{"id":10020688,"date":"2025-03-26T21:56:22","date_gmt":"2025-03-26T13:56:22","guid":{"rendered":"https:\/\/itmax.cn\/?p=10020688"},"modified":"2025-03-26T21:56:22","modified_gmt":"2025-03-26T13:56:22","slug":"%e6%99%ba%e8%83%bd%e7%9f%a5%e8%af%86%e5%9b%be%e8%b0%b1%ef%bc%9a%e5%a4%a7%e6%a8%a1%e5%9e%8b%e5%a6%82%e4%bd%95%e5%ae%9e%e7%8e%b0%e9%ab%98%e6%95%88%e5%ae%9e%e4%bd%93%e8%af%86%e5%88%ab%e4%b8%8e%e5%85%b3","status":"publish","type":"post","link":"https:\/\/itmax.cn\/?p=10020688","title":{"rendered":"\u667a\u80fd\u77e5\u8bc6\u56fe\u8c31\uff1a\u5927\u6a21\u578b\u5982\u4f55\u5b9e\u73b0\u9ad8\u6548\u5b9e\u4f53\u8bc6\u522b\u4e0e\u5173\u7cfb\u62bd\u53d6"},"content":{"rendered":"<h2>\u6458\u8981<\/h2>\n<p>\u77e5\u8bc6\u56fe\u8c31\uff08Knowledge Graph, KG\uff09\u662f\u4eba\u5de5\u667a\u80fd\u9886\u57df\u7684\u91cd\u8981\u6280\u672f\u4e4b\u4e00\uff0c\u5e7f\u6cdb\u5e94\u7528\u4e8e\u641c\u7d22\u5f15\u64ce\u3001\u63a8\u8350\u7cfb\u7edf\u3001\u95ee\u7b54\u7cfb\u7edf\u7b49\u573a\u666f\u3002\u7136\u800c\uff0c\u4f20\u7edf\u77e5\u8bc6\u56fe\u8c31\u6784\u5efa\u4f9d\u8d56\u5927\u91cf\u4eba\u5de5\u6807\u6ce8\uff0c\u6210\u672c\u9ad8\u4e14\u6548\u7387\u4f4e\u3002\u8fd1\u5e74\u6765\uff0c\u968f\u7740\u5927\u6a21\u578b\uff08\u5982GPT\u3001BERT\u7b49\uff09\u7684\u5feb\u901f\u53d1\u5c55\uff0c\u5229\u7528\u5927\u6a21\u578b\u81ea\u52a8\u5316\u751f\u6210\u77e5\u8bc6\u56fe\u8c31\u6210\u4e3a\u53ef\u80fd\u3002\u672c\u6587\u5c06\u8be6\u7ec6\u8bb2\u89e3\u5982\u4f55\u5229\u7528\u5927\u6a21\u578b\u5b9e\u73b0\u77e5\u8bc6\u56fe\u8c31\u7684\u81ea\u52a8\u5316\u6784\u5efa\uff0c\u5305\u62ec\u5b9e\u4f53\u8bc6\u522b\u3001\u5173\u7cfb\u62bd\u53d6\u548c\u56fe\u8c31\u66f4\u65b0\uff0c\u5e76\u63d0\u4f9b\u53ef\u8fd0\u884c\u7684\u793a\u4f8b\u4ee3\u7801\u548c\u76f8\u5173\u914d\u56fe\u3002<\/p>\n<h2>\u5f15\u8a00<\/h2>\n<p>\u77e5\u8bc6\u56fe\u8c31\u662f\u4e00\u79cd\u4ee5\u56fe\u7ed3\u6784\u8868\u793a\u77e5\u8bc6\u7684\u6280\u672f\uff0c\u5176\u6838\u5fc3\u662f\u5b9e\u4f53\uff08Entity\uff09\u548c\u5b9e\u4f53\u4e4b\u95f4\u7684\u5173\u7cfb\uff08Relation\uff09\u3002\u4f20\u7edf\u77e5\u8bc6\u56fe\u8c31\u6784\u5efa\u9700\u8981\u5927\u91cf\u4eba\u5de5\u53c2\u4e0e\uff0c\u5305\u62ec\u6570\u636e\u6807\u6ce8\u3001\u5173\u7cfb\u5b9a\u4e49\u548c\u56fe\u8c31\u7ef4\u62a4\uff0c\u5bfc\u81f4\u6210\u672c\u9ad8\u3001\u6548\u7387\u4f4e\u3002\u5927\u6a21\u578b\uff08\u5982GPT\u3001BERT\u7b49\uff09\u51ed\u501f\u5176\u5f3a\u5927\u7684\u81ea\u7136\u8bed\u8a00\u7406\u89e3\u80fd\u529b\uff0c\u53ef\u4ee5\u663e\u8457\u63d0\u5347\u77e5\u8bc6\u56fe\u8c31\u6784\u5efa\u7684\u81ea\u52a8\u5316\u6c34\u5e73\u3002\u672c\u6587\u5c06\u63a2\u8ba8\u5982\u4f55\u5229\u7528\u5927\u6a21\u578b\u5b9e\u73b0\u77e5\u8bc6\u56fe\u8c31\u7684\u81ea\u52a8\u5316\u751f\u6210\uff0c\u5e76\u901a\u8fc7\u4ee3\u7801\u793a\u4f8b\u5c55\u793a\u5177\u4f53\u5b9e\u73b0\u65b9\u6cd5\u3002<\/p>\n<h2>\u5927\u6a21\u578b\u5728\u77e5\u8bc6\u56fe\u8c31\u6784\u5efa\u4e2d\u7684\u4f5c\u7528<\/h2>\n<p>\u5927\u6a21\u578b\u5728\u77e5\u8bc6\u56fe\u8c31\u6784\u5efa