n\n<p>\u25cf \u652f\u6301\u591a\u79cd\u8bed\u8a00\u7684\u6587\u672c\u5904\u7406\u3002<\/p>\n\n\n\n<p>\u7f3a\u70b9\uff1a<\/p>\n\n\n\n<p>\u25cf \u4f7f\u7528\u9759\u6001\u8bcd\u5411\u91cf\uff0c\u65e0\u6cd5\u6355\u6349\u4e0a\u4e0b\u6587\u76f8\u5173\u7684\u8bcd\u4e49\u53d8\u5316\u3002<\/p>\n\n\n\n<p>\u5e94\u7528\u573a\u666f\uff1a<\/p>\n\n\n\n<p>\u25cf \u6587\u672c\u5206\u7c7b<\/p>\n\n\n\n<p>\u25cf \u5173\u952e\u8bcd\u63d0\u53d6<\/p>\n\n\n\n<p>\u25cf \u62fc\u5199\u9519\u8bef\u7ea0\u6b63<\/p>\n\n\n\n<p><strong>ELMo\uff08\u51fa\u73b0\u4e8e2018\u5e74\uff09<\/strong><\/p>\n\n\n\n<p>ELMo\uff08Embeddings from Language Models\uff09\u662f\u4e00\u79cd\u7531Allen Institute for AI\u7684\u7814\u7a76\u8005Matthew E. Peters\u7b49\u57282018\u5e74\u63d0\u51fa\u7684embedding\u65b9\u6cd5\uff0c\u901a\u8fc7\u4e0a\u4e0b\u6587\u4e2d\u7684\u8bcd\u8bed\u6765\u6355\u6349\u8bed\u4e49\u3002ELMo\u7684\u4f5c\u8005\u8ba4\u4e3a\u4e4b\u524d\u7684\u6587\u672c\u5411\u91cf\u5316\u65b9\u5f0f\u6ca1\u6709\u8003\u8651\u4e00\u4e2a\u8bcd\u5728\u4e0d\u540c\u4e0a\u4e0b\u6587\u4e2d\u610f\u4e49\u7684\u53d8\u5316\uff0c\u6240\u4ee5\u4e0eWord2Vec\u548cGloVe\u4e0d\u540c\uff0cELMo\u80fd\u591f\u6355\u6349\u5230\u8bcd\u5728\u4e0d\u540c\u4e0a\u4e0b\u6587\u4e2d\u7684\u4e0d\u540c\u610f\u4e49\uff0c\u4ece\u800c\u751f\u6210\u66f4\u4e3a\u4e30\u5bcc\u548c\u7cbe\u786e\u7684embedding\u3002<\/p>\n\n\n\n<p>ELMo\u4f7f\u7528\u53cc\u5411LSTM\uff08Long Short-Term Memory\uff09\u7f51\u7edc\u4f5c\u4e3a\u57fa\u7840\u67b6\u6784\u3002\u8be5\u7f51\u7edc\u5305\u62ec\u4e00\u4e2a\u5411\u524d\u7684LSTM\u548c\u4e00\u4e2a\u5411\u540e\u7684LSTM\uff0c\u5b83\u4eec\u5206\u522b\u4ece\u5de6\u5230\u53f3\u548c\u4ece\u53f3\u5230\u5de6\u904d\u5386\u6587\u672c\u5e8f\u5217\uff0c\u5b66\u4e60\u6bcf\u4e2a\u8bcd\u7684\u524d\u5411\u548c\u540e\u5411\u4e0a\u4e0b\u6587\u3002\u8fd9\u4e2a\u53cc\u5411LSTM\u88ab\u8bad\u7ec3\u4e3a\u4e00\u4e2a\u8bed\u8a00\u6a21\u578b\uff0c\u76ee\u6807\u662f\u9884\u6d4b\u53e5\u5b50\u4e2d\u7684\u4e0b\u4e00\u4e2a\u8bcd\uff08\u5bf9\u4e8e\u524d\u5411LSTM\uff09\u548c\u524d\u4e00\u4e2a\u8bcd\uff08\u5bf9\u4e8e\u540e\u5411LSTM\uff09\u3002\u5bf9\u4e8e\u7ed9\u5b9a\u7684\u6587\u672c\u5e8f\u5217\uff0cELMo\u4f1a\u4e3a\u6bcf\u4e2a\u8bcd\u63d0\u53d6\u591a\u5c42\u