content/chapter2/ChatGPT使用指南——相似匹配.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "62bb6ea1",
   "metadata": {
    "toc": true
   },
   "source": [
    "<h1>Table of Contents<span class=\"tocSkip\"></span></h1>\n",
    "<div class=\"toc\"><ul class=\"toc-item\"><li><span><a href=\"#ChatGPT使用指南——相似匹配\" data-toc-modified-id=\"ChatGPT使用指南——相似匹配-1\">ChatGPT使用指南——相似匹配</a></span><ul class=\"toc-item\"><li><span><a href=\"#1.-何为Embedding\" data-toc-modified-id=\"1.-何为Embedding-1.1\">1. 何为Embedding</a></span></li><li><span><a href=\"#2.-相关API\" data-toc-modified-id=\"2.-相关API-1.2\">2. 相关API</a></span><ul class=\"toc-item\"><li><span><a href=\"#2.1-LMAS-Embedding-API\" data-toc-modified-id=\"2.1-LMAS-Embedding-API-1.2.1\">2.1 LMAS Embedding API</a></span></li><li><span><a href=\"#2.2-ChatGPT-Style\" data-toc-modified-id=\"2.2-ChatGPT-Style-1.2.2\">2.2 ChatGPT Style</a></span></li></ul></li><li><span><a href=\"#3.-Embedding应用\" data-toc-modified-id=\"3.-Embedding应用-1.3\">3. Embedding应用</a></span><ul class=\"toc-item\"><li><span><a href=\"#3.1-QA\" data-toc-modified-id=\"3.1-QA-1.3.1\">3.1 QA</a></span></li><li><span><a href=\"#3.2-聚类\" data-toc-modified-id=\"3.2-聚类-1.3.2\">3.2 聚类</a></span></li><li><span><a href=\"#3.3-推荐\" data-toc-modified-id=\"3.3-推荐-1.3.3\">3.3 推荐</a></span></li></ul></li><li><span><a href=\"#相关文献\" data-toc-modified-id=\"相关文献-1.4\">相关文献</a></span></li></ul></li></ul></div>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a3df9188",
   "metadata": {},
   "source": [
    "# ChatGPT使用指南——相似匹配"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "435621e7",
   "metadata": {},
   "source": [
    "## 1. 何为Embedding"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "649f4059",
   "metadata": {},
   "source": [
    "&emsp;&emsp;对于自然语言，因为它的输入是一段文本，在中文里就是一个一个字，或一个一个词，行业内把这个字或词叫Token。如果要使用模型，拿到一段文本的第一件事就是把它Token化，当然，可以按字、也可以按词，或按你想要的其他方式，比如每两个字一组（Bi-Gram）。举个例子：\n",
    "\n",
    "- 给定文本：我们相信AI可以让世界变得更美好。\n",
    "- 按字Token化：我/们/相/信/A/I/可/以/让/世/界/变/得/更/美/好/。\n",
    "- 按词Token化：我们/相信/AI/可以/让/世界/变得/更/美好/。\n",
    "- 按Bi-Gram Token化：我们/们相/相信/信A/AI/I可/可以/以让/让世/世界/界变/变得/得更/更美/美好/好。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "05c1990f",
   "metadata": {},
   "source": [
    "&emsp;&emsp;那自然就有一个新的问题：我们应该怎么选择Token化方式？其实每种不同的方法都有自己的优点和不足，在大模型之前，按词的方式比较常见。但在有了大模型之后，基本都是按字来了，不用再纠结这个点了。\n",
    "\n",
    "&emsp;&emsp;Token化后，第二件事就是要怎么表示这些Token，我们知道计算机只能处理数字，所以要想办法把这些Token给变成计算机「认识」的数字才行。读者不妨思考一下如果要你来做这件事会怎么做。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "20de24d0",
   "metadata": {},
   "source": [
    "&emsp;&emsp;其实很简单很直观，把所有字作为一个字典，序号就代表它自己。我们还是以上面的句子为例，假设词表就包含上面那些字，那么词表就可以用一个txt文件存储，内容如下：\n",
    "\n",
    "```text\n",
    "我\n",
    "们\n",
    "相\n",
    "信\n",
    "A\n",
    "I\n",
    "可\n",
    "以  \n",
    "让  \n",
    "世  \n",
    "界  \n",
    "变  \n",
    "得  \n",
    "更  \n",
    "美  \n",
    "好  \n",
    "。 \n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "72ce2918",
   "metadata": {},
   "source": [
    "&emsp;&emsp;一行一个字，每个字作为一个Token，此时，0=我，1=们，……，以此类推。拿中文来说，这个词表可能只要几千行，即使包含各种特殊符号、生僻字，也就2万个多点，我们假设词表大小为N。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5b7a47d4",
   "metadata": {},
   "source": [
    "&emsp;&emsp;接下来我们考虑如何用这些数字来表示一段文本。最简单的方法就是用它的ID直接串起来，这样也不是不行，但这种表示方法的特征是一维的，也就是说只能表示一个特征。这种方法不太符合实际情况，效果也不理想。所以，研究人员就想到另一种表示方法：One-Hot编码。其实，将文本变为数字表示的过程本质上就是一种编码过程。One-Hot的意思是，对每一个字都有N（词表大小）个特征，除了该字的ID位置值为1，其余都为0。我们依然用上面的例子来说明，把整个词表表示为下面的形式：\n",
    "\n",
    "\n",
    "> 我 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0  \n",
    "们 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0  \n",
    "相 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0  \n",
    "信 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0  \n",
    "……下面省略"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3c8a0a2b",
   "metadata": {},
   "source": [
