{"id":9850,"date":"2016-08-01T01:00:00","date_gmt":"2016-07-31T17:00:00","guid":{"rendered":"http:\/\/swarma.org\/%swarma-paper%\/"},"modified":"2018-12-06T09:02:33","modified_gmt":"2018-12-06T01:02:33","slug":"%e8%87%aa%e7%84%b6%e8%af%ad%e8%a8%80%e7%90%86%e8%a7%a3%e4%b8%8e%e6%b7%b1%e5%ba%a6%e5%ad%a6%e4%b9%a0%e8%af%be%e5%a0%82%e7%ac%94%e8%ae%b01-word2vec","status":"publish","type":"post","link":"https:\/\/swarma.org\/?p=9850","title":{"rendered":"\u81ea\u7136\u8bed\u8a00\u7406\u89e3\u4e0e\u6df1\u5ea6\u5b66\u4e60\u8bfe\u5802\u7b14\u8bb01: Word2Vec"},"content":{"rendered":"
\n
\n
\n
\"\u96c6\u667a\" <\/p>\n
\n
\n
\n
\n
2016\u5e747\u670821\u65e5 <\/section>\n
\uff08AndreaSunshine\u3001\u7504\u6167\u73b2\u3001\u7941\u8f89\u8bb0\u5f55\uff0cJake\u6574\u7406\u3001\u7efc\u5408\uff09<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n

\"\u96c6\u667a\" <\/section>\n<\/section>\n

\n
\n
\n
\n
\n
\n
\n
\n
\n
\n
\u4e3b\u9898\uff1aWord2Vec:Basics and Model<\/section>\n
Natural Language Processing in the Era of Deep Learning<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\n
\"\u96c6\u667a\" <\/p>\n
\n
\u4e3b\u8bb2\u4eba\uff1aYanran Li<\/strong> <\/section>\n
\uff08The Hong Kong Polytechnic University\uff09<\/span><\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\n
\n
\u8bfe\u7a0b\u4e3b\u9875\uff0c\u4e91\u7aef\u8d44\u6599\u53ca\u52a9\u6559\u90ae\u7bb1<\/section>\n<\/section>\n
<\/section>\n<\/section>\n<\/section>\n
\n
\n
\n
Course Page:<\/span><\/section>\n

https:\/\/swarma.github.io<\/p>\n

\u25b2 All resources (syllabus,slides,suggested readings,homework) are shared via<\/span><\/p>\n


<\/section>\n
http:\/\/pan.baidu.com\/s\/1qYN9yji<\/section>\n

Password:  j7hu<\/span><\/p>\n

(\u53ef\u70b9\u51fb\u6587\u672b<\/span>\u9605\u8bfb\u539f\u6587<\/span>\u4e0b\u8f7d\uff09<\/span><\/span><\/p>\n


<\/section>\n
Class monitor:<\/section>\n
 swarma_era@163.com<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\n
\n
\n

<\/strong><\/section>\n
\u8fd9\u6b21\u8bfe\u7a0b\u4e3b\u8981\u4ecb\u7ecd\u5b66\u4e60Word embedding \u548c word2vec \u524d\u9700\u8981\u4e86\u89e3\u7684\u77e5\u8bc6\u3002<\/strong><\/section>\n

<\/span><\/section>\n<\/section>\n
\n
<\/section>\n
<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\n
\n

<\/strong><\/section>\n

<\/strong><\/section>\n
\u4e00\u3001\u80cc\u666f<\/strong><\/strong><\/strong><\/span>
<\/section>\n

<\/strong><\/section>\n
\n

\u4f20\u7edf\u81ea\u7136\u8bed\u8a00\u5904\u7406\u7684\u8868\u8fbe\uff0c\u4e00\u822c\u4ee5\u8bcd\u4e3a\u5355\u4f4d\u3002\u81ea\u7136\u8bed\u8a00\u5b57\u7b26\u8868\u8fbe\u6210\u8ba1\u7b97\u673a\u80fd\u5904\u7406\u7684\u4fe1\u606f\uff0c\u91c7\u53d6\u79bb\u6563\u8868\u8fbe\u3002\u5047\u8bbe\u6709\u4e2a\u8bcd\u8868\uff0c\u53ef\u4ee5\u7406\u89e3\u4e3a\u5b57\u5178\uff0c\u5982\u4f55\u5728\u8fd9\u4e00\u4e07\u4e2a\u5b57\u4e2d\u8868\u8fbe\u8fd9\u4e00\u4e2a\u5b57\u3002\u7528\u4e00\u4e07\u7ef4\u7684\u5411\u91cf\u8868\u8fbe\u8fd9\u4e2a\u5b57\uff0c\u67d0\u4e2a\u5b57\u53ea\u53ef\u80fd\u5728\u67d0\u4e00\u7ef4\u5730\u65b9\u4e3a1\uff0c\u5176\u4ed6\u4f4d\u7f6e\u90fd\u4e3a0\u3002<\/p>\n


<\/p>\n

\u8fd9\u79cd\u65b9\u6cd5\u7279\u70b9\uff1a\u6839\u636e\u8bcd\u8868\u5927\u5c0f\u51b3\u5b9a\u5411\u91cf\u957f\u5ea6\u3002\u5f88\u5360\u7a7a\u95f4\uff0c\u5373\u4f7f\u7a00\u758f\u8868\u793a\u3002\u7f3a\u4e4f\u6cdb\u5316\u6027\uff0c\u5373lack of generalization\u3002\u8fd9\u79cd\u8868\u793a\u6ca1\u6cd5\u8868\u8fbe\u76f8\u4f3c\uff08cos\u51fd\u6570\uff09\u3002\u6240\u4ee5\u5e0c\u671b\u80fd\u591f\u6709\u4e00\u79cd\u65b9\u5f0f\u8ba1\u7b97\u8fd9\u79cd\u6cdb\u5316\u6027\uff0c\u7a20\u5bc6\u5411\u91cf\u3002<\/p>\n


