{"id":16881,"date":"2025-01-01T08:49:11","date_gmt":"2025-01-01T00:49:11","guid":{"rendered":"https:\/\/www.aisharenet.com\/?p=16881"},"modified":"2025-01-01T08:55:02","modified_gmt":"2025-01-01T00:55:02","slug":"ruheweiragyingyongben","status":"publish","type":"post","link":"https:\/\/www.kdjingpai.com\/de\/ruheweiragyingyongben\/","title":{"rendered":"\u5982\u4f55\u4e3aRAG\u5e94\u7528\u9009\u62e9\u6700\u4f73Embedding\u6a21\u578b"},"content":{"rendered":"<p><img decoding=\"async\" src=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/01\/824b2d6903d8ac0.png\" \/><\/p>\n<p>\u5411\u91cfEmbedding\u662f\u76ee\u524d\u68c0\u7d22\u589e\u5f3a\u751f\u6210\uff08RAG\uff09\u5e94\u7528\u7a0b\u5e8f\u7684\u6838\u5fc3\u3002\u5b83\u4eec\u6355\u83b7\u6570\u636e\u5bf9\u8c61\uff08\u5982\u6587\u672c\uff0c\u56fe\u50cf\u7b49\uff09\u7684\u8bed\u4e49\u4fe1\u606f\uff0c\u5e76\u4ee5\u6570\u5b57\u6570\u7ec4\u8868\u793a\u3002\u5728\u65f6\u4e0b\u7684\u751f\u6210\u5f0fAI\u5e94\u7528\u4e2d\uff0c\u8fd9\u4e9b\u5411\u91cfEmbedding\u901a\u5e38\u7531Embedding\u6a21\u578b\u751f\u6210\u3002\u5982\u4f55\u4e3aRAG\u5e94\u7528\u7a0b\u5e8f\u9009\u62e9\u5408\u9002\u7684Embedding\u6a21\u578b\u5462\uff1f\u603b\u4f53\u6765\u8bf4\uff0c\u8fd9\u53d6\u51b3\u4e8e\u5177\u4f53\u7528\u4f8b\u4ee5\u53ca\u5177\u4f53\u9700\u6c42\u3002\u63a5\u4e0b\u6765\uff0c\u8ba9\u6211\u4eec\u62c6\u5206\u6b65\u9aa4\u6765\u5206\u522b\u6765\u770b\u3002<\/p>\n<p>&nbsp;<\/p>\n<h2>01.\u786e\u5b9a\u5177\u4f53\u7528\u4f8b<\/h2>\n<p>\u6211\u4eec\u57fa\u4e8eRAG\u5e94\u7528\u7a0b\u5e8f\u9700\u6c42\u8003\u8651\u4ee5\u4e0b\u95ee\u9898\uff1a<\/p>\n<p>\u9996\u5148\uff0c\u901a\u7528\u6a21\u578b\u662f\u5426\u8db3\u4ee5\u6ee1\u8db3\u9700\u6c42\uff1f<\/p>\n<p>\u5176\u6b21\uff0c\u662f\u5426\u6709\u7279\u5b9a\u7684\u9700\u6c42\uff1f\u4f8b\u5982\u6a21\u6001\uff08\u5982\u4ec5\u6587\u672c\u6216\u56fe\u7247\uff0c\u5173\u4e8e\u591a\u6a21\u6001Embedding\u9009\u62e9\u53c2\u89c1\u300a<a href=\"https:\/\/www.kdjingpai.com\/ruhexuanzeheshiai\/\">\u5982\u4f55\u9009\u62e9\u5408\u9002\u7684 Embedding \u6a21\u578b<\/a>\u300b\uff09\u3001\u7279\u5b9a\u9886\u57df\uff08\u5982\u6cd5\u5f8b\u3001\u533b\u5b66\u7b49\uff09<\/p>\n<p>\u5728\u5927\u591a\u6570\u60c5\u51b5\u4e0b\uff0c\u9488\u5bf9\u6240\u9700\u6a21\u6001\uff0c\u901a\u5e38\u9009\u62e9\u901a\u7528\u6a21\u578b\u3002<\/p>\n<p>&nbsp;<\/p>\n<h2>02.