    "&emsp;&emsp;此时，对每一个Token（字），它的表示就变成了一个一维向量，比如「我」：`[1,0,...0]`，这个向量的长度就是词表的大小N，它被称为「我」这个字的One-Hot表示。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "37c8f5cd",
   "metadata": {},
   "source": [
    "&emsp;&emsp;对于一段文本，我们一般会将每个Token的表示结合起来，结合方式可以采用求和或平均。这样，对于任意长度的任意文本，我们都能将其表示为固定大小的向量，非常方便进行各种矩阵或张量（三维以上的数组）计算，这对深度学习至关重要。\n",
    "\n",
    "&emsp;&emsp;举个例子，比如有这么一句话：让世界更美好。现在我们使用刚刚的方法将其表示为一个向量，采用平均的方式。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fe6407b8",
   "metadata": {},
   "source": [
    "&emsp;&emsp;首先，列出每个字的向量：\n",
    "\n",
    "> 让 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0  \n",
    "世 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0  \n",
    "界 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0  \n",
    "更 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0  \n",
    "美 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0  \n",
    "好 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0  \n",
    "。 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1  "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e200e77f",
   "metadata": {},
   "source": [
    "&emsp;&emsp;然后针对每一列取平均，结果为：`0 0 0 0 0 0 0 1/7 1/7 1/7 0 0 1/7 1/7 1/7 1/7`。不难发现，对任两句话，只要其中包含的字不完全一样，最终得到的向量表示也不会完全一样。注①"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "de8e1b17",
   "metadata": {},
   "source": [
    "&emsp;&emsp;当然，在实际使用时，往往不会这么简单的用1/0来表示，因为每个字在句子中的作用是不一样的，所以一般会给不同的Token赋予不同的权重。最常见的方法是使用在句子中出现的频率，那些高频的（但不是「的」「更」这样的虚词）被认为是重要的。更多可以参考【相关文献1】。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4ccfbff3",
   "metadata": {},
   "source": [
    "&emsp;&emsp;这种方法不错，在深度学习之前很长一段时间里都是这样的，不过它有两个很大的问题：\n",
    "\n",
    "1. 数据维度太高：太高的维度会导致向量在空间中聚集在一个非常狭窄的角落，模型难以训练。\n",
    "2. 数据稀疏，向量之间缺乏语义上的交互（语义鸿沟）：比如「我爱吃苹果」和「我爱用苹果」，前者是水果，后者是手机，怎么判断出来的呢？根据上下文。但由于现在这种表示方式，导致上下文之间是孤立的，所以模型学不到这个知识点。还有类似「我喜欢你」和「你喜欢我」这样会得到同样的表示，但其实是不同的意思。注①\n",
    "\n",
    "\n",
    "更多可以参考【相关文献 1】。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ba6ad3bd",
   "metadata": {},
   "source": [
    "&emsp;&emsp;终于轮到我们的主角Embedding登场了，它的主要思想是这样的：\n",
    "\n",
    "- 把特征固定在某一个维度D，比如256、300、768等等，这个不重要，总之不再是词表那么大的数字。这就避免了维度过高的问题。\n",
    "- 利用自然语言文本的上下文关系学习一个稠密表示。也就是说，每个Token的表示不再是预先算好的了，而是在过程中学习到的，元素也不再是很多个0，而是每个位置都有一个小数，这D个小数构成了一个Token表示。至于D个特征到底是什么，不知道，也不重要。我们只需要知道这D个小数就表示这个Token。\n",
    "\n",
    "&emsp;&emsp;还是继续以前面的例子来说明，这时候词表的表示变成下面这样了：\n",
    "\n",
    "\n",
    "> 我 0.xxx0, 0.yyy0, 0.zzz0, ... D个小数  \n",
    "们 0.xxx1, 0.yyy1, 0.zzz1, ... D个小数  \n",
    "相 0.xxx2, 0.yyy2, 0.zzz2, ... D个小数  \n",
    "信 0.xxx3, 0.yyy3, 0.zzz3, ... D个小数  \n",
    "……下面省略  \n",
    "\n",
    "&emsp;&emsp;这些小数怎么来的？简单，随机来的。就像下面这样："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "f1385d52",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "76dd94f7",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(16, 256)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rng = np.random.default_rng(42)\n",
    "# 词表大小N=16，维度D=256\n",
    "table = rng.uniform(size=(16, 256))\n",
    "table.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "7477dad3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.77395605, 0.43887844, 0.85859792, ..., 0.24783956, 0.23666236,\n",
       "        0.74601428],\n",
       "       [0.81656876, 0.10527808, 0.06655886, ..., 0.11585672, 0.07205915,\n",
       "        0.84199321],\n",
       "       [0.05556792, 0.28061144, 0.33413004, ..., 0.00925978, 0.18832197,\n",
       "        0.03128351],\n",
       "       ...,\n",
       "       [0.50647331, 0.22303613, 0.94414565, ..., 0.79202324, 0.40169878,\n",
       "        0.72247782],\n",
       "       [0.9151384 , 0.80071297, 0.39044651, ..., 0.03994193, 0.79502741,\n",
       "        0.28297954],\n",
       "       [0.68255979, 0.64272531, 0.65262805, ..., 0.18645529, 0.21927175,\n",
       "        0.32320729]])"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "table"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "854a4065",