\u4e2d\u7684\u6838\u5fc3\u4f5c\u7528\u5305\u62ec\uff1a<\/p>\n<ul>\n<li><strong>\u5b9e\u4f53\u8bc6\u522b<\/strong>\uff1a\u4ece\u6587\u672c\u4e2d\u81ea\u52a8\u8bc6\u522b\u51fa\u5b9e\u4f53\u3002<\/li>\n<li><strong>\u5173\u7cfb\u62bd\u53d6<\/strong>\uff1a\u8bc6\u522b\u5b9e\u4f53\u4e4b\u95f4\u7684\u5173\u7cfb\u3002<\/li>\n<li><strong>\u56fe\u8c31\u66f4\u65b0<\/strong>\uff1a\u52a8\u6001\u66f4\u65b0\u77e5\u8bc6\u56fe\u8c31\u4e2d\u7684\u5b9e\u4f53\u548c\u5173\u7cfb\u3002<\/li>\n<\/ul>\n<h2>\u5b9e\u4f53\u8bc6\u522b<\/h2>\n<p>\u5b9e\u4f53\u8bc6\u522b\u662f\u77e5\u8bc6\u56fe\u8c31\u6784\u5efa\u7684\u7b2c\u4e00\u6b65\uff0c\u76ee\u6807\u662f\u8bc6\u522b\u6587\u672c\u4e2d\u7684\u5173\u952e\u5b9e\u4f53\uff08\u5982\u4eba\u540d\u3001\u5730\u540d\u3001\u7ec4\u7ec7\u7b49\uff09\u3002\u5927\u6a21\u578b\u53ef\u4ee5\u901a\u8fc7\u9884\u8bad\u7ec3\u7684\u8bed\u8a00\u6a21\u578b\uff08\u5982BERT\uff09\u5b9e\u73b0\u9ad8\u6548\u7684\u5b9e\u4f53\u8bc6\u522b\u3002<\/p>\n<h3>\u57fa\u4e8eBERT\u7684\u5b9e\u4f53\u8bc6\u522b<\/h3>\n<pre><code class=\"python\">from transformers import BertTokenizer, BertForTokenClassification\nimport torch\n\n# \u52a0\u8f7d\u9884\u8bad\u7ec3\u7684BERT\u6a21\u578b\u548c\u5206\u8bcd\u5668\ntokenizer = BertTokenizer.from_pretrained('bert-base-uncased')\nmodel = BertForTokenClassification.from_pretrained('bert-base-uncased', num_labels=9)\n\n# \u8f93\u5165\u6587\u672c\ntext = &quot;Apple is a technology company headquartered in Cupertino, California.&quot;\n\n# \u5206\u8bcd\u5e76\u8f6c\u6362\u4e3a\u6a21\u578b\u8f93\u5165\ninputs = tokenizer(text, return_tensors=&quot;pt&quot;, truncation=True, padding=True)\ntokens = tokenizer.convert_ids_to_tokens(inputs['input_ids'][0])\n\n# \u6a21\u578b\u9884\u6d4b\noutputs = model(**inputs).logits\npredictions = torch.argmax(outputs, dim=2)\n\n# \u8f93\u51fa\u5b9e\u4f53\u8bc6\u522b\u7ed3\u679c\nentities = []\nfor token, prediction in zip(tokens, predictions[0]):\n    if prediction != 0:  # \u8fc7\u6ee4\u975e\u5b9e\u4f53\n        entities.append((token, prediction.item()))\nprint(&quot;\u8bc6\u522b\u5230\u7684\u5b9e\u4f53\uff1a&quot;, entities)<\/code><\/pre>\n<h3>\u5b9e\u4f53\u8bc6\u522b\u6d41\u7a0b<\/h3>\n<p>+&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+<br \/>|   \u8f93\u5165\u6587\u672c         |       |   \u6587\u672c\u9884\u5904\u7406       |       |   \u6a21\u578b\u8f93\u5165         |<br \/>|                   | &#8212;-&gt; |                   | &#8212;-&gt; |                   |<br \/>| &#8220;Apple is a       |       | \u5206\u8bcd\u3001\u53bb\u505c\u7528\u8bcd\u7b49   |       | \u5c06\u6587\u672c\u7f16\u7801\u4e3a\u6a21\u578b    |<br \/>| technology company|       |                   |       | \u8f93\u5165\u683c\u5f0f           |<br \/>| headquartered in  |       |                   |       |                   |<br \/>| Cupertino,        |       |                   |       |                   |<br \/>| California.