\u9009\u62e9\u901a\u7528\u6a21\u578b<\/h2>\n<p>\u5982\u4f55\u9009\u62e9\u901a\u7528\u6a21\u578b\u5462\uff1fHuggingFace\u4e2dMassive Text Embedding Benchmark\uff08MTEB\uff09\u6392\u884c\u699c\u7f57\u5217\u4e86\u5f53\u524d\u5404\u79cd\u4e13\u6709\u548c\u5f00\u6e90\u6587\u672cEmbedding\u6a21\u578b\uff0c\u5bf9\u4e8e\u6bcf\u4e2aEmbedding\u6a21\u578b\uff0cMTEB\u5217\u51fa\u4e86\u5404\u79cd\u6307\u6807\uff0c\u5305\u62ec\u6a21\u578b\u53c2\u6570\u3001\u5185\u5b58\u3001Embedding\u7ef4\u5ea6\u3001\u6700\u5927token\u6570\u91cf\uff0c\u4ee5\u53ca\u5176\u5728\u68c0\u7d22\u3001\u6458\u8981\u7b49\u4efb\u52a1\u4e2d\u7684\u5f97\u5206\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/01\/468ee6cc688904c.png\" \/><\/p>\n<p>\u5728\u4e3aRAG\u5e94\u7528\u7a0b\u5e8f\u9009\u62e9Embedding\u6a21\u578b\u7684\u65f6\u5019\uff0c\u9700\u8981\u8003\u8651\u4ee5\u4e0b\u56e0\u7d20\uff1a<\/p>\n<p><strong>\u4efb\u52a1<\/strong>\uff1a\u5728MTEB\u6392\u884c\u699c\u9876\u90e8\uff0c\u6211\u4eec\u4f1a\u770b\u5230\u5404\u79cd\u4efb\u52a1\u9009\u9879\u5361\u3002\u5bf9\u4e8e\u4e00\u4e2aRAG\u5e94\u7528\u7a0b\u5e8f\uff0c\u6211\u4eec\u53ef\u80fd\u9700\u8981\u66f4\u5173\u6ce8\u201c\u68c0\u7d22\u201d\u4efb\u52a1\uff0c\u6211\u4eec\u53ef\u4ee5\u9009\u62e9 <code>Retrial<\/code> \u8fd9\u4e2a\u9009\u9879\u5361\u3002<\/p>\n<p><strong>\u8bed\u8a00<\/strong>\uff1a\u57fa\u4e8eRAG\u5e94\u7528\u7684\u6570\u636e\u96c6\u8bed\u8a00\u6765\u9009\u62e9\u5bf9\u5e94\u8bed\u8a00\u7684Embedding\u6a21\u578b\u3002<\/p>\n<p><strong>\u5f97\u5206<\/strong>\uff1a\u8868\u793a\u6a21\u578b\u5728\u7279\u5b9a\u57fa\u51c6\u6570\u636e\u96c6\u6216\u591a\u4e2a\u57fa\u51c6\u6570\u636e\u96c6\u4e0a\u7684\u8868\u73b0\u3002\u6839\u636e\u4efb\u52a1\u7684\u4e0d\u540c\uff0c\u4f7f\u7528\u4e0d\u540c\u7684\u8bc4\u4f30\u6307\u6807\u3002\u901a\u5e38\uff0c\u8fd9\u4e9b\u6307\u6807\u7684\u53d6\u503c\u8303\u56f4\u57280\u52301\u4e4b\u95f4\uff0c\u503c\u8d8a\u9ad8\u8868\u793a\u6027\u80fd\u8d8a\u597d\u3002<\/p>\n<p><strong>\u6a21\u578b\u89c4\u6a21\u548c\u5185\u5b58\u7528\u91cf<\/strong>\uff1a\u8fd9\u4e9b\u6307\u6807\u8ba9\u6211\u4eec\u4e86\u89e3\u6a21\u578b\u8fd0\u884c\u65f6\u6240\u9700\u7684\u8ba1\u7b97\u8d44\u6e90\u3002\u867d\u7136\u68c0\u7d22\u6027\u80fd\u968f\u7740\u6a21\u578b\u89c4\u6a21\u7684\u589e\u52a0\u800c\u63d0\u9ad8\uff0c\u4f46\u9700\u8981\u6ce8\u610f\u7684\u662f\uff0c\u6a21\u578b\u89c4\u6a21\u4e5f\u76f4\u63a5\u5f71\u54cd\u5ef6\u8fdf\u3002\u6b64\u5916\uff0c\u8f83\u5927\u7684\u6a21\u578b\u53ef\u80fd\u4f1a\u8fc7\u62df\u5408\uff0c\u5176\u6cdb\u5316\u6027\u80fd\u4f4e\uff0c\u4ece\u800c\u5728\u751f\u4ea7\u4e2d\u8868\u73b0\u4e0d\u4f73\u3002\u56e0\u6b64\uff0c\u6211\u4eec\u9700\u8981\u5728\u751f\u4ea7\u73af\u5883\u4e2d\u5bfb\u6c42\u6027\u80fd\u548c\u5ef6\u8fdf\u4e4b\u95f4\u7684\u5e73\u8861\u3002\u4e00\u822c\u6765\u8bf4\uff0c\u6211\u4eec\u53ef\u4ee5\u5148\u4ece\u4e00\u4e2a\u5c0f\u7684\u3001\u8f7b\u91cf\u7ea7\u7684\u6a21\u578b\u5f00\u59cb\uff0c\u5148\u5feb\u901f\u6784\u5efaRAG\u5e94