   "metadata": {},
   "source": [
    "&emsp;&emsp;在模型训练过程中，会根据不同的上下文不断地更新这个参数，最后模型训练完后得到的这个矩阵就是Token的表示。我们完全可以把它当成一个黑盒子，输入一个X，根据标签Y不断更新参数，最终就得到一组参数，这些参数的名字就叫「模型」。\n",
    "\n",
    "&emsp;&emsp;这种表示方法在深度学习早期（2013-2015年左右）比较流行，不过由于这个矩阵训练好后就固定不变了，这在有些时候就不合适。比如「你好坏」这句话在不同的情况下可能完全是不同的意思。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9f7588b2",
   "metadata": {},
   "source": [
    "&emsp;&emsp;我们知道，句子才是语义的最小单位，因此相比Token，我们其实更加关注和需要句子的表示，我们期望可以根据不同上下文动态地获得句子表示。这中间当然经历了比较多的探索，一直到如今的大模型时代，对模型输入任意一句话，它都能给我们返回一个非常不错的表示，而且依然是固定长度的向量。\n",
    "\n",
    "&emsp;&emsp;如果对这方面感兴趣，可以进一步阅读【相关文献2】和【相关文献3】。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "81092f69",
   "metadata": {},
   "source": [
    "&emsp;&emsp;我们总结一下，Embedding本质就是一组稠密向量，用来表示一段文本（可以是字、词、句、段等），获取到这个表示后，我们就可以进一步做一些任务。大家不妨先思考一下，当给定任意句子并获得到它的固定长度的语义表示时，我们可以干什么？我们在下一节将先介绍一下OpenAI提供的接口，以及一些后面任务可能用到的概念。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "02b8719e",
   "metadata": {},
   "source": [
    "## 2. 相关API"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "13cb04d8",
   "metadata": {},
   "source": [
    "### 2.1 LMAS Embedding API"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "aeb026c8",
   "metadata": {},
   "outputs": [],
   "source": [
    "from openai import OpenAI\n",
    "OPENAI_API_KEY = \"填入专属的API key\"\n",
    "client = OpenAI(api_key=OPENAI_API_KEY)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "01354566",
   "metadata": {},
   "outputs": [],
   "source": [
    "text = \"我喜欢你\"\n",
    "model = \"text-embedding-ada-002\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "0db3dfc7",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "emb_req = client.embeddings.create(input=[text], model=model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "ff56be5e",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1536, list)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "emb = emb_req.data[0].embedding\n",
    "len(emb), type(emb)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f3c49506",
   "metadata": {},
   "source": [
    "以下是智谱的API，用法不能说类似，只能说一样。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9f4835a2",
   "metadata": {},
   "outputs": [],
   "source": [
    "from zhipuai import ZhipuAI\n",
    " \n",
    "zhipu_client = ZhipuAI(api_key=\"your api key\") "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "id": "9ab985f0",
   "metadata": {},
   "outputs": [],
   "source": [
    "emb_req = zhipu_client.embeddings.create(\n",
    "    model=\"embedding-2\", #填写需要调用的模型名称\n",
    "    input=text,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "id": "8081e175",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1024, list)"
      ]
     },
     "execution_count": 155,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "emb = emb_req.data[0].embedding\n",
    "len(emb), type(emb)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0544dc2e",
   "metadata": {},
   "source": [
    "&emsp;&emsp;与Embedding息息相关的一个概念是「相似度」，准确来说是「语义相似度」。在自然语言处理领域，我们一般使用**cosine相似度**作为语义相似度的度量，评估两个向量在语义空间上的分布情况。\n",
    "\n",
    "&emsp;&emsp;具体来说就是下面这个式子：\n",
    "\n",
    "$$\n",
    "\\text{cosine}(v,w) = \\frac {v·w}{|v||w|} = \\frac {\\displaystyle \\sum_{i=1}^N v_iw_i} {\\displaystyle \\sqrt{\\sum_{i=1}^N v_i^2} \\sqrt{\\sum_{i=1}^N w_i^2}}\n",
    "$$"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "026f569b",
   "metadata": {},
   "source": [
    "&emsp;&emsp;我们举个例子："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "e96e1087",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9925833339709301"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import numpy as np\n",
    "a = [0.1, 0.2, 0.3]\n",
    "b = [0.2, 0.3, 0.4]\n",
    "cosine_ab = (0.1*0.2+0.2*0.3+0.3*0.4)/(np.sqrt(0.1**2+0.2**2+0.3**2) * np.sqrt(0.2**2+0.3**2+0.4**2))\n",
    "cosine_ab"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f7fc7f26",