7684\u8868\u793a\u3002\u5bf9\u4e8e\u6bcf\u4e2a\u8bcd\uff0cELMo\u63d0\u4f9b\u4e09\u5c42\u7684embedding\u8f93\u51fa\uff1a\u4e00\u5c42\u6765\u81eaembedding\u5c42\uff08\u7c7b\u4f3c\u4e8e\u4f20\u7edf\u7684embedding\uff09\uff0c\u53e6\u5916\u4e24\u5c42\u6765\u81ea\u53cc\u5411LSTM\u7684\u5404\u81ea\u8f93\u51fa\u3002\u6700\u540e\uff0cELMo\u4f7f\u7528\u8fd9\u4e9b\u5c42\u7684\u52a0\u6743\u548c\u4f5c\u4e3a\u7ed3\u679c\uff0c\u516c\u5f0f\u5982\u4e0b\uff1a<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/p3-sign.toutiaoimg.com\/tos-cn-i-6w9my0ksvp\/997da44336e544c3ba7ac2b4e86eb180~noop.image?_iz=58558&amp;from=article.pc_detail&amp;lk3s=953192f4&amp;x-expires=1717655407&amp;x-signature=vdc%2Bh5D0jjl0P7mlxur6sv4RqKQ%3D\" alt=\"\"\/><\/figure>\n\n\n\n<p>\u5176\u4e2d\u03b3\u662f\u4e00\u4e2a\u53ef\u5b66\u4e60\u7684\u6807\u91cf\uff1bsj\u662f\u5c42j\u7684\u53ef\u5b66\u4e60\u6743\u91cd\uff1bht(j)\u662f\u8bcdt\u5728\u5c42j\u7684\u8868\u793a\u3002\u4e3a\u4e86\u4f7f\u8fd9\u4e9b\u8f93\u51fa\u88ab\u5e94\u7528\u5728\u4e0d\u540c\u4efb\u52a1\u4e2d\uff0cELMo\u63d0\u51fa\u4e86\u4e00\u79cd\u52a8\u6001\u52a0\u6743\u7684\u65b9\u6cd5\uff0c\u53ef\u4ee5\u4e3a\u4e0d\u540c\u5c42\u7684\u8f93\u51fa\u5206\u914d\u4e0d\u540c\u7684\u6743\u91cd\u3002\u8fd9\u6837\uff0c\u6a21\u578b\u4e0d\u4ec5\u4f7f\u7528\u4e86\u6240\u6709\u5c42\u7684\u4fe1\u606f\uff0c\u8fd8\u6839\u636e\u4efb\u52a1\u52a8\u6001\u8c03\u6574\u4e86\u6bcf\u5c42\u4fe1\u606f\u7684\u91cd\u8981\u6027\u3002<\/p>\n\n\n\n<p>\u4f18\u70b9\uff1a<\/p>\n\n\n\n<p>\u25cf \u57fa\u4e8e\u6574\u4e2a\u53e5\u5b50\u7684\u4e0a\u4e0b\u6587\u4fe1\u606f\u751f\u6210embedding\uff0c\u80fd\u591f\u66f4\u597d\u5730\u5904\u7406\u591a\u4e49\u8bcd\u548c\u4e0a\u4e0b\u6587\u4f9d\u8d56\u6027\u5f3a\u7684\u8bcd\u3002<\/p>\n\n\n\n<p>\u25cf \u80fd\u591f\u6355\u6349\u590d\u6742\u7684\u53e5\u5b50\u7ed3\u6784\u548c\u957f\u8ddd\u79bb\u4f9d\u8d56\u5173\u7cfb\u3002<\/p>\n\n\n\n<p>\u25cf \u9884\u8bad\u7ec3\u7684ELMo\u6a21\u578b\u53ef\u4ee5\u5f88\u5bb9\u6613\u5730\u8fc1\u79fb\u5230\u5404\u79cd\u4e0b\u6e38\u81ea\u7136\u8bed\u8a00\u5904\u7406\u4efb\u52a1\u4e2d\u3002<\/p>\n\n\n\n<p>\u7f3a\u70b9\uff1a<\/p>\n\n\n\n<p>\u25cf \u6a21\u578b\u8f83\u5927\uff0c\u8bad\u7ec3\u548c\u63a8\u7406\u8fc7\u7a0b\u9700\u8981\u5927\u91cf\u7684\u8ba1\u7b97\u8d44\u6e90\u3002<\/p>\n\n\n\n<p>\u25cf \u7531\u4e8e\u6a21\u578b\u590d\u6742\uff0c\u63a8\u7406\u901f\u5ea6\u8f83\u6162\u3002<\/p>\n\n\n\n<p>\u5e94\u7528\u573a\u666f\uff1a<\/p>\n\n\n\n<p>\u25cf \u60c5\u611f\u5206\u6790<\/p>\n\n\n\n<p>\u25cf \u95ee\u7b54\u7cfb\u7edf<\/p>\n\n\n\n<p>\u25cf \u6587\u672c\u5206\u7c7b<\/p>\n\n\n\n<p>\u25cf \u673a\u5668\u7ffb\u8bd1<\/p>\n\n\n\n<p><strong>\u57fa\u4e8eTransformer\u7ed3\u6784\u7684embedding\u6a21\u578b\uff08\u51fa\u73b0\u4e8e2018\u5e74\u540e\uff09<\/strong><\/p>\n\n\n\n<p>\u8c37\u6b4c\u7684\u7814\u7a76\u4eba\u5458Ashish Vaswani\u7b49\u57282017\u5e74\u63d0\u51fa\u4e86\u5e26\u6709\u6ce8\u610f\u529b\u673a\u5236\u7684Transformer\u6a21\u578b\u7ed3\u6784\uff0c\u53ef\u4ee5\u5229\u7528\u8f93\u5165\u5e8f\u5217\u7684\u5168\u5c40\u4fe1\u606f\u8fdb\u884c\u8f93\u51fa\uff0c\u6539\u5584\u4e86\u5728RNN\u548cLSTM\u4e2d\u7684\u957f\u8ddd\u79bb\u4f9d\u8d56\u95ee\u9898\uff0c\u540c\u65f6\u63d0\u9ad8\u4e86\u5e76\u884c\u8ba1\u7b97\u80fd\u529b\u4f7f\u8bad\u7ec3\u8fc7\u7a0b\u5927\u5e45\u52a0\u901f\u3002<\/p>\n\n\n\n<p>\u968f\u7740transformer\u7ed3\u6784\u7684\u63a8\u51fa\uff0c\u57282018\u5e74\u5f00\u59cb\u51fa\u73b0\u4e86\u8bb8\u591a\u5404\u5177\u7279\u8272\u7684\u57fa\u4e8eTransformer\u7ed3\u6784\u7684embedding\u6a21\u578b\u3002\u5176\u4e2dBERT (Bidirectional Encoder Representations from Transformers) \u662f\u4e00\u79cd\u53cc\u5411Transformer\u7f16\u7801\u5668\uff0c\u7528\u4e8e\u751f\u6210\u4e0a\u4e0b\u6587\u611f\u77e5\u7684\u8bcd\u5411\u91cf\u3002GPT (Generative Pre-trained Transformer) \u7cfb\u5217\u5219\u662f\u5355\u5411\uff0c\u4e14\u4e3b\u8981\u7528\u4e8e\u6587\u672c\u751f\u6210\u4efb\u52a1\u3002T5 (Text-To-Text Transfer Transformer) \u5c06\u6240\u6709\u4efb\u52a1\u7edf\u4e00\u4e3a\u6587\u672c\u5230\u6587\u672c\u7684\u8f6c\u6362\u3002RoBERTa (Robustly Optimized BERT Approach) \u5bf9BERT\u8fdb\u884c\u4e86\u4f18\u5316\uff0c\u63d0\u5347\u4e86\u6027\u80fd\u3002DistilBERT \u5219\u662fBERT\u7684\u8f7b\u91cf\u7ea7\u7248\u672c\uff0c\u65e8\u5728\u51cf\u5c11\u8ba1\u7b97\u8d44\u6e90\u3002ALBERT (A Lite