\u7528\u3002\u5728\u5e94\u7528\u7684\u57fa\u7840\u6d41\u7a0b\u6b63\u5e38\u8fd0\u884c\u4e4b\u540e\uff0c\u6211\u4eec\u53ef\u4ee5\u5207\u6362\u5230\u66f4\u5927\u3001\u6027\u80fd\u66f4\u9ad8\u7684\u6a21\u578b\uff0c\u6765\u5bf9\u5e94\u7528\u505a\u8fdb\u4e00\u6b65\u4f18\u5316\u3002<\/p>\n<p><strong>Embedding\u7ef4\u5ea6<\/strong>\uff1a\u8fd9\u662fEmbedding\u5411\u91cf\u7684\u957f\u5ea6\u3002\u867d\u7136\u66f4\u5927\u7684Embedding\u7ef4\u5ea6\u53ef\u4ee5\u6355\u83b7\u6570\u636e\u4e2d\u66f4\u7ec6\u5fae\u7684\u7ec6\u8282\uff0c\u4f46\u662f\u6548\u679c\u4e0d\u4e00\u5b9a\u662f\u6700\u4f73\u3002\u4f8b\u5982\uff0c\u6211\u4eec\u662f\u5426\u771f\u7684\u9700\u89818192\u7ef4\u5ea6\u6765\u5904\u7406\u6587\u6863\u6570\u636e\u5462\uff1f\u53ef\u80fd\u4e0d\u9700\u8981\u3002\u53e6\u4e00\u65b9\u9762\uff0c\u8f83\u5c0f\u7684Embedding\u7ef4\u5ea6\u63d0\u4f9b\u4e86\u66f4\u5feb\u7684\u63a8\u7406\u901f\u5ea6\uff0c\u5e76\u4e14\u5728\u5b58\u50a8\u548c\u5185\u5b58\u65b9\u9762\u66f4\u9ad8\u6548\u3002\u56e0\u6b64\uff0c\u6211\u4eec\u9700\u8981\u5728\u6355\u83b7\u6570\u636e\u5185\u5bb9\u548c\u6267\u884c\u6548\u7387\u4e4b\u95f4\u5bfb\u6c42\u826f\u597d\u7684\u5e73\u8861\u3002<\/p>\n<p><strong>\u6700\u5927token\u6570\u91cf<\/strong>\uff1a\u8868\u793a\u5355\u4e2aEmbedding\u7684\u6700\u5927token\u6570\u91cf\u3002\u5bf9\u4e8e\u5e38\u89c1\u7684RAG\u5e94\u7528\u7a0b\u5e8f\uff0cEmbedding\u8f83\u597d\u7684\u5206\u5757\u5927\u5c0f\u901a\u5e38\u662f\u5355\u4e2a\u6bb5\u843d\uff0c\u6b64\u65f6\uff0c\u6700\u5927token\u4e3a512\u7684Embedding\u6a21\u578b\u5e94\u8be5\u8db3\u591f\u3002\u7136\u800c\uff0c\u5728\u67d0\u4e9b\u7279\u6b8a\u60c5\u51b5\u4e0b\uff0c\u6211\u4eec\u53ef\u80fd\u9700\u8981token\u6570\u91cf\u66f4\u5927\u7684\u6a21\u578b\u6765\u5904\u7406\u66f4\u957f\u7684\u6587\u672c\u3002<\/p>\n<p>&nbsp;<\/p>\n<h2>03.\u5728RAG\u5e94\u7528\u4e2d\u8bc4\u4f30\u6a21\u578b<\/h2>\n<p>\u867d\u7136\u6211\u4eec\u53ef\u4ee5\u4eceMTEB\u6392\u884c\u699c\u627e\u5230\u901a\u7528\u6a21\u578b\uff0c\u4f46\u6211\u4eec\u9700\u8981\u8c28\u614e\u5bf9\u5f85\u5176\u7ed3\u679c\u3002\u8c28\u8bb0\uff0c\u8fd9\u4e9b\u7ed3\u679c\u662f\u6a21\u578b\u7684\u81ea\u6211\u62a5\u544a\uff0c\u6709\u53ef\u80fd\u67d0\u4e9b\u6a21\u578b\u4ea7\u751f\u4e86\u5938\u5927\u5176\u6027\u80fd\u7684\u5206\u6570\uff0c\u56e0\u4e3a\u5b83\u4eec\u53ef\u80fd\u5728\u8bad\u7ec3\u6570\u636e\u4e2d\u5305\u542b\u4e86MTEB\u6570\u636e\u96c6\uff0c\u6bd5\u7adf\u8fd9\u4e9b\u662f\u516c\u5f00\u7684\u6570\u636e\u96c6\u3002\u53e6\u5916\uff0c\u6a21\u578b\u4f7f\u7528\u57fa\u51c6\u7684\u6570\u636e\u96c6\u53ef\u80fd\u65e0\u6cd5\u51c6\u786e\u4ee3\u8868\u6211\u4eec\u5e94\u7528\u91cc\u4f7f\u7528\u7684\u6570\u636e\u3002\u56e0\u6b64\uff0c\u6211\u4eec\u9700\u8981\u5728\u81ea\u5df1\u7684\u6570\u636e\u96c6\u4e0a\u8bc4\u4f30Embedding\u6a21\u578b\u3002<\/p>\n<p><strong>3.1 \u6570\u636e\u96c6<\/strong><\/p>\n<p>\u6211\u4eec\u53ef\u4ee5\u4eceRAG\u5e94\u7528\u4f7f\u7528\u7684\u6570\u636e\u4e2d\uff0c\u751f\u6210\u4e00\u4e2a\u5c0f\u578b\u6807\u8bb0\u6570\u636e\u96c6\u3002\u6211\u4eec\u4ee5\u5982\u4e0b\u6570\u636e\u96c6\u4e3a\u4f8b\u3002<\/p>\n<table>\n<thead>\n<tr>\n<th>Language<\/th>\n<th>Description<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>C\/C++<\/td>\n<td>A general-purpose programming language known for its performance and efficiency. It provides low-level memory manipulation capabilities and is widely used in system\/software development, game development, and applications requiring high performance.<\/td>\n<\/tr>\n<tr>\n<td>Java<\/td>\n<td>A versatile, object-oriented programming language designed to have as few implementation dependencies as possible. It is widely used for building enterprise-scale applications, mobile applications (especially Android), and web applications due to its portability and robustness.<\/td>\n<\/tr>\n<tr>\n<td>Python<\/td>\n<td>A high-level, interpreted programming language known for its readability and simplicity. It supports multiple programming paradigms and is widely used in web development, data analysis, artificial intelligence, scientific computing, and automation.<\/td>\n<\/tr>\n<tr>\n<td>JavaScript<\/td>\n<td>A high-level, dynamic programming language primarily used for creating interactive and dynamic content on the web. It is an essential technology for front-end web development and is increasingly used on the server-side with environments like Node.js.<\/td>\n<\/tr>\n<tr>\n<td>C#<\/td>\n<td>A modern, object-oriented programming language developed by Microsoft. It is used for developing a wide range of applications, including web, desktop, mobile, and games, particularly within the Microsoft ecosystem.<\/td>\n<\/tr>\n<tr>\n<td>SQL<\/td>\n<td>A domain-specific language used in programming and managing relational databases. It is essential for querying, updating, and managing data in databases, and is widely used in data analysis and business intelligence.<\/td>\n<\/tr>\n<tr>\n<td>PHP<\/td>\n<td>A server-side scripting language designed primarily for web development. It is embedded into HTML and is widely used for building dynamic web pages and applications, with a strong presence in content management systems like WordPress.