   "metadata": {},
   "source": [
    "&emsp;&emsp;OpenAI官方提供了一个集成接口，使用起来更加简单（但其实你也可以自己写一个）："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 165,
   "id": "60dc3fdd",
   "metadata": {},
   "outputs": [],
   "source": [
    "# 计算两个向量的余弦相似度\n",
    "def cosine_similarity(vec1, vec2):\n",
    "    return np.dot(vec1, vec2) / (np.linalg.norm(vec1) * np.linalg.norm(vec2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "id": "fed6baf7",
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_embedding(text,model = \"text-embedding-ada-002\"):\n",
    "    emb_req = client.embeddings.create(model=model,input = text)\n",
    "    return emb_req.data[0].embedding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "id": "937071de",
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_embedding_zhipu(text):\n",
    "    emb_req = zhipu_client.embeddings.create(\n",
    "        model=\"embedding-2\",\n",
    "        input=text,\n",
    "    )\n",
    "    return emb_req.data[0].embedding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "id": "bbfa5de2",
   "metadata": {},
   "outputs": [],
   "source": [
    "# 注意它支持多种模型，可以通过接口查看\n",
    "text1 = \"我喜欢你\"\n",
    "text2 = \"我钟意你\"\n",
    "text3 = \"我不喜欢你\"\n",
    "emb1 = get_embedding(text1, \"text-embedding-3-large\")\n",
    "emb2 = get_embedding(text2, \"text-embedding-3-large\")\n",
    "emb3 = get_embedding(text3, \"text-embedding-3-large\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "id": "04b9e465",
   "metadata": {},
   "outputs": [],
   "source": [
    "emb1_zhipu = get_embedding_zhipu(text1)\n",
    "emb2_zhipu = get_embedding_zhipu(text2)\n",
    "emb3_zhipu = get_embedding_zhipu(text3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 164,
   "id": "c545bfb2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(3072, list, 1024, list)"
      ]
     },
     "execution_count": 164,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(emb1), type(emb1), len(emb1_zhipu), type(emb1_zhipu)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "id": "b96b516a",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.6413979806416552, 0.7285185590498366)"
      ]
     },
     "execution_count": 167,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cosine_similarity(emb1, emb2), cosine_similarity(emb1_zhipu, emb2_zhipu)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "id": "8598a5c6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.6305954523234731, 0.8184623856653748)"
      ]
     },
     "execution_count": 168,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cosine_similarity(emb1, emb3), cosine_similarity(emb1_zhipu, emb3_zhipu)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 170,
   "id": "da81c2fc",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.4753375880949686, 0.6273458560395342)"
      ]
     },
     "execution_count": 170,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cosine_similarity(emb2, emb3), cosine_similarity(emb2_zhipu, emb3_zhipu)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2b6940d5",
   "metadata": {},
   "source": [
    "OpenAI还有其他Embedding模型，比如："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 171,
   "id": "c0518791",
   "metadata": {},
   "outputs": [],
   "source": [
    "text1 = \"我喜欢你\"\n",
    "text2 = \"我钟意你\"\n",
    "text3 = \"我不喜欢你\"\n",
    "emb1 = get_embedding(text1, \"text-embedding-ada-002\")\n",
    "emb2 = get_embedding(text2, \"text-embedding-ada-002\")\n",
    "emb3 = get_embedding(text3, \"text-embedding-ada-002\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 175,
   "id": "63dfb1ba",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8939290846547182"
      ]
     },
     "execution_count": 175,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cosine_similarity(emb1, emb2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 176,