<\/p>\n

\u201cYou shall know a word by the company it keeps.\u201d–J.R. Firth(1957)\u57fa\u4e8e\u540c\u4e00\u79cd\u601d\u60f3\uff0c\u751f\u51fa\u5f88\u591a\u65b9\u6cd5\uff0c\u4e3b\u8981\u6709\u4e24\u79cd\u5047\u8bf4\u3002<\/p>\n


<\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n

\n
\"\u96c6\u667a\"<\/section>\n<\/section>\n
\n
\"\u96c6\u667a\"<\/section>\n<\/section>\n
\n
\n
\n

<\/strong><\/section>\n

<\/strong><\/section>\n
\n

\u5047\u8bf4\u4e00\uff1athe distributional hypothesis \u5206\u5e03\u5f0f\u5047\u8bf4 <\/strong><\/p>\n


<\/p>\n

\u4e00\u4e2a\u8bcd\u7531\u5468\u56f4\u8bcd\u6765\u63a8\u65ad\u3002\u76f8\u4f3c\u7684\u8bcd\u4f1a\u51fa\u73b0\u5728\u76f8\u4f3c\u7684\u8bed\u5883\u91cc\u3002\u7531\u8bcd\u8ddf\u8bed\u5883\u7684\u5173\u7cfb\u6765\u5224\u65ad\u76f8\u4f3c\u5ea6\u3002\u4e0b\u9762\u662f\u7531\u8fd9\u79cd\u5047\u8bf4\uff0c\u4ea7\u751f\u7684\u5f88\u591a\u79cd\u65b9\u6cd5\u6765\u8ba1\u7b97\u8fd9\u4e2a\u5b9e\u503c\u5411\u91cf\u30021954\u5e74BOW\u6a21\u578b\u4e0d\u8003\u8651\u8bed\u5e8f\uff0c\u51fa\u73b0\u4e00\u6b21\u52a01\u30022003\u5e74LDA\u6a21\u578b\uff0c\u662f\u4e3b\u9898\u6a21\u578b\u4e2d\u7684\u67d0\u4e00\u4e2a\u7279\u4f8b\u3002<\/p>\n


<\/p>\n

Distributional models \u57fa\u4e8e\u90a3\u4e2a\u5047\u8bbe\uff0c\u7ec6\u5316\u5230\u81ea\u5df1\u7684\u6a21\u578b\u4e2d\uff0c\u610f\u601d\u662f\u76f8\u5173\u7684\u8bcd\u51fa\u73b0\u5728\u76f8\u540c\u7684\u53e5\u5b50\u91cc\u9762\u3002\u4f8b\u5982\u4eca\u665a\u7684\u5929\u7a7a\u6709\u5f88\u591a\u661f\u661f\u3002\u53e5\u5b50\u4e2d\u5929\u7a7a\u548c\u661f\u661f\u76f8\u5173\uff0c\u6a2a\u5411\u3002\u5171\u73b0\uff0c\u76f8\u5173\u8bcd\u5171\u540c\u51fa\u73b0\u51e0\u6b21\uff0csvd\u77e9\u9635\u5206\u89e3\u65b9\u5f0f\u62c6\u9664\u5b9e\u503c\u5411\u91cf\u7ed3\u679c\u3002<\/p>\n


<\/p>\n


<\/p>\n

\u5047\u8bf4\u4e8c\uff1aDistributed models  <\/strong><\/p>\n


<\/p>\n

\u76f8\u4f3c\u8bcd\u5728\u76f8\u4f3c\u8bed\u5883contex\u4e0b\u3002\u4f8b\u5982\u4eca\u5929\u5929\u7a7a\u6709\u5f88\u591a\u661f\u661f\u3002\u4eca\u5929\u5929\u7a7a\u6709\u4e2a\u592a\u9633\u3002\u8fd9\u4e24\u4e2a\u53e5\u5b50\u4e2d\u7684\u661f\u661f\u548c\u592a\u9633\u8fd9\u4e24\u4e2a\u8bcd\u3002\u7eb5\u5411\u76f8\u4f3c\u6027\u3002\u7531\u795e\u7ecf\u7f51\u7edc\u8bed\u8a00\u6a21\u578b\u5f97\u51fa\u7684\u526f\u4ea7\u7269\uff0c\u6df1\u5ea6\u5b66\u4e60\u526f\u4ea7\u7269\uff0c\u6df1\u5ea6\u5b66\u4e60\u4e2d\u57fa\u4e8e\u9519\u8bef\u4f20\u5bfc\u7684\u9884\u6d4b\u65b9\u5f0f\u3002<\/p>\n


<\/p>\n

Context \u53ef\u4ee5\u540c\u4e00\u4e2a\u53e5\u5b50\uff0c\u4e5f\u53ef\u4ee5\u7eb5\u5411\u3002\u8fd9\u4e24\u8005\u7684\u533a\u522b\u548c\u8054\u7cfb\u5982\u4e0b\u56fe\uff1a<\/p>\n


<\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n

\n
\"\u96c6\u667a\"<\/section>\n<\/section>\n
\n
\n
\n
<\/span><\/section>\n

<\/section>\n
\u540e\u6765Bangio\u5728\u4e00\u7bc7\u8bba\u6587\u4e2d\u63d0\u51fa\u4e86\u4e00\u79cd\u5229\u7528\u795e\u7ecf\u7f51\u7edc\u7684\u65b9\u6cd5\u6765\u505a\u81ea\u7136\u8bed\u8a00\u7406\u89e3\u7684\u601d\u8def\uff1a<\/section>\n

<\/span><\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\"\u96c6\u667a\"<\/section>\n<\/section>\n
\n
\n
\n

<\/section>\n
\u5c3d\u7ba1\u8fd9\u4e00\u6846\u67b6\u672c\u8eab\u5e76\u4e0d\u5b9e\u7528\uff0c\u4f46\u662f\u5374\u5305\u542b\u4e86\u5c06\u5355\u8bcd\u8fdb\u884c\u5411\u91cf\u7f16\u7801\u7684\u64cd\u4f5c\uff0c\u800c\u8fd9\u5374\u6210\u4e3a\u4e86\u540e\u6765\u8bcd\u5411\u91cf\u7b97\u6cd5\u7684\u524d\u8eab\u3002<\/section>\n