<\/td>\n<\/tr>\n<tr>\n<td>Golang<\/td>\n<td>A statically typed, compiled programming language designed by Google. Known for its simplicity and efficiency, it is used for building scalable and high-performance applications, particularly in cloud services and distributed systems.<\/td>\n<\/tr>\n<tr>\n<td>Rust<\/td>\n<td>A systems programming language focused on safety and concurrency. It provides memory safety without using a garbage collector and is used for building reliable and efficient software, particularly in systems programming and web assembly.<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p><strong>3.2 \u521b\u5efaEmbedding<\/strong><\/p>\n<p>\u63a5\u4e0b\u6765\uff0c\u6211\u4eec\u91c7\u7528<code>pymilvus[model]<\/code>\u5bf9\u4e8e\u4e0a\u8ff0\u6570\u636e\u96c6\u751f\u6210\u76f8\u5e94\u7684\u5411\u91cfEmbedding\u3002\u5173\u4e8e <code>pymilvus[model]<\/code> \u4f7f\u7528\uff0c\u53c2\u89c1https:\/\/milvus.io\/blog\/introducing-pymilvus-integrations-with-embedding-models.md<\/p>\n<pre>def\u00a0gen_embedding(model_name):\r\n\u00a0\u00a0\u00a0\u00a0openai_ef\u00a0=\u00a0model.dense.OpenAIEmbeddingFunction(\r\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0model_name=model_name,\u00a0\r\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0api_key=os.environ[\"OPENAI_API_KEY\"]\r\n\u00a0\u00a0\u00a0\u00a0)\r\n\u00a0\u00a0\u00a0\u00a0\r\n\u00a0\u00a0\u00a0\u00a0docs_embeddings\u00a0=\u00a0openai_ef.encode_documents(df['description'].tolist())\r\n\r\nreturn\u00a0docs_embeddings,\u00a0openai_ef<\/pre>\n<p>\u7136\u540e\uff0c\u628a\u751f\u6210\u7684Embedding\u5b58\u5165\u5230Milvus \u7684collection\u3002<\/p>\n<pre><code>def\u00a0save_embedding(docs_embeddings,\u00a0collection_name,\u00a0dim):\r\n\u00a0\u00a0\u00a0\u00a0data\u00a0=\u00a0[\r\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0{\"id\":\u00a0i,\u00a0\"vector\":\u00a0docs_embeddings[i].data,\u00a0\"text\":\u00a0row.language}\r\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0for\u00a0i,\u00a0row\u00a0in\u00a0df.iterrows()\r\n\u00a0\u00a0\u00a0\u00a0]\r\n\u00a0\u00a0\u00a0\u00a0\r\n\u00a0\u00a0\u00a0\u00a0if\u00a0milvus_client.has_collection(collection_name=collection_name):\r\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0milvus_client.drop_collection(collection_name=collection_name)\r\n\u00a0\u00a0\u00a0\u00a0milvus_client.create_collection(collection_name=collection_name,\u00a0dimension=dim)\r\n\u00a0\u00a0\u00a0\u00a0\r\n\u00a0\u00a0\u00a0\u00a0res\u00a0=\u00a0milvus_client.insert(collection_name=collection_name,\u00a0data=data)\r\n<\/code><\/pre>\n<p><strong>3.3 \u67e5\u8be2<\/strong><\/p>\n<p>\u6211\u4eec\u5b9a\u4e49\u67e5\u8be2\u51fd\u6570\uff0c\u65b9\u4fbf\u5bf9\u4e8e\u5411\u91cfEmbedding\u8fdb\u884c\u53ec\u56de\u3002<\/p>\n<pre>def\u00a0query_results(query,\u00a0collection_name,\u00a0openai_ef):\r\n\u00a0\u00a0\u00a0\u00a0query_embeddings\u00a0=\u00a0openai_ef.encode_queries(query)\r\n\u00a0\u00a0\u00a0\u00a0res\u00a0=\u00a0milvus_client.search(\r\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0collection_name=collection_name,\u00a0\r\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0data=query_embeddings,\u00a0\r\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0limit=4,\u00a0\u00a0\r\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0output_fields=[\"text\"],\r\n\u00a0\u00a0\u00a0\u00a0)\r\n\r\nresult\u00a0=\u00a0{}\r\n\u00a0\u00a0\u00a0\u00a0for\u00a0items\u00a0in\u00a0res:\r\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0for\u00a0item\u00a0in\u00a0items:\r\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0result[item.get(\"entity\").get(\"text\")]\u00a0=\u00a0item.get('distance')\r\n\r\nreturn\u00a0result<\/pre>\n<p><strong>3.4 \u8bc4\u4f30Embedding\u6a21\u578b\u6027\u80fd<\/strong><\/p>\n<p>\u6211\u4eec\u91c7\u7528 OpenAI\u7684\u4e24\u4e2a Embedding\u6a21\u578b\uff0c<code>text-embedding-3-small<\/code> \u548c <code>text-embedding-3-large<\/code>\uff0c\u5bf9\u4e8e\u5982\u4e0b\u4e24\u4e2a\u67e5\u8be2\u8fdb\u884c\u6bd4\u8f83\u3002\u6709\u5f88\u591a\u8bc4\u4f30\u6307\u6807\uff0c\u4f8b\u5982\u51c6\u786e\u7387\u3001\u53ec\u56de\u7387\u3001MRR\u3001MAP\u7b49\u3002\u5728\u8fd9\u91cc\uff0c\u6211\u4eec\u91c7\u7528\u51c6\u786e\u7387\u548c\u53ec\u56de\u7387\u3002<\/p>\n<p>\u51c6\u786e\u7387\uff08Precision\uff09 \u8bc4\u4f30\u68c0\u7d22\u7ed3\u679c\u4e2d\u7684\u771f\u6b63\u76f8\u5173\u5185\u5bb9\u7684\u5360\u6bd4\uff0c\u5373\u8fd4\u56de\u7684\u7ed3\u679c\u4e2d\u6709\u591a\u5c11\u4e0e\u641c\u7d22\u67e5\u8be2\u76f8\u5173\u3002<\/p>\n<p>Precision = TP \/ (TP + FP)<\/p>\n<p>\u5176\u4e2d\uff0c\u68c0\u7d22\u7ed3\u679c\u4e2d\u4e0e\u67e5\u8be2\u771f\u6b63\u76f8\u5173\u7684\u5185\u5bb9 True Positives\uff08TP\uff09, \u800c False Positives\uff08FP\uff09 \u6307\u7684\u662f\u68c0\u7d22\u7ed3\u679c\u4e2d\u4e0d\u76f8\u5173\u7684\u5185\u5bb9\u3002<\/p>\n<p>\u53ec\u56de\u7387 \uff08Recall\uff09\u8bc4\u4f30\u4ece\u6574\u4e2a\u6570\u636e\u96c6\u4e2d\u6210\u529f\u68c0\u7d22\u5230\u76f8\u5173\u5185\u5bb9\u7684\u6570\u91cf\u3002<\/p>\n<p>Recall = TP \/ (TP + FN)<\/p>\n<p>\u5176\u4e2d\uff0cFalse Negatives \uff08FN\uff09 \u6307\u7684\u662f\u6240\u6709\u672a\u5305\u542b\u5728\u6700\u7ec8\u7ed3\u679c\u96c6\u4e2d\u7684\u76f8\u5173\u9879\u76ee<\/p>\n<p><a href=\"https:\/\/zilliz.com\/learn\/information-retrieval-metrics\">\u5bf9\u4e8e\u8fd9\u4e24\u4e2a\u6982\u5ff5\u66f4\u8be6\u7ec6\u7684\u89e3\u91ca<\/a><\/p>\n<p><strong>\u67e5\u8be2 