   "id": "59f7f701",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9260708403768584"
      ]
     },
     "execution_count": 176,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cosine_similarity(emb1, emb3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 177,
   "id": "de5f3155",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8457468167009362"
      ]
     },
     "execution_count": 177,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cosine_similarity(emb2, emb3)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3c8ac608",
   "metadata": {},
   "source": [
    "更多模型可以在这里查看：[New and improved embedding model](https://openai.com/blog/new-and-improved-embedding-model)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b32ebee1",
   "metadata": {},
   "source": [
    "### 2.2 ChatGPT Style"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c8391ff2",
   "metadata": {},
   "source": [
    "&emsp;&emsp;接下来我们用万能的ChatGPT尝试一下，注意它不会给你返回Embedding，它是尝试直接告诉你答案！"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 181,
   "id": "d958fdca",
   "metadata": {},
   "outputs": [],
   "source": [
    "content = \"请告诉我下面三句话的相似程度：\\n1. 我喜欢你。\\n2. 我钟意你。\\n3.我不喜欢你。\\n\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 182,
   "id": "0c2c7a6a",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1. 我喜欢你。\n",
      "2. 我钟意你。\n",
      "这两句话的意思非常相似，表达了对对方的好感和迷恋。\n",
      "\n",
      "3. 我不喜欢你。\n",
      "这句话与前两句相比是相反的意思，表达了对对方的不喜欢或者厌恶之情。与前两句相比，相似程度较低。\n"
     ]
    }
   ],
   "source": [
    "response = client.chat.completions.create(\n",
    "    model=\"gpt-3.5-turbo\", \n",
    "    messages=[{\"role\": \"user\", \"content\": content}]\n",
    ")\n",
    "\n",
    "print(response.choices[0].message.content)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "282a4202",
   "metadata": {},
   "source": [
    "智谱AI结果如下："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 184,
   "id": "0b2e4b77",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "这三句话的情感倾向是不同的。第一句“我喜欢你”和第二句“我钟意你”都是表达正面的喜欢或爱意，只不过用词稍有差异。第三句“我不喜欢你”则表达了否定的情感，即不喜欢。在情感表达上，第一句和第二句相似，与第三句相反。\n"
     ]
    }
   ],
   "source": [
    "response = zhipu_client.chat.completions.create(\n",
    "    model=\"glm-3-turbo\",\n",
    "    messages=[\n",
    "        {\"role\": \"user\", \"content\": content},\n",
    "    ],\n",
    ")\n",
    "print(response.choices[0].message.content)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "77d769fd",
   "metadata": {},
   "source": [
    "牛逼，不过这个格式不太好，我们调整一下："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 193,
   "id": "410ec980",
   "metadata": {},
   "outputs": [],
   "source": [
    "content = \"\"\"请告诉我下面三句话的相似程度：\n",
    "1. 我喜欢你。\n",
    "2. 我钟意你。\n",
    "3. 我不喜欢你。\n",
    "第一句话用a表示，第二句话用b表示，第三句话用c表示。\n",
    "请以json格式输出两两语义相似度。仅输出json，不要输出其他任何内容。\n",
    "\"\"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 194,
   "id": "07b2de75",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"a-b\": 0.9,\n",
      "  \"a-c\": 0.2,\n",
      "  \"b-c\": 0.1\n",
      "}\n"
     ]
    }
   ],
   "source": [
    "response = client.chat.completions.create(\n",
    "    model=\"gpt-3.5-turbo\", \n",
    "    messages=[{\"role\": \"user\", \"content\": content}],\n",
    ")\n",
    "\n",
    "print(response.choices[0].message.content)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0b8ae4d1",
   "metadata": {},
   "source": [
    "智谱AI结果如下："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 186,
   "id": "bd27b98d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "```json\n",
      "{\n",
      "  \"a-b\": 0.9,\n",
      "  \"a-c\": -1,\n",
      "  \"b-c\": -1\n",
      "}\n",
      "```\n"
     ]
    }
   ],
   "source": [
    "response = zhipu_client.chat.completions.create(\n",
    "    model=\"glm-3-turbo\",\n",
    "    messages=[\n",
    "        {\"role\": \"user\", \"content\": content},\n",