<\/section>\n
\u4e8e\u662f\uff0c\u5230\u4e862013\u5e74Google\u7684Mikolov\u63d0\u51fa\u4e86\u4e24\u5957\u5b9e\u7528\u7684\u7b97\u6cd5\uff0c\u5b83\u4eec\u7edf\u79f0\u4e3aWord2Vec\u3002Word2vec \u662f\u4ee3\u7801\u9879\u76ee\u7684\u540d\u5b57\uff0c\u53ea\u662f\u8ba1\u7b97word embedding\u7684\u4e00\u4e2a\u5de5\u5177\uff0c\u662fCBOW\u548cSkip-Gram\u8fd9\u4e24\u4e2a\u6a21\u578b\u5f00\u6e90\u51fa\u6765\u7684\u5de5\u5177\u3002\u8fde\u7eed\u5b9e\u503c\u8bcd\u8868\u8fbe\u4e5f\u53eb\u8bcd\u5d4c\u5165 word embedding\u3002<\/strong><\/section>\n

<\/section>\n
\u5982\u4e0b\u56fe\u6240\u793a\uff0c\u5de6\u8fb9\u7684CBOW\u662f\u5229\u7528\u8bcd\u7684\u4e0a\u4e0b\u6587\u9884\u6d4b\u5f53\u524d\u7684\u5355\u8bcd\uff1b\u800cSkip-Gram\u5219\u662f\u5229\u7528\u5f53\u524d\u8bcd\u6765\u9884\u6d4b\u4e0a\u4e0b\u6587\u3002<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\"\u96c6\u667a\"<\/section>\n<\/section>\n
\n
\n
\n

<\/section>\n

<\/section>\n

<\/section>\n
\u4e8c\u3001Skipgram\u7684\u539f\u7406<\/span><\/strong><\/section>\n
 <\/section>\n
\u5728\u8fd9\u8282\u8bfe\u4e2d\uff0c\u6211\u4eec\u91cd\u70b9\u4ecb\u7ecd\u540e\u4e00\u79cd\uff0c\u5373Skipgram\uff0c\u5b83\u7684\u7ed3\u6784\u5982\u56fe\uff1a<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\"\u96c6\u667a\"<\/section>\n<\/section>\n
\n
\n
\n

<\/section>\n
\u5b83\u7684\u5de5\u4f5c\u539f\u7406\u5982\u4e0b\uff1a<\/section>\n
 <\/section>\n
\u9996\u5148\u8981\u6709\u8bed\u6599\u5e93\u548c\u8bcd\u5e93\uff0c\u5927\u6982\u8bcd\u5e93\u9700\u8981\u5305\u542b\u6240\u6709\u8bed\u6599\u5e93\u4e2d\u7684\u8bcd\uff0c\u6216\u8005\u628a\u6ca1\u56ca\u62ec\u7684\u8bcdencode\u6210\u4e00\u4e2a\u7edf\u4e00\u7684\u503c\u3002\u5982\u4e0b\u56fe\u6240\u793a\uff0c\u6bcf\u4e00\u4e2ainput\u90fd\u662f\u4e00\u4e2a1-hot encoded\u7684\u8bcd\u3002\u6bcf\u4e00\u4e2aoutput\u662f\u4e00\u4e2a\u6982\u7387Vector\u8868\u793aVector\u4f4d\u7f6e\u4e0a\u7684\u8bcd\u51fa\u73b0\u5728\u5f53\u524doutput\u4f4d\u7f6e\u7684\u6982\u7387\uff0cVector\u7684\u4f4d\u7f6e\u548cinput\u7684\u8bcd\u7684\u4f4d\u7f6e\u76f8\u7b26\u3002<\/section>\n

<\/section>\n
\u5f88\u7ed5\u53e3\uff0c\u5f53\u65f6\u542c\u8bfe\u8fd9\u90e8\u5206\u4e5f\u6ca1\u542c\u660e\u767d\uff0c\u4e3e\u4e2a\u4f8b\u5b50\u5427\uff0c\u5982\u679c\u6709\u8fd9\u6837\u4e00\u4e2a\u8bcd\u5e8f\u5217 \uff08\u6211\u7231\u4e2d\u56fd\uff09\u90a3\u4e48\u5c31\u4f1a\u6709\u56db\u4e2a1-hot encoded input vector 1000, 0100,0010,0001\u3002<\/section>\n

<\/section>\n
 \u8bb0\u5f97\u90a3\u4e2a\u725b\u4eba\u7684\u540d\u8a00\u4e48\uff0c\u8bcd\u7684\u542b\u4e49\u5305\u542b\u5728\u8fd9\u4e2a\u8bcd\u5468\u56f4\u7684\u5c0f\u4f19\u4f34\u4e2d\u3002\u90a3\u4e48\u5c31\u8981\u53d6\u5468\u56f4\u7684\u5c0f\u4f19\u4f34\uff0c\u6bd4\u5982\u8bf4\u53d6\u524d\u540e\u5404\u4e00\u4e2a\u3002<\/section>\n