1<\/strong>\uff1a<code>auto garbage collection<\/code><\/p>\n<p>\u76f8\u5173\u9879\uff1aJava\uff0c Python\uff0c JavaScript\uff0c Golang<\/p>\n<table>\n<thead>\n<tr>\n<th>Rank<\/th>\n<th>text-embedding-3-small<\/th>\n<th>text-embedding-3-large<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>1<\/td>\n<td>\u274e Rust<\/td>\n<td>\u274e Rust<\/td>\n<\/tr>\n<tr>\n<td>2<\/td>\n<td>\u274e C\/C++<\/td>\n<td>\u274e C\/C++<\/td>\n<\/tr>\n<tr>\n<td>3<\/td>\n<td>\u2705 Golang<\/td>\n<td>\u2705 Java<\/td>\n<\/tr>\n<tr>\n<td>4<\/td>\n<td>\u2705 Java<\/td>\n<td>\u2705 Golang<\/td>\n<\/tr>\n<tr>\n<td>Precision<\/td>\n<td>0.50<\/td>\n<td>0.50<\/td>\n<\/tr>\n<tr>\n<td>Recall<\/td>\n<td>0.50<\/td>\n<td>0.50<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p><strong>\u67e5\u8be2 2<\/strong>\uff1a<code>suite for web backend server development<\/code><\/p>\n<p>\u76f8\u5173\u9879\uff1aJava, JavaScript, PHP, Python (\u7b54\u6848\u5305\u542b\u4e3b\u89c2\u5224\u65ad)<\/p>\n<table>\n<thead>\n<tr>\n<th>Rank<\/th>\n<th>text-embedding-3-small<\/th>\n<th>text-embedding-3-large<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>1<\/td>\n<td>\u2705 PHP<\/td>\n<td>\u2705 JavaScript<\/td>\n<\/tr>\n<tr>\n<td>2<\/td>\n<td>\u2705 Java<\/td>\n<td>\u2705 Java<\/td>\n<\/tr>\n<tr>\n<td>3<\/td>\n<td>\u2705 JavaScript<\/td>\n<td>\u2705 PHP<\/td>\n<\/tr>\n<tr>\n<td>4<\/td>\n<td>\u274e C#<\/td>\n<td>\u2705Python<\/td>\n<\/tr>\n<tr>\n<td>Precision<\/td>\n<td>0.75<\/td>\n<td>1.0<\/td>\n<\/tr>\n<tr>\n<td>Recall<\/td>\n<td>0.75<\/td>\n<td>1.0<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u5728\u8fd9\u4e24\u4e2a\u67e5\u8be2\u4e2d\uff0c\u6211\u4eec\u901a\u8fc7\u51c6\u786e\u7387\u548c\u53ec\u56de\u7387\u5bf9\u6bd4\u4e86\u4e24\u4e2aEmbedding\u6a21\u578b <code>text-embedding-3-small<\/code> \u548c <code>text-embedding-3-large<\/code> \u3002\u6211\u4eec\u53ef\u4ee5\u4ee5\u6b64\u4e3a\u8d77\u70b9\uff0c\u589e\u52a0\u6570\u636e\u96c6\u4e2d\u6570\u636e\u5bf9\u8c61\u7684\u6570\u91cf\u4ee5\u53ca\u67e5\u8be2\u7684\u6570\u91cf\uff0c\u5982\u6b64\u624d\u80fd\u66f4\u6709\u6548\u5730\u8bc4\u4f30Embedding\u6a21\u578b\u3002<\/p>\n<p>&nbsp;<\/p>\n<h2>04.\u603b\u7ed3<\/h2>\n<p>\u5728\u68c0\u7d22\u589e\u5f3a\u751f\u6210\uff08RAG\uff09\u5e94\u7528\u4e2d\uff0c\u9009\u62e9\u5408\u9002\u7684\u5411\u91cfEmbedding\u6a21\u578b\u81f3\u5173\u91cd\u8981\u3002\u672c\u6587\u9610\u8ff0\u4e86\u4ece\u5b9e\u9645\u4e1a\u52a1\u9700\u6c42\u51fa\u53d1\uff0c\u4eceMTEB\u9009\u62e9\u901a\u7528\u6a21\u578b\u4e4b\u540e\uff0c\u91c7\u7528\u51c6\u786e\u7387\u548c\u53ec\u56de\u7387\u5bf9\u4e8e\u6a21\u578b\u57fa\u4e8e\u7279\u5b9a\u4e1a\u52a1\u7684\u6570\u636e\u96c6\u8fdb\u884c\u6d4b\u8bd5\uff0c\u4ece\u800c\u9009\u62e9\u6700\u5408\u9002\u7684Embedding\u6a21\u578b\uff0c\u8fdb\u800c\u6709\u6548\u5730\u63d0\u9ad8RAG\u5e94\u7528\u7684\u53ec\u56de\u51c6\u786e\u7387\u3002<\/p>\n<p>\u5b8c\u6574\u4ee3\u7801\u53ef\u901a\u8fc7\u4e0b\u8f7d\u83b7\u53d6<\/p>\n\n<div