    "    ],\n",
    ")\n",
    "print(response.choices[0].message.content)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6b725bd9",
   "metadata": {},
   "source": [
    "牛逼+++！"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1c907870",
   "metadata": {},
   "source": [
    "## 3. Embedding应用"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "91da530f",
   "metadata": {},
   "source": [
    "&emsp;&emsp;有读者可能会疑惑，既然有了这么牛逼Plus的ChatGPT了，为什么还要介绍Embedding这种看起来好像是「低级货」的技术呢？这里主要有两个原因：\n",
    "\n",
    "1. 有些问题使用Embedding解决（或其他非ChatGPT的方式）会更加合理。通俗来说就是「杀鸡焉用牛刀」。\n",
    "2. ChatGPT性能方面不是特别友好，毕竟是一个Token一个Token吐出来的。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8178341b",
   "metadata": {},
   "source": [
    "&emsp;&emsp;关于第一点，我们要额外多说几句。选择技术方案就跟选对象一样，合适最重要。只要你的问题（需求）没变，能解决的技术就是好技术。比如你的任务就是一个二分类，明明一个很简单的模型就能解决，就没必要非得上个很复杂的。除非ChatGPT这样的LLM已经大众到一定阶段（任何人都能够非常流畅、自由地使用），我们从统一性角度考虑。\n",
    "\n",
    "&emsp;&emsp;言归正传，使用Embedding的应用大多跟语义相关，我们这里介绍与此相关的几个经典任务以及衍生的应用。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ecbb79ea",
   "metadata": {},
   "source": [
    "### 3.1 QA"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4533f70c",
   "metadata": {},
   "source": [
    "&emsp;&emsp;QA是问答的意思，Q表示Question，A表示Answer，QA是NLP非常基础和常用的任务。简单来说，就是当用户提出一个问题时，我们能从已有的问题库中找到一个最相似的，并把它的答案返回给用户。这里有两个关键点：\n",
    "\n",
    "1. 事先需要有一个QA库。\n",
    "2. 用户提问时，系统要能够在QA库中找到一个最相似的。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cab7c796",
   "metadata": {},
   "source": [
    "&emsp;&emsp;ChatGPT（或生成方式）做这类任务相对有点麻烦，尤其是当：\n",
    "\n",
    "- QA库非常庞大时\n",
    "- 给用户的答案是固定的，不允许自由发挥时\n",
    "\n",
    "&emsp;&emsp;生成方式做起来是事倍功半。但是Embedding确实天然的非常适合，因为该任务的核心就是在一堆文本中找出给定文本最相似的。简单来说，其实就是个相似度计算问题。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "af750917",
   "metadata": {},
   "source": [
    "&emsp;&emsp;我们使用Kaggle提供的Quora数据集：[FAQ Kaggle dataset! | Data Science and Machine Learning](https://www.kaggle.com/general/183270)，先把它给读进来。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 195,
   "id": "f9caddeb",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 196,
   "id": "5d52818e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1166, 4)"
      ]
     },
     "execution_count": 196,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv(\"dataset/Kaggle related questions on Qoura - Questions.csv\")\n",
    "df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 197,
   "id": "11c46874",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Questions</th>\n",
       "      <th>Followers</th>\n",
       "      <th>Answered</th>\n",
       "      <th>Link</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>How do I start participating in Kaggle competi...</td>\n",
       "      <td>1200</td>\n",
       "      <td>1</td>\n",
       "      <td>/How-do-I-start-participating-in-Kaggle-compet...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Is Kaggle dead?</td>\n",
       "      <td>181</td>\n",
       "      <td>1</td>\n",
       "      <td>/Is-Kaggle-dead</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>How should a beginner get started on Kaggle?</td>\n",
       "      <td>388</td>\n",
       "      <td>1</td>\n",
       "      <td>/How-should-a-beginner-get-started-on-Kaggle</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>What are some alternatives to Kaggle?</td>\n",
       "      <td>201</td>\n",
       "      <td>1</td>\n",
       "      <td>/What-are-some-alternatives-to-Kaggle</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>What Kaggle competitions should a beginner sta...</td>\n",
       "      <td>273</td>\n",
       "      <td>1</td>\n",
       "      <td>/What-Kaggle-competitions-should-a-beginner-st...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                           Questions  Followers  Answered  \\\n",
       "0  How do I start participating in Kaggle competi...       1200         1   \n",