<\/section>\n
\u90a3\u4e48\u5c31\u4f1a\u6709\u4e09\u4e2aoutput\uff0c\u521a\u624d\u8bf4\u4e86\u6bcf\u4e2aoutput\u662f\u4e00\u4e2a\u6982\u7387Vector\uff0c \u5047\u8bbe\u8fd9\u4e09\u4e2aoutput Vector\u662f\uff080.3, 0.5, 0.7, 0.2\uff09\uff0c \uff080.1\uff0c0.9\uff0c0.1\uff0c0.2\uff09\uff0c \uff080.3\uff0c0.1\uff0c0.8\uff0c0.1\uff09\u5982\u679c\u5f53\u524d\u8f93\u5165\u7684\u8bcd\u662f0100\u7684\u8bdd\uff0c\u4e5f\u5c31\u662f\u201c\u7231\u201d\u8fd9\u4e2a\u8bcd\uff0c\u90a3\u4e48\u4e09\u4e2aVector\u4e2d\u7684\u7b2c\u4e00\u4e2a\uff080.3\uff0c0.5\uff0c0.7\uff0c0.2\uff09\u8868\u793a\u7684\u5c31\u662f\u201c\u6211\u201d\u51fa\u73b0\u5728\u201c\u7231\u201d\u4e4b\u524d\u4e00\u4f4d\u7684\u6982\u7387\u662f0.3\uff0c \u201c\u7231\u201d\u51fa\u73b0\u5728\u201c\u7231\u201d\u4e4b\u524d\u4e00\u4f4d\u7684\u6982\u7387\u662f0.5, “\u4e2d”\u51fa\u73b0\u5728\u201c\u7231\u201d\u4e4b\u524d\u4e00\u4f4d\u7684\u6982\u7387\u662f0.7\uff0c\u201c\u56fd\u201d\u51fa\u73b0\u5728\u201c\u7231\u201d\u4e4b\u524d\u4e00\u4f4d\u7684\u6982\u7387\u662f0.2 \u3002\u540c\u6837\u5bf9output\u4e2d\u7b2c\u4e8c\u4e2aVector\u5c31\u662f\u51fa\u73b0\u5728\u5f53\u524d\u201c\u7231\u201d\u7684\u4f4d\u7f6e\u7684\u6982\u7387\uff0c\u4e5f\u8bb8\u6709\u5c0f\u4f19\u4f34\u95ee\uff0c\u8fd9\u4e2a\u4f4d\u7f6e\u4e0d\u5e94\u8be5\u662f\u786e\u5b9a\u7684\u8bcd\u201c\u7231\u201d\u5417? \u6211\u4e5f\u6709\u540c\u6837\u7684\u7591\u95ee\uff0c\u6682\u65f6\u641e\u4e0d\u6e05\u3002\u7b2c\u4e09\u4e2aoutput\u5c31\u662f\u6bcf\u4e2a\u8bcd\u51fa\u73b0\u5728\u201c\u7231\u201d\u4e4b\u540e\u7684\u6982\u7387\u4e86\u3002<\/span><\/section>\n

<\/section>\n
\u5982\u679c\u8fd9\u90e8\u5206\u7406\u89e3\u4e86\uff0c\u90a3\u4e48\u6211\u4eec\u5c31\u53ef\u4ee5\u6709\u4e00\u4e2a\u4f18\u5316\u7684\u76ee\u6807\u4e86\uff0c\u5c31\u662f\u4f7f\u5f97\u8fd9\u4e00\u7cfb\u5217\u7684Output Vector\u6700\u63a5\u8fd1\u771f\u5b9e\u7684\u60c5\u51b5\u3002\u6240\u8c13\u771f\u5b9e\u7684\u60c5\u51b5\u5c31\u662f\u5404\u4e2a\u6587\u672c\u4e86\u3002\u4e5f\u5c31\u662f\u8bf4\u5728\u6536\u96c6\u7684\u4e0a\u767e\u4e07\u5343\u4e07\u771f\u5b9e\u7684\u6587\u672c\u4e2d\uff0c\u5bf9\u4e8e\u201c\u7231\u201d\u8fd9\u4e2a\u8bcd\uff0c\u4e4b\u524d\uff08\u7b2c\u4e00\u4e2aVector\uff09\u51fa\u73b0\u201c\u6211\u201d\uff0c\u201c\u7231\u201d\uff0c\u201c\u4e2d\u201d\uff0c\u201c\u56fd\u201d\u7684\u6982\u7387\u5230\u5e95\u6709\u591a\u5927\u3002<\/section>\n
 <\/section>\n
Input \u548c output\u4f18\u5316\u76ee\u6807\u90fd\u6709\u4e86\uff0c\u90a3\u4e48\u901a\u8fc7\u4e00\u7cfb\u5217\u590d\u6742\u7684\u516c\u5f0f\uff0c\u4ee5\u53ca\u53c2\u7167\u4e0b\u56fe\uff0c\u5c31\u4f1a\u5f97\u5230\u4e24\u4e2a\u53c2\u6570W\u548cW\u2019\u3002 W\u6211\u4eec\u53ef\u4ee5\u7b80\u5355\u7684\u7406\u89e3\u4e3a\u8bcd\u5411\u91cf\uff0c\u4e5f\u5c31\u662f\u8bf4\u8fd9\u6837\u7684\u5f62\u5f0f\uff1a<\/section>\n

<\/section>\n
\u6211 \uff080.9\uff0c-0.2\uff0c0.1\uff09<\/section>\n
\u7231 \uff080.8\uff0c 0.2\uff0c0.2\uff09<\/section>\n
\u4e2d \uff080.1\uff0c 0.3\uff0c0.4\uff09<\/section>\n
\u56fd \uff080.2\uff0c 0.3\uff0c0.6\uff09<\/section>\n
 <\/section>\n
\u90a3\u4e48\u8fd9\u4e00\u7cfb\u5217\u7684\u590d\u6742\u8fd0\u7b97\u662f\u600e\u4e48\u505a\u7684\u5462\uff1f\u7b54\u6848\u5c31\u5728\u4e8e\u4f18\u5316\u3002\u6211\u4eec\u8ba9\u8bad\u7ec3\u8fd9\u4e2a\u7f51\u7edc\u7684\u6700\u7ec8\u76ee\u6807\u5c31\u662f\u53bb\u6839\u636e\u5f53\u524d\u8bcd\u6765\u4f18\u5316\u9884\u6d4b\u51fa\u5b83\u7684\u4e0a\u4e0b\u6587\u7684\u6982\u7387\uff0c\u4e8e\u662f\uff0c\u6211\u4eec\u5c31\u53ef\u4ee5\u5199\u51fa\u76ee\u6807\u51fd\u6570\uff1a<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\"\u96c6\u667a\"<\/section>\n<\/section>\n
\n
\n
\n