&#8221;      |       |                   |       |                   |<br \/>+&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+<\/p>\n<pre><code>                                                             |\n                                                             v<\/code><\/pre>\n<p>+&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+<br \/>|   \u5b9e\u4f53\u8bc6\u522b\u6a21\u578b     |       |   \u5b9e\u4f53\u6807\u6ce8         |       |   \u5b9e\u4f53\u63d0\u53d6         |<br \/>|                   | &#8212;-&gt; |                   | &#8212;-&gt; |                   |<br \/>| \u4f7f\u7528BERT\u7b49\u6a21\u578b\u8fdb\u884c |       | \u8f93\u51fa\u6bcf\u4e2a\u8bcd\u7684\u5b9e\u4f53   |       | \u63d0\u53d6\u5e76\u8fc7\u6ee4\u5b9e\u4f53     |<br \/>| \u5b9e\u4f53\u6807\u6ce8           |       | \u6807\u7b7e              |       |                   |<br \/>+&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+<\/p>\n<pre><code>                                                             |\n                                                             v<\/code><\/pre>\n<table>\n<thead>\n<tr>\n<th>\u8f93\u51fa\u5b9e\u4f53\u5217\u8868<\/th>\n<\/tr>\n<\/thead>\n<thead>\n<tr>\n<th> <\/th>\n<\/tr>\n<\/thead>\n<thead>\n<tr>\n<th>ORG: Apple<\/th>\n<\/tr>\n<\/thead>\n<thead>\n<tr>\n<th>LOC: Cupertino,<\/th>\n<\/tr>\n<\/thead>\n<thead>\n<tr>\n<th>California<\/th>\n<\/tr>\n<\/thead>\n<\/tbody>\n<\/table>\n<h2>\u7cfb\u62bd\u53d6<\/h2>\n<p>\u5173\u7cfb\u62bd\u53d6\u7684\u76ee\u6807\u662f\u4ece\u6587\u672c\u4e2d\u8bc6\u522b\u5b9e\u4f53\u4e4b\u95f4\u7684\u5173\u7cfb\u3002\u5927\u6a21\u578b\u53ef\u4ee5\u901a\u8fc7\u8054\u5408\u5b66\u4e60\uff08Joint Learning\uff09\u6216\u7ba1\u9053\u5f0f\u65b9\u6cd5\uff08Pipeline\uff09\u5b9e\u73b0\u5173\u7cfb\u62bd\u53d6\u3002<\/p>\n<h3>\u57fa\u4e8e GPT \u7684\u5173\u7cfb\u62bd\u53d6<\/h3>\n<pre><code class=\"python\">from transformers import GPT2Tokenizer, GPT2LMHeadModel\n\n# \u52a0\u8f7d\u9884\u8bad\u7ec3\u7684GPT\u6a21\u578b\u548c\u5206\u8bcd\u5668\ntokenizer = GPT2Tokenizer.from_pretrained('gpt2')\nmodel = GPT2LMHeadModel.from_pretrained('gpt2')\n\n# \u8f93\u5165\u6587\u672c\ntext = &quot;Steve Jobs founded Apple in 1976.&quot;\n\n# \u751f\u6210\u5173\u7cfb\u62bd\u53d6\u63d0\u793a\nprompt = f&quot;Extract relations from the text: {text}\/nRelations:&quot;\n\n# \u6a21\u578b\u751f\u6210\ninputs = tokenizer(prompt, return_tensors=&quot;pt&quot;, truncation=True, padding=True)\noutputs = model.generate(**inputs, max_length=50)\n\n# \u89e3\u7801\u8f93\u51fa\nrelations = tokenizer.decode(outputs[0], skip_special_tokens=True)\nprint(&quot;\u62bd\u53d6\u5230\u7684\u5173\u7cfb\uff1a&quot;, relations)<\/code><\/pre>\n<h3>\u5173\u7cfb\u62bd\u53d6\u6d41\u7a0b<\/h3>\n<p>+&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+<br \/>|   \u8f93\u5165\u6587\u672c         |       |   \u6587\u672c\u9884\u5904\u7406       |       |   \u5b9e\u4f53\u8bc6\u522b         |<br \/>|                   | &#8212;-&gt; |                   | &#8212;-&gt; |                   |<br \/>| &#8220;Steve Jobs founded|       | \u5206\u8bcd\u3001\u53bb\u505c\u7528\u8bcd\u7b49   |       | \u8bc6\u522b\u5b9e\u4f53\uff1aSteve Jobs|<br \/>| Apple in 1976.