BERT) \u901a\u8fc7\u53c2\u6570\u5171\u4eab\u548c\u56e0\u5b50\u5206\u89e3\uff0c\u8fdb\u4e00\u6b65\u51cf\u5c0f\u4e86\u6a21\u578b\u5927\u5c0f\u3002\u8fd9\u4e9b\u6a21\u578b\u540c\u4f20\u7edf\u7684embedding\u65b9\u6cd5\u76f8\u6bd4\u6548\u679c\u663e\u8457\u63d0\u9ad8\uff0c\u88ab\u5e7f\u6cdb\u5e94\u7528\u4e8e\u81ea\u7136\u8bed\u8a00\u5904\u7406\u7684\u5404\u79cd\u4efb\u52a1\u3002<\/p>\n\n\n\n<p><strong>\u7ed3\u8bed<\/strong><\/p>\n\n\n\n<p>\u5728\u5927\u6a21\u578b\u65f6\u4ee3\uff0cembedding\u6210\u4e3a\u4e00\u9879\u5fc5\u4e0d\u53ef\u5c11\u7684\u6280\u672f\uff0c\u4e3b\u8981\u662f\u56e0\u4e3a\u5b83\u80fd\u591f\u6709\u6548\u5730\u5c06\u5927\u91cf\u9ad8\u7ef4\u6570\u636e\uff08\u5982\u6587\u672c\u3001\u56fe\u50cf\u548c\u97f3\u9891\u4fe1\u606f\uff09\u8f6c\u6362\u6210\u66f4\u4f4e\u7ef4\u7684\u3001\u5bc6\u96c6\u7684\u5411\u91cf\u8868\u793a\u3002\u8fd9\u79cd\u8868\u793a\u53ef\u4ee5\u6355\u6349\u5230\u6570\u636e\u7684\u672c\u8d28\u7279\u5f81\uff0c\u5e76\u4e14\u5728\u5904\u7406\u548c\u8ba1\u7b97\u4e0a\u66f4\u4e3a\u9ad8\u6548\u3002\u901a\u8fc7embedding\uff0c\u6a21\u578b\u53ef\u4ee5\u5728\u8fd9\u4e9b\u538b\u7f29\u7684\u5411\u91cf\u7a7a\u95f4\u4e2d\u5b66\u4e60\u6570\u636e\u4e4b\u95f4\u7684\u590d\u6742\u5173\u7cfb\u548c\u6a21\u5f0f\uff0c\u4ece\u800c\u63d0\u9ad8\u4e86\u5904\u7406\u901f\u5ea6\u548c\u51c6\u786e\u6027\u3002\u6b64\u5916\uff0cembedding\u4f7f\u5f97\u6a21\u578b\u80fd\u591f\u5728\u8f83\u5c0f\u7684\u8868\u793a\u7a7a\u95f4\u4e2d\u8fdb\u884c\u64cd\u4f5c\uff0c\u51cf\u5c11\u4e86\u8ba1\u7b97\u8d44\u6e90\u7684\u9700\u6c42\uff0c\u540c\u65f6\u4e5f\u4fbf\u4e8e\u6a21\u578b\u7684\u8bad\u7ec3\u548c\u63a8\u7406\uff0c\u8fd9\u5bf9\u4e8e\u5904\u7406\u73b0\u4ee3\u5927\u89c4\u6a21\u6570\u636e\u96c6\u5c24\u4e3a\u5173\u952e\u3002Embedding\u5728\u5927\u6a21\u578b\u65f6\u4ee3\u626e\u6f14\u7740\u6865\u6881\u7684\u89d2\u8272\uff0c\u8de8\u8d8a\u4e86\u539f\u59cb\u6570\u636e\u548c\u673a\u5668\u5b66\u4e60\u7b97\u6cd5\u4e4b\u95f4\u7684\u5de8\u5927\u9e3f\u6c9f\u3002<\/p>\n\n\n\n<p>\u5728\u81ea\u7136\u8bed\u8a00\u5904\u7406\u9886\u57df\uff0c\u6587\u672c\u5411\u91cf\u5316\u6280\u672f\u7684\u53d1\u5c55\u5e26\u6765\u4e86\u9769\u547d\u6027\u7684\u53d8\u5316\u3002\u4ece\u6700\u65e9\u7684One-Hot\u7f16\u7801\u5230\u5f53\u524d\u57fa\u4e8eTransformer\u7ed3\u6784\u7684embedding\u6a21\u578b\uff0c\u6bcf\u4e00\u6b21\u521b\u65b0\u90fd\u5e26\u6765\u4e86\