<\/section>\n
\u5176\u4e2d\uff0cT\u662f\u8bed\u6599\u5e93\u5355\u8bcd\u7684\u603b\u4e2a\u6570\uff0cp(wt+j|wt)\u662f\u5df2\u77e5\u5f53\u524d\u8bcdt\uff0c\u9884\u6d4b\u5468\u56f4\u8bcd\u7684\u603b\u6982\u7387\u5bf9\u6570\u503c\u3002<\/section>\n

<\/section>\n
\uff081\uff09\u4f9d\u7136\u6709\u4e00\u4e2a tips: \u8fd9\u91cc\u503c\u5f97\u6ce8\u610f\u7684\u4e0d\u662f\u51fd\u6570\u672c\u8eab\uff0c\u56e0\u4e3a\u4e4b\u6240\u4ee5\u5728\u76ee\u6807\u51fd\u6570\u4e2d\u7528\u5bf9\u6570\uff0c\u5b8c\u5168\u662f\u4e3a\u4e86\u8ba1\u7b97\u65b9\u4fbf\uff0c\u503c\u5f97\u6ce8\u610f\u7684\u662f\u7b2c\u4e8c\u4e2a\u6c42\u548c\u53f7\u7684\u4e0b\u6807\uff0c\u9700\u8981\u5728\u7b2c j \u4e2a\u8bcd\u7684\u5de6\u53f3\u5206\u522b\u6709 c\u4e2a\u3002<\/span><\/section>\n

<\/span><\/section>\n
\u8fd9\u91cc,\u6982\u7387\u7684\u8ba1\u7b97\u65b9\u5f0f\u5982\u4e0b\uff1a<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\"\u96c6\u667a\"<\/section>\n<\/section>\n
\n
\n
\n

<\/section>\n
\uff082\uff09\u518d\u6765\u4e00\u4e2a tips\uff1a\u8fd9\u91cc\u65e0\u8bba\u662f v \u8fd8\u662f v\u2019\u5bf9\u5e94\u7684\u90fd\u662f\u56fe 2 \u4e2d\u7684 v-dim\uff0c\u56e0\u6b64\u8fd9\u91cc\u7684\u4e58\u6cd5\u662f\u5411\u91cf\u4e58\u6cd5\u3002<\/span><\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\n
\n

<\/section>\n

<\/section>\n
\u4e09\u3001\u8d1f\u91c7\u6837<\/span><\/strong><\/section>\n
 <\/section>\n
\u800c\u540e\uff0c\u4f9d\u7136\u53ef\u4ee5\u91c7\u7528\u53cd\u5411\u4f20\u64ad\u7b49\u65b9\u5f0f\u8ba1\u7b97\u3002\u7136\u800c\uff0c\u5728\u5b9e\u9645\u7684\u8ba1\u7b97\u8fc7\u7a0b\u4e2d\uff0c\u8fd0\u7b97\u91cf\u8fc7\u4e8e\u5de8\u5927\uff0c\u4e8e\u662f\u4eba\u4eec\u5c31\u50cf\u51fa\u4e86\u4e00\u4e2a\u7edd\u5999\u7684\u89e3\u51b3\u529e\u6cd5\uff0c\u4ece\u800c\u5feb\u901f\u5730\u5b8c\u6210\u4efb\u52a1\u3002\u8fd9\u79cd\u65b9\u6cd5\u5c31\u88ab\u79f0\u4e3a\u201c\u8d1f\u91c7\u6837\u201d\u3002<\/section>\n
 <\/section>\n
\u8d1f\u91c7\u6837\u7684\u6838\u5fc3\u601d\u60f3\u662f\uff0c\u5982\u679c\u5c06\u81ea\u7136\u8bed\u8a00\u770b\u4f5c\u662f\u4e00\u4e32\u5355\u8bcd\u7684\u968f\u673a\u7ec4\u5408\uff0c\u90a3\u4e48\u5b83\u7684\u51fa\u73b0\u6982\u7387\u662f\u5f88\u5c0f\u7684\u3002\u4e8e\u662f\uff0c\u5982\u679c\u6211\u4eec\u5c06\u62fc\u51d1\u7684\u5355\u8bcd\u968f\u673a\u7ec4\u5408\uff08\u8d1f\u91c7\u6837\uff09\u8d77\u6765\u5c06\u4f1a\u4ee5\u5f88\u5927\u7684\u6982\u7387\u4e0d\u4f1a\u51fa\u73b0\u5728\u5f53\u524d\u6587\u7ae0\u4e2d\u3002\u4e8e\u662f\uff0c\u6211\u4eec\u5f88\u663e\u7136\u5e94\u8be5\u81f3\u5c11\u8ba9\u6211\u4eec\u7684\u6a21\u578b\u5728\u8fd9\u4e9b\u8d1f\u91c7\u6837\u51fa\u6765\u7684\u5355\u8bcd\u7ec4\u5408\u4e0a\u9762\u51fa\u73b0\u6982\u7387\u5e94\u8be5\u5c3d\u53ef\u80fd\u5730\u5c0f\uff0c\u540c\u65f6\u8981\u8ba9\u771f\u6b63\u51fa\u73b0\u5728\u6587\u4e2d\u7684\u5355\u8bcd\u7ec4\u5408\u51fa\u73b0\u6982\u7387\u5927\u3002\u8fd9\u6837\u6211\u4eec\u7684\u6a21\u578b\u624d\u8db3\u591f\u6709\u6548\u3002\u4e8e\u662f\uff0c\u6211\u4eec\u7684\u76ee\u6807\u53d8\u6210\u4e86\u4f18\u5316\u65b0\u7684\u6982\u7387\u51fd\u6570\uff0c\u4ece\u800c\u5f97\u5230\u6700\u4f18\u7684\u8bcd\u5411\u91cf\u5d4c\u5165\uff1a<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\"\u96c6\u667a\"<\/section>\n<\/section>\n
\n
\n
\n