&#8221;    |       |                   |       | \u548c Apple           |<br \/>+&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+<\/p>\n<pre><code>                                                             |\n                                                             v<\/code><\/pre>\n<p>+&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+<br \/>|   \u5173\u7cfb\u62bd\u53d6\u6a21\u578b     |       |   \u5173\u7cfb\u5206\u7c7b         |       |   \u8f93\u51fa\u5173\u7cfb\u4e09\u5143\u7ec4   |<br \/>|                   | &#8212;-&gt; |                   | &#8212;-&gt; |                   |<br \/>| \u751f\u6210\u5019\u9009\u5173\u7cfb       |       | \u5206\u7c7b\u5173\u7cfb\uff1afounded  |       | (Steve Jobs,       |<br \/>|                   |       |                   |       |  founded, Apple)   |<br \/>+&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+<\/p>\n<h2>\u56fe\u8c31\u66f4\u65b0<\/h2>\n<p>\u77e5\u8bc6\u56fe\u8c31\u9700\u8981\u52a8\u6001\u66f4\u65b0\u4ee5\u53cd\u6620\u6700\u65b0\u7684\u77e5\u8bc6\u3002\u5927\u6a21\u578b\u53ef\u4ee5\u901a\u8fc7\u589e\u91cf\u5b66\u4e60\u6216\u5728\u7ebf\u5b66\u4e60\u5b9e\u73b0\u56fe\u8c31\u7684\u5b9e\u65f6\u66f4\u65b0\u3002<\/p>\n<h3>\u57fa\u4e8e\u589e\u91cf\u5b66\u4e60\u7684\u56fe\u8c31\u66f4\u65b0<\/h3>\n<pre><code class=\"python\">from sklearn.feature_extraction.text import TfidfVectorizer\nfrom sklearn.cluster import KMeans\n\n# \u5047\u8bbe\u5df2\u6709\u77e5\u8bc6\u56fe\u8c31\u6570\u636e\nexisting_entities = [&quot;Apple&quot;, &quot;Steve Jobs&quot;, &quot;Cupertino&quot;]\nnew_entities = [&quot;Tim Cook&quot;, &quot;iPhone&quot;, &quot;California&quot;]\n\n# \u5408\u5e76\u65b0\u65e7\u5b9e\u4f53\nall_entities = existing_entities + new_entities\n\n# \u4f7f\u7528TF-IDF\u5411\u91cf\u5316\nvectorizer = TfidfVectorizer()\nX = vectorizer.fit_transform(all_entities)\n\n# \u805a\u7c7b\u4ee5\u53d1\u73b0\u65b0\u5173\u7cfb\nkmeans = KMeans(n_clusters=2)\nkmeans.fit(X)\n\n# \u8f93\u51fa\u805a\u7c7b\u7ed3\u679c\nprint(&quot;\u5b9e\u4f53\u805a\u7c7b\u7ed3\u679c\uff1a&quot;, kmeans.labels_)<\/code><\/pre>\n<h3>\u56fe\u8c31\u66f4\u65b0\u6d41\u7a0b<\/h3>\n<p>+&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+<br \/>|   \u8f93\u5165\u65b0\u6570\u636e       |       |   \u6570\u636e\u9884\u5904\u7406       |       |   \u5b9e\u4f53\u8bc6\u522b         |<br \/>|                   | &#8212;-&gt; |                   | &#8212;-&gt; |                   |<br \/>| &#8220;Tim Cook is the  |       | \u6e05\u6d17\u3001\u5206\u8bcd\u3001\u53bb\u505c\u7528\u8bcd|       | \u8bc6\u522b\u5b9e\u4f53\uff1aTim Cook |<br \/>| CEO of Apple.