       "1                                    Is Kaggle dead?        181         1   \n",
       "2       How should a beginner get started on Kaggle?        388         1   \n",
       "3              What are some alternatives to Kaggle?        201         1   \n",
       "4  What Kaggle competitions should a beginner sta...        273         1   \n",
       "\n",
       "                                                Link  \n",
       "0  /How-do-I-start-participating-in-Kaggle-compet...  \n",
       "1                                    /Is-Kaggle-dead  \n",
       "2       /How-should-a-beginner-get-started-on-Kaggle  \n",
       "3              /What-are-some-alternatives-to-Kaggle  \n",
       "4  /What-Kaggle-competitions-should-a-beginner-st...  "
      ]
     },
     "execution_count": 197,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "15ddf462",
   "metadata": {},
   "source": [
    "&emsp;&emsp;这里，我们就把Link当做答案构造数据对。基本的流程如下：\n",
    "\n",
    "- 对每个Question计算Embedding\n",
    "- 存储Embedding，同时存储每个Question对应的答案\n",
    "- 从存储的地方检索最相似的Question"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b0796354",
   "metadata": {},
   "source": [
    "&emsp;&emsp;第一步我们将借助OpenAI的Embedding接口，但是后两步得看实际情况了。如果Question的数量比较少，比如只有几万条甚至几千条，那我们可以把计算好的Embedding直接存储成文件，每次服务启动时直接加载到内存或缓存里就好了。使用时，挨个计算输入问题和存储的所有问题的相似度，然后给出最相似的问题的答案。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9cdc80cc",
   "metadata": {},
   "source": [
    "&emsp;&emsp;为了快速演示，我们只取前5个句子为例："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 198,
   "id": "a4751337",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "id": "6fecd512",
   "metadata": {},
   "outputs": [],
   "source": [
    "vec_base = []\n",
    "for v in df.head().itertuples():\n",
    "    emb = get_embedding(v.Questions)\n",
    "    im = {\n",
    "        \"question\": v.Questions,\n",
    "        \"embedding\": emb,\n",
    "        \"answer\": v.Link\n",
    "    }\n",
    "    vec_base.append(im)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 199,
   "id": "fcc82cd1",
   "metadata": {},
   "outputs": [],
   "source": [
    "vec_base_zhipu = []\n",
    "for v in df.head().itertuples():\n",
    "    emb = get_embedding_zhipu(v.Questions)\n",
    "    im = {\n",
    "        \"question\": v.Questions,\n",
    "        \"embedding\": emb,\n",
    "        \"answer\": v.Link\n",
    "    }\n",
    "    vec_base_zhipu.append(im)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9e0fa92a",
   "metadata": {},
   "source": [
    "&emsp;&emsp;然后给定输入，比如：\"is kaggle alive?\"，我们先获取它的Embedding，然后逐个遍历`vec_base`计算相似度，并取最高的作为响应。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 205,
   "id": "d90db14b",
   "metadata": {},
   "outputs": [],
   "source": [
    "query = \"is kaggle alive?\"\n",
    "q_emb = get_embedding(query)\n",
    "q_emb_zhpu = get_embedding_zhipu(query)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 206,
   "id": "e93aacc9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.7879602774711745,\n",
       " 0.9542302479388203,\n",
       " 0.8227517970462344,\n",
       " 0.8398752285928424,\n",
       " 0.8109667304826295]"
      ]
     },
     "execution_count": 206,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sims = [cosine_similarity(q_emb, v[\"embedding\"]) for v in vec_base]\n",
    "sims"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 207,
   "id": "73c62d9e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.47816811389316444,\n",
       " 0.8856716112337668,\n",
       " 0.6596088197402513,\n",
       " 0.6973234108766252,\n",
       " 0.5534577913085058]"
      ]
     },
     "execution_count": 207,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sims_zhipu = [cosine_similarity(q_emb_zhpu, v[\"embedding\"]) for v in vec_base_zhipu]\n",
    "sims_zhipu"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "89de319c",