<\/section>\n
\u8fd9\u91ccD\u662f\u51fa\u73b0\u5728\u6587\u7ae0\u4e2d\u7684\u90a3\u4e9b\u5355\u8bcd\u7ec4\u5408\uff0cDbar\u662f\u6240\u6709\u7684\u672a\u51fa\u73b0\u7684\uff0c\u4f46\u662f\u7531\u673a\u5668\u751f\u6210\u7684\u968f\u673a\u5355\u8bcd\u7ec4\u5408\u3002Theta\u5219\u662f\u5305\u542b\u8bcd\u5411\u91cf\u7684\u5f85\u4f18\u5316\u53c2\u6570\u3002<\/section>\n
 <\/section>\n
\u5b9e\u9a8c\u8bc1\u660e\uff0c\u91c7\u7528\u4e86\u65b0\u7684\u76ee\u6807\u51fd\u6570\u4ee5\u540e\uff0c\u6211\u4eec\u6574\u4e2a\u8ba1\u7b97\u8fc7\u7a0b\u4f1a\u975e\u5e38\u5feb\uff0c\u800c\u8bcd\u5411\u91cf\u5d4c\u5165\u7684\u6548\u679c\u4e5f\u5f88\u597d\u3002<\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\n
\n

<\/section>\n

<\/section>\n
\u56db\u3001\u8054\u60f3\uff1a\u4e0e\u7edf\u8ba1\u7269\u7406\u7684\u8054\u7cfb<\/span><\/strong> <\/section>\n
 <\/section>\n
\u4ece\u7b2c\u4e00\u6b21\u8bfe\u7a0b\u5bf9 word2vec \u7684\u201c\u524d\u4e16\u4eca\u751f\u201d\u7684\u5256\u6790\u6211\u4eec\u4f3c\u4e4e\u53ef\u770b\u51fa\uff1a\u8fd9\u662f\u4e00\u4e2a\u5de5\u4e1a\u754c\u81f3\u5c11\u662f\u5b9e\u9a8c\u5ba4\u7684 project \u4e2d\u5f88\u597d\u7528\u7684\u5de5\u5177\uff0c\u5b83\u7684\u539f\u7406\u5df2\u7ecf\u5f88\u6e05\u695a\u3002\u4f46\u662f\uff0c\u5173\u4e8e word2vec \u6216\u8005 \u8bf4\u5173\u4e8e distributed representation \u53ef\u4ee5\u505a\u7684\u5de5\u4f5c\u8fdc\u8fdc\u6ca1\u6709\u7ed3\u675f\uff0c\u5c24\u5176\u662f\u4ece\u7edf\u8ba1\u7269\u7406\u7684\u89d2\u5ea6\u800c\u8a00\u3002<\/section>\n
 <\/section>\n
\u7b2c I \u90e8\u5206\u7684\u6700\u540e\u4e00\u4e2a tips \u4e2d\u6211\u7b80\u5355\u63d0\u5230\u5982\u679c\u628a\u5206\u6bcd\u90fd\u7406\u89e3\u6210\u914d\u5206\u51fd\u6570\u7684\u8bdd\uff0c word2vec \u4e0e RBM \u7684\u5173\u7cfb\u3002\u4e0b\u9762\u6211\u5c06\u7528\u4e00\u90e8\u5206\u7684\u6587\u5b57\u66f4\u52a0\u8be6\u7ec6\u7684\u8bf4\u660e\u3002\u56e0\u4e3a\uff0c\u4e0e RBM \u7684\u8054\u7cfb\uff0c\u8fdc\u8fdc\u4e0d\u9650\u4e8e\u914d\u5206\u51fd\u6570\u3002<\/section>\n

<\/section>\n
\u6211\u770b\u5230 distributed representation \u7684\u7b2c\u4e00\u53cd\u5e94\uff0c\u5c31\u662f\u548c\u7edf\u8ba1\u7269\u7406\u53ca\u673a\u5668\u5b66\u4e60\u7b49\u90fd\u5bc6 \u5207\u76f8\u5173\u7684 representation learning\u3002 \u4e8b\u5b9e\u4e0a\uff0cBengio \u5728 2013 \u5e74\u53d1\u8868\u7684 review\uff1a\u300aRepresentation Learning: A Review and New Perspectives\u300b\u4e2d\u5c31\u5df2\u7ecf\u6307\u51fa\u4e86\u57fa\u4e8e\u80fd\u91cf\u7684\u6a21\u578b Boltzmann machine \u548c Restricted Boltzmann machine \u4e0e appropriate objectives for learning good representations \u4e4b\u95f4\u5343\u4e1d\u4e07\u7f15\u7684\u8054\u7cfb\u3002<\/section>\n

<\/section>\n
\u4eba\u4eec\u4e3a\u4e86\u7814\u7a76 representation \u7684\u79cd\u79cd\u8bad\u7ec3\u7b97\u6cd5\uff0c\u4ee5\u5f97\u5230\u66f4\u597d\u7684\u8bed\u8a00\u6a21\u578b,\u5f88\u591a\u5de5\u4f5c\u90fd \u5df2\u7ecf\u5f00\u5c55\u3002\u4f8b\u5982\uff0c\u4e3a\u4e86\u5efa\u7acb deep representations\uff0c Hinton \u548c\u5b66\u751f\u5728 2009 \u5e74\u5c31\u7814\u7a76\u4e86 Deep Boltzmann Machine(DBM)\uff0c\u8fd9\u91cc\u6240\u63d0\u5230\u7684 Boltzmann Machine\uff08BM\uff09 \u4e0e\u4e0a\u6587\u8bf4\u5230\u7684RBM \u4e00\u6837\uff0c\u90fd\u662f\u4e00\u4e2a\u57fa\u4e8e\u80fd\u91cf\u7684\u6a21\u578b\uff0c\u751a\u81f3\u53ef\u4ee5\u8bf4 RBM \u662f BM \u7684\u4e00\u4e2a\u7279\u4f8b\uff1a\u56e0\u4e3a\u5b83\u4eec\u4e24\u4e2a\u90fd\u662f\u4e24\u5c42\u795e\u7ecf\u7f51\u7edc\uff0c\u4f46\u662f BM \u662f\u5c42\u95f4\u548c\u5c42\u5185\u90fd\u5168\u8fde\u63a5\uff0c\u800c RBM \u662f\u5c42\u5185\u65e0\u8fde\u63a5\uff0c\u5c42\u95f4\u5168\u8fde\u63a5\u3002\u4e0b\u56fe\u5c55\u793a\u4e86 RBM \u7684\u7f51\u7edc\u6a21\u578b\u7ed3\u6784\u3002 <\/section>\n