&#8221;    |       | \u7b49\u64cd\u4f5c             |       | \u548c Apple           |<br \/>+&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+<\/p>\n<pre><code>                                                             |\n                                                             v<\/code><\/pre>\n<p>+&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+<br \/>|   \u5173\u7cfb\u62bd\u53d6         |       |   \u77e5\u8bc6\u56fe\u8c31\u67e5\u8be2     |       |   \u51b2\u7a81\u68c0\u6d4b\u4e0e\u6d88\u89e3   |<br \/>|                   | &#8212;-&gt; |                   | &#8212;-&gt; |                   |<br \/>| \u62bd\u53d6\u5173\u7cfb\uff1a         |       | \u68c0\u67e5\u65b0\u5b9e\u4f53\u548c\u5173\u7cfb   |       | \u68c0\u6d4b\u51b2\u7a81\u5e76\u6d88\u89e3     |<br \/>| (Tim Cook, CEO of, |       | \u662f\u5426\u5df2\u5b58\u5728         |       |                   |<br \/>| Apple)            |       |                   |       |                   |<br \/>+&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+<\/p>\n<pre><code>                                                             |\n                                                             v<\/code><\/pre>\n<p>+&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+<br \/>|   \u56fe\u8c31\u66f4\u65b0         |       |   \u8f93\u51fa\u66f4\u65b0\u540e\u7684\u77e5\u8bc6\u56fe\u8c31 |<br \/>|                   | &#8212;-&gt; |                   |<br \/>| \u6dfb\u52a0\u65b0\u5b9e\u4f53\u548c\u5173\u7cfb   |       | \u5305\u542b\u65b0\u589e\u7684\u5b9e\u4f53\u548c\u5173\u7cfb |<br \/>+&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+       +&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;-+<\/p>\n<h2>QA\u73af\u8282<\/h2>\n<p><strong>Q1: \u5927\u6a21\u578b\u5728\u77e5\u8bc6\u56fe\u8c31\u6784\u5efa\u4e2d\u7684\u4f18\u52bf\u662f\u4ec0\u4e48\uff1f<\/strong>  <br \/>A1: \u5927\u6a21\u578b\u5177\u6709\u5f3a\u5927\u7684\u81ea\u7136\u8bed\u8a00\u7406\u89e3\u80fd\u529b\uff0c\u53ef\u4ee5\u663e\u8457\u63d0\u5347\u5b9e\u4f53\u8bc6\u522b\u548c\u5173\u7cfb\u62bd\u53d6\u7684\u51c6\u786e\u6027\u548c\u6548\u7387\uff0c\u51cf\u5c11\u4eba\u5de5\u5e72\u9884\u3002<\/p>\n<p><strong>Q2: \u5982\u4f55\u5904\u7406\u591a\u8bed\u8a00\u77e5\u8bc6\u56fe\u8c31\u7684\u6784\u5efa\uff1f<\/strong>  <br \/>A2: \u53ef\u4ee5\u4f7f\u7528\u591a\u8bed\u8a00\u5927\u6a21\u578b\uff08\u5982mBERT\u3001XLM-R\uff09\u6765\u5904\u7406\u591a\u8bed\u8a00\u6587\u672c\uff0c\u5e76\u901a\u8fc7\u7edf\u4e00\u7684\u8868\u793a\u65b9\u6cd5\u6784\u5efa\u591a\u8bed\u8a00\u77e5\u8bc6\u56fe\u8c31\u3002<\/p>\n<h2>\u603b\u7ed3<\/h2>\n<p>\u672c\u6587\u8be6\u7ec6\u4ecb\u7ecd\u4e86\u5982\u4f55\u5229\u7528\u5927\u6a21\u578b\u6784\u5efa\u9ad8\u6548\u7684\u77e5\u8bc6\u56fe\u8c31\uff0c\u5305\u62ec\u5b9e\u4f53\u8bc6\u522b\u3001\u5173\u7cfb\u62bd\u53d6\u548c\u56fe\u8c31\u66f4\u65b0\u3002\u901a\u8fc7\u5927\u6a21\u578b\u7684\u81ea\u52a8\u5316\u80fd\u529b\uff0c\u53ef\u4ee5\u663e\u8457\u964d\u4f4e\u77e5\u8bc6\u56fe\u8c31\u6784\u5efa\u7684\u6210\u672c\u548c\u590d\u6742\u5ea6\u3002\u4ee3\u7801\u793a\u4f8b\u5c55\u793a\u4e86\u5177\u4f53\u5b9e\u73b0\u65b9\u6cd5\uff0c\u4e3a\u8bfb\u8005\u63d0\u4f9b\u4e86\u