u66f4\u9ad8\u6548\u3001\u66f4\u7cbe\u51c6\u7684\u6587\u672c\u8868\u793a\u65b9\u5f0f\uff0c\u4e3a\u6587\u672c\u5904\u7406\u63d0\u4f9b\u4e86\u66f4\u4e3a\u7075\u6d3b\u3001\u9ad8\u6548\u7684\u89e3\u51b3\u65b9\u6848\u3002\u672a\u6765\uff0c\u968f\u7740\u6280\u672f\u7684\u4e0d\u65ad\u8fdb\u6b65\u548c\u5e94\u7528\u573a\u666f\u7684\u4e0d\u65ad\u62d3\u5c55\uff0c\u6587\u672c\u5411\u91cf\u5316\u6280\u672f\u5c06\u7ee7\u7eed\u53d1\u6325\u7740\u91cd\u8981\u7684\u4f5c\u7528\uff0c\u5e76\u4e3a\u4eba\u5de5\u667a\u80fd\u5728\u6587\u672c\u7406\u89e3\u548c\u5904\u7406\u9886\u57df\u5e26\u6765\u66f4\u5e7f\u9614\u7684\u53d1\u5c55\u7a7a\u95f4\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u6587\u672c\u5411\u91cf\u5316\u662f\u5c06\u6587\u672c\u6570\u636e\u8f6c\u6362\u4e3a\u6570\u503c\u5411\u91cf\u7684\u8fc7\u7a0b\u3002\u8fd9\u79cd\u8f6c\u6362\u4f7f\u5f97\u8ba1\u7b97\u673a\u80fd\u591f\u7406\u89e3\u548c\u5904\u7406\u6587\u672c\u6570\u636e\uff0c\u4ece\u800c\u53ef\u4ee5\u5e94\u7528\u5404\u79cd\u673a\u5668\u5b66\u4e60 [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[30],"tags":[],"class_list":["post-1955","post","type-post","status-publish","format-standard","hentry","category-30"],"_links":{"self":[{"href":"http:\/\/blog.xtaa.cn\/index.php\/wp-json\/wp\/v2\/posts\/1955","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/blog.xtaa.cn\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/blog.xtaa.cn\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/blog.xtaa.cn\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/blog.xtaa.cn\/index.php\/wp-json\/wp\/v2\/comments?post=1955"}],"version-history":[{"count":1,"href":"http:\/\/blog.xtaa.cn\/index.php\/wp-json\/wp\/v2\/posts\/1955\/revisions"}],"predecessor-version":[{"id":1956,"href":"http:\/\/blog.xtaa.cn\/index.php\/wp-json\/wp\/v2\/posts\/1955\/revisions\/1956"}],"wp:attachment":[{"href":"http:\/\/blog.xtaa.cn\/index.php\/wp-json\/wp\/v2\/media?parent=1955"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/blog.xtaa.cn\/index.php\/wp-json\/wp\/v2\/categories?post=1955"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/blog.xtaa.cn\/index.php\/wp-json\/wp\/v2\/tags?post=1955"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}