id=\"J_MPDLCont\" class=\"magicpost-cont-wp\">\n  <div class=\"magicpost-cont-inner\">\n    <div class=\"magicpost-cont-hd\">\n      <svg class=\"wb-icon-magicpost wbsico-magicpost-local\">\n        <use xlink:href=\"#wbsico-magicpost-local\"><\/use>\n      <\/svg> <span>\u76f8\u5173\u6587\u4ef6\u4e0b\u8f7d\u5730\u5740<\/span>\n    <\/div>\n\n    <div class=\"magicpost-cont-bd\">\n                <div class=\"wb-tips\">\u8be5\u8d44\u6e90\u9700\u767b\u5f55\u540e\u4e0b\u8f7d\uff0c\u53bb<a class=\"link\" href=\"https:\/\/www.kdjingpai.com\/wp-login.php?redirect_to=https%3A%2F%2Fwww.kdjingpai.com%2Fde%2Fruheweiragyingyongben%2F\">\u767b\u5f55<\/a><\/div>\n\n            <\/div>\n\n    <div class=\"magicpost-cont-ft\">\n      &copy;\u4e0b\u8f7d\u8d44\u6e90\u7248\u6743\u5f52\u4f5c\u8005\u6240\u6709\uff1b\u672c\u7ad9\u6240\u6709\u8d44\u6e90\u5747\u6765\u6e90\u4e8e\u7f51\u7edc\uff0c\u4ec5\u4f9b\u5b66\u4e60\u4f7f\u7528\uff0c\u8bf7\u652f\u6301\u6b63\u7248\uff01    <\/div>\n  <\/div>\n<\/div>","protected":false},"excerpt":{"rendered":"<p>\u5411\u91cfEmbedding\u662f\u76ee\u524d\u68c0\u7d22\u589e\u5f3a\u751f\u6210\uff08RAG\uff09\u5e94\u7528\u7a0b\u5e8f\u7684\u6838\u5fc3\u3002\u5b83\u4eec\u6355\u83b7\u6570\u636e\u5bf9\u8c61\uff08\u5982\u6587\u672c\uff0c\u56fe\u50cf\u7b49\uff09\u7684\u8bed\u4e49\u4fe1\u606f\uff0c\u5e76\u4ee5\u6570\u5b57\u6570\u7ec4\u8868\u793a\u3002\u5728\u65f6\u4e0b\u7684\u751f\u6210\u5f0fAI\u5e94\u7528\u4e2d\uff0c\u8fd9\u4e9b\u5411\u91cfEmbedding\u901a\u5e38\u7531Embedding\u6a21\u578b\u751f\u6210\u3002\u5982\u4f55\u4e3aRAG\u5e94\u7528\u7a0b\u5e8f\u9009\u62e9&#8230;<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[34],"tags":[],"class_list":["post-16881","post","type-post","status-publish","format-standard","hentry","category-knowledge"],"_links":{"self":[{"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/posts\/16881","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/comments?post=16881"}],"version-history":[{"count":0,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/posts\/16881\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/media?parent=16881"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/categories?post=16881"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/tags?post=16881"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}