<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\"\u96c6\u667a\"<\/section>\n<\/section>\n
\n
\n
\n
\u25b2<\/span>RBM \u7684\u7f51\u7edc\u7ed3\u6784<\/span><\/em><\/span>
<\/section>\n


<\/p>\n

\u8fd9\u91cc\uff0cW \u662f\u8fde\u63a5\u9690\u85cf\u5c42\u548c\u663e\u5c42\u4e4b\u95f4\u7684\u6743\u503c\u77e9\u9635\u3002<\/p>\n

 <\/section>\n
\u5f53\u7136\uff0c\u5f53\u5e74\u4e5f\u662f Hinton \u7b49\u4eba\u4e3a\u4e86\u8bad\u7ec3 RBM \u5e76\u63d0\u51fa\u4e86 CD-k \u7b97\u6cd5\uff0c\u5f53\u7136\uff0c\u8fd9\u662f\u53e6\u5916\u4e00\u4e2a\u5185\u5bb9\u4e86\u3002\u81f3\u4e8e\u6587\u732e\u4e2d\u5e38\u5e38\u80fd\u770b\u5230\u7684 Deep Belief Networks \uff08DBNs\uff09\u548c DBM \u4e00\u6837,\u90fd\u662f\u57fa\u4e8e RBM \u7684\u4e00\u79cd\u6269\u5c55\u3002\u8fd9\u91cc\uff0c\u6211\u4eec\u4e3b\u8981\u8bb2 distributed representation \u548c RBM \u7684\u5173\u7cfb\uff0c\u81f3\u4e8e DBN \u548c DBM\uff0c\u7559\u4e0b\u8282\u8bfe\u7684\u8bfe\u540e\u4f5c\u4e3a\u8111\u6d1e\u5206\u522b\u8bf4\u660e\u3002<\/section>\n

<\/section>\n

\u901a\u8fc7\u4e0a\u9762\u7684\u6574\u7406\u6211\u4eec\u770b\u51fa\uff0c\u65e0\u8bba\u662f word2vec \u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u7684\u76ee\u6807\u51fd\u6570\uff0c\u8fd8\u662f\u76ee\u524d\u5de5\u4e1a\u754c\u591a\u7528\u4e8e\u5206\u7c7b\u7684 RBM\uff08\u5f53\u7136\u5305\u62ec DBN\uff0cDBM\uff09\uff0c\u8fd8\u662f planted \u81ea\u65cb\u73bb\u7483\u6a21\u578b\uff0c\u6700\u540e\u90fd\u53ef\u4ee5\u5f52\u7ed3\u5230\u7edf\u8ba1\u63a8\u65ad\u95ee\u9898\u4e0a\u9762\u6765\u3002\u6240\u4ee5\uff0c\u662f\u4e0d\u662f\u53ef\u4ee5\u5c06\u5927\u5bb6\u7684\u5206\u6790\u65b9\u5f0f\u8fdb\u884c\u76f8\u4e92\u501f\u9274\uff1f\u5c06\u7edf\u8ba1\u7269\u7406\u4e2d\u6c42\u89e3\u914d\u5206\u51fd\u6570\u7684\u65b9\u5f0f\u7528\u5230\uff081\uff09\u5f0f\u4e2d\uff0c\u800c\u5c06\u8fd9\u91cc\u7684\u8d1f\u91c7\u6837\u7684\u60f3\u6cd5\u7528\u5230\u914d\u5206\u51fd\u6570\u7684\u6570\u503c\u6a21\u62df\u4e2d\uff0c\u5f53\u7136\uff0c\u8fd9\u91cc\uff0c\u5982\u679c\u5b9a\u4e49\u201c\u8d1f\u201d\u624d\u662f\u95ee\u9898\u7684\u6838\u5fc3\u3002 <\/span><\/p>\n


<\/p>\n


<\/p>\n

\uff08<\/span>END<\/span>\uff09<\/span><\/p>\n


<\/p>\n<\/section>\n<\/section>\n<\/section>\n

\n
\n
\n
\n
\n

<\/span><\/strong>
<\/p>\n

\u5173\u4e8e\u300a\u8d70\u8fd12050\uff1a\u6ce8\u610f\u529b\u3001\u4e92\u8054\u7f51\u4e0e\u4eba\u5de5\u667a\u80fd\u300b<\/span><\/strong><\/p>\n


<\/span><\/strong><\/p>\n

\"\u96c6\u667a\"
<\/span><\/strong><\/p>\n

\u25bc<\/p>\n

<\/p>\n
\"\u96c6\u667a\"
<\/section>\n
\"\u96c6\u667a\"\u2190<\/span>\u8bc6\u522b\u4e8c\u7ef4\u7801\u8d2d\u4e70<\/section>\n