5b9e\u8df5\u53c2\u8003\u3002<\/p>\n<p>\u672a\u6765\uff0c\u968f\u7740\u5927\u6a21\u578b\u6280\u672f\u7684\u8fdb\u4e00\u6b65\u53d1\u5c55\uff0c\u77e5\u8bc6\u56fe\u8c31\u6784\u5efa\u5c06\u66f4\u52a0\u667a\u80fd\u5316\u548c\u81ea\u52a8\u5316\u3002\u4ee5\u4e0b\u662f\u4e00\u4e9b\u503c\u5f97\u63a2\u7d22\u7684\u65b9\u5411\uff1a<\/p>\n<ol>\n<li><strong>\u591a\u6a21\u6001\u77e5\u8bc6\u56fe\u8c31<\/strong>\uff1a\u7ed3\u5408\u6587\u672c\u3001\u56fe\u50cf\u3001\u89c6\u9891\u7b49\u591a\u6a21\u6001\u6570\u636e\u6784\u5efa\u77e5\u8bc6\u56fe\u8c31\u3002<\/li>\n<li><strong>\u5b9e\u65f6\u56fe\u8c31\u66f4\u65b0<\/strong>\uff1a\u5229\u7528\u5728\u7ebf\u5b66\u4e60\u6280\u672f\u5b9e\u73b0\u77e5\u8bc6\u56fe\u8c31\u7684\u5b9e\u65f6\u66f4\u65b0\u3002<\/li>\n<li><strong>\u77e5\u8bc6\u56fe\u8c31\u4e0e\u751f\u6210\u5f0fAI\u7ed3\u5408<\/strong>\uff1a\u5c06\u77e5\u8bc6\u56fe\u8c31\u4e0e\u751f\u6210\u5f0fAI\uff08\u5982ChatGPT\uff09\u7ed3\u5408\uff0c\u63d0\u5347\u95ee\u7b54\u7cfb\u7edf\u548c\u63a8\u8350\u7cfb\u7edf\u7684\u6027\u80fd\u3002<\/li>\n<\/ol>\n<h2>\u53c2\u8003\u8d44\u6599<\/h2>\n<ol>\n<li>Devlin, J., et al. &#8220;BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding.&#8221; arXiv:1810.04805, 2018.<\/li>\n<li>Radford, A., et al. &#8220;Language Models are Few-Shot Learners.&#8221; arXiv:2005.14165, 2020.<\/li>\n<li>Wang, Z., et al. &#8220;Knowledge Graph Embedding: A Survey of Approaches and Applications.&#8221; IEEE Transactions on Knowledge and Data Engineering, 2017.<\/li>\n<\/ol>\n<p><a href=\"https:\/\/segmentfault.com\/a\/1190000046386371\">https:\/\/segmentfault.com\/a\/1190000046386371<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u6458\u8981 \u77e5\u8bc6\u56fe\u8c31\uff08Knowledge Graph, KG\uff09\u662f\u4eba\u5de5\u667a\u80fd\u9886\u57df\u7684\u91cd\u8981\u6280\u672f\u4e4b\u4e00\uff0c\u5e7f\u6cdb\u5e94\u7528\u4e8e\u641c\u7d22\u5f15\u64ce\u3001\u63a8\u8350 [&hellip;]<\/p>\n","protected":false},"author":7,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[4],"tags":[],"class_list":["post-10020688","post","type-post","status-publish","format-standard","hentry","category-share"],"jetpack_sharing_enabled":true,"jetpack_featured_media_url":"","_links":{"self":[{"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/posts\/10020688","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/users\/7"}],"replies":[{"embeddable":true,"href":"https:\/\/itmax.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=10020688"}],"version-history":[{"count":0,"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/posts\/10020688\/revisions"}],"wp:attachment":[{"href":"https:\/\/itmax.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=10020688"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/itmax.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=10020688"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/itmax.cn\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=10020688"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}