<\/span><\/span><\/section>\n

\u968f\u7740\u4eba\u5de5\u667a\u80fd\u7a0b\u5e8fAlphaGo\u4ee54:1\u7684\u5927\u6bd4\u5206\u6218\u80dc\u4eba\u7c7b\u56f4\u68cb\u4e16\u754c\u51a0\u519b\u674e\u4e16\u77f3\uff0c\u673a\u5668\u5c06\u5f81\u670d\u4eba\u7c7b\u7684\u62c5\u5fe7\u6b63\u5728\u751a\u56a3\u5c18\u4e0a\u3002\u300a\u8d70\u8fd12050\uff1a\u6ce8\u610f\u529b\u3001\u4e92\u8054\u7f51\u4e0e\u4eba\u5de5\u667a\u80fd\u300b\u5219\u4e3a\u6211\u4eec\u63cf\u7ed8\u4e86\u4e00\u5e45\u4eba\u673a\u548c\u8c10\u5171\u751f\u3001\u534f\u540c\u6f14\u5316\u7684\u5168\u65b0\u573a\u666f\u3002\u5728\u770b\u5f97\u89c1\u7684\u672a\u6765\uff0c\u4eba\u7c7b\u5c06\u8d8a\u6765\u8d8a\u591a\u5730\u6c89\u6d78\u4e8e\u4e94\u82b1\u516b\u95e8\u7684\u865a\u62df\u4e16\u754c\u4ee5\u83b7\u53d6\u5404\u5f0f\u4f53\u9a8c\uff0c\u4e0e\u6b64\u540c\u65f6\uff0c\u6211\u4eec\u5c06\u5fc3\u7518\u60c5\u613f\u5730\u5c06\u81ea\u5df1\u7684\u6ce8\u610f\u529b\u6e90\u6e90\u4e0d\u65ad\u5730\u8f93\u5165\u7ed9\u673a\u5668\u4e16\u754c\u4ee5\u4fc3\u4f7f\u5b83\u4eec\u8fdb\u5316\u3002<\/span><\/p>\n


<\/p>\n

<\/span><\/p>\n

\u672c\u4e66\u4ece\u6ce8\u610f\u529b\u7684\u89d2\u5ea6\u89e3\u8bfb\u4e86\u5305\u62ec\u4e92\u8054\u7f51\u3001\u4eba\u5de5\u667a\u80fd\u3001\u4f17\u5305\u3001\u4eba\u7c7b\u8ba1\u7b97\u3001\u8ba1\u7b97\u673a\u6e38\u620f\u3001\u865a\u62df\u73b0\u5b9e\u5728\u5185\u7684\u6280\u672f\u9886\u57df\u53ca\u5176\u5bf9\u793e\u4f1a\u751f\u6d3b\u7684\u5f71\u54cd\uff0c\u8fd8\u521b\u9020\u6027\u5730\u63d0\u51fa\u4e86\u4e00\u7cfb\u5217\u5168\u65b0\u7684\u6982\u5ff5\uff1a\u5360\u610f\u7406\u8bba\u3001\u56fe\u7075\u673a-\u53c2\u4e0e\u8005\u6a21\u578b\u3001\u201c\u6e38\u620f\uff0b\u201d\u65f6\u4ee3\u3001\u610f\u672c\u5bb6\u3001\u81ea\u52a8\u6e38\u620f\u8bbe\u8ba1\u3001\u81ea\u52a8\u5316\u521b\u4e1a\u3001\u5360\u610f\u901a\u8d27\u3001\u8bb8\u613f\u6811\uff0c\u7b49\u7b49\u3002\u6240\u6709\u8fd9\u4e9b\u5c06\u4e3a\u6211\u4eec\u7406\u89e3\u6280\u672f\u4e0e\u4eba\u7c7b\u7684\u5173\u7cfb\u3001\u900f\u89c6\u4eba\u7c7b\u793e\u4f1a\u7684\u672a\u6765\u53d1\u5c55\u548c\u8d70\u5411\u63d0\u4f9b\u6df1\u523b\u7684\u6d1e\u5bdf\u3002<\/span><\/p>\n


<\/span><\/p>\n


<\/p>\n

\"\u96c6\u667a\"<\/p>\n


<\/p>\n

\"\u96c6\u667a\"<\/p>\n

\u8ba9\u82f9\u679c\u7838\u5f97\u66f4\u731b\u70c8\u4e9b\u5427\uff01\uff01\uff01<\/span><\/p>\n

\"\u96c6\u667a\"<\/span><\/p>\n


<\/p>\n

\n
\"\u96c6\u667a\"
<\/section>\n<\/section>\n

\u957f\u6309\u8bc6\u522b\u4e8c\u7ef4\u7801\uff0c\u5173\u6ce8\u96c6\u667aClub\uff0c<\/span><\/p>\n

\u8ba9\u6211\u4eec\u79bb\u79d1\u5b66\u63a2\u7d22\u66f4\u8fd1\u4e00\u6b65\u3002<\/span><\/p>\n


<\/span><\/p>\n


<\/span><\/p>\n

\u25bc\u70b9\u6b64\u4e0b\u8f7d\u4e91\u7aef\u8d44\u6599\uff0c\u5bc6\u7801\uff1aj7hu<\/span><\/span><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n

\n

\u59cb\u53d1\u4e8e\u5fae\u4fe1\u516c\u4f17\u53f7\uff1a \u96c6\u667a\u4ff1\u4e50\u90e8 <\/p>\n<\/blockquote>\n<\/div>\n","protected":false},"excerpt":{"rendered":"

2016\u5e747\u670821\u65e5  \uff08AndreaSunshine\u3001\u7504\u6167\u73b2\u3001\u7941\u8f89\u8bb0\u5f55\uff0cJake\u6574\u7406\u3001\u7efc\u5408\uff09 \u4e3b\u9898\uff1aWord2Vec:Basics and Model Natural Language Processing in the Era of Deep Learning \u4e3b\u8bb2\u4eba\uff1aYanran L…<\/p>\n","protected":false},"author":1,"featured_media":9851,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[19],"tags":[],"special":[],"_links":{"self":[{"href":"https:\/\/swarma.org\/index.php?rest_route=\/wp\/v2\/posts\/9850"}],"collection":[{"href":"https:\/\/swarma.org\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/swarma.org\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/swarma.org\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/swarma.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=9850"}],"version-history":[{"count":1,"href":"https:\/\/swarma.org\/index.php?rest_route=\/wp\/v2\/posts\/9850\/revisions"}],"predecessor-version":[{"id":9874,"href":"https:\/\/swarma.org\/index.php?rest_route=\/wp\/v2\/posts\/9850\/revisions\/9874"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/swarma.org\/index.php?rest_route=\/wp\/v2\/media\/9851"}],"wp:attachment":[{"href":"https:\/\/swarma.org\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=9850"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/swarma.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=9850"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/swarma.org\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=9850"},{"taxonomy":"special","embeddable":true,"href":"https:\/\/swarma.org\/index.php?rest_route=%2Fwp%2Fv2%2Fspecial&post=9850"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}