   "metadata": {},
   "source": [
    "&emsp;&emsp;我们返回第二个即可："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 210,
   "id": "315242e4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Is Kaggle dead? /Is-Kaggle-dead Is Kaggle dead? /Is-Kaggle-dead\n"
     ]
    }
   ],
   "source": [
    "print(\n",
    "    vec_base[1][\"question\"], vec_base[1][\"answer\"], \n",
    "    vec_base_zhipu[1][\"question\"], vec_base_zhipu[1][\"answer\"]\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f2822f29",
   "metadata": {},
   "source": [
    "&emsp;&emsp;当然，在实际中，我们不建议使用循环，大家可以使用NumPy进行批量计算（智谱AI一样，不赘述）。\n",
    "\n",
    ">调用Embedding API，也要查看对应文档看是否支持Batch方法。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "id": "08599520",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(5, 1536)"
      ]
     },
     "execution_count": 81,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "arr = np.array(\n",
    "    [v[\"embedding\"] for v in vec_base]\n",
    ")\n",
    "arr.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "id": "fcea762e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1, 1536)"
      ]
     },
     "execution_count": 82,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "q_arr = np.expand_dims(q_emb, 0)\n",
    "q_arr.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "id": "b05378e2",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics.pairwise import cosine_similarity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "id": "76d82885",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.78796028],\n",
       "       [0.95423025],\n",
       "       [0.8227518 ],\n",
       "       [0.83987523],\n",
       "       [0.81096673]])"
      ]
     },
     "execution_count": 84,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cosine_similarity(arr, q_arr)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "983a9f96",
   "metadata": {},
   "source": [
    "&emsp;&emsp;不过，当Question非常多，比如上百万甚至上亿时，这种方式就不合适了。一个是内存里可能放不下，另一个是算起来也很慢。这时候就必须借助一些专门用来做语义检索的工具了。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "00b719f0",
   "metadata": {},
   "source": [
    "&emsp;&emsp;比较常用的工具有：\n",
    "\n",
    "- [facebookresearch/faiss: A library for efficient similarity search and clustering of dense vectors.](https://github.com/facebookresearch/faiss)\n",
    "- [milvus-io/milvus: Vector database for scalable similarity search and AI applications.](https://github.com/milvus-io/milvus)\n",
    "- [Vector similarity | Redis](https://redis.io/docs/stack/search/reference/vectors/)\n",
    "\n",
    "此处，我们以Redis为例，其他工具用法类似。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ad9756f3",
   "metadata": {},
   "source": [
    "&emsp;&emsp;首先，我们需要一个redis，建议使用docker直接运行：\n",
    "\n",
    "```shell\n",
    "docker run -p 6379:6379 -it redis/redis-stack:latest\n",
    "```\n",
    "\n",
    "执行后，docker会自动从hub把镜像拉到本地，默认是6379端口。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f5799c26",
   "metadata": {},
   "source": [
    "&emsp;&emsp;然后安装redis-py，也就是Redis的Python客户端：\n",
    "\n",
    "```shell\n",
    "pip install redis\n",
    "```\n",
    "\n",
    "这样我们就可以用Python和Redis进行交互了。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "da297f4c",
   "metadata": {},
   "source": [
    "&emsp;&emsp;先来个最简单的例子："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "id": "86ed73d6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 88,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import redis\n",
    "r = redis.Redis()\n",
    "r.set(\"key\", \"value\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "2b4b7fb8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "b'value'"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "r.get(\"key\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "eea1fbf2",
   "metadata": {},
   "source": [
    "&emsp;&emsp;如果大家使用过ElasticSearch，接下来的内容会非常容易理解。总的来说，和刚刚的步骤差不多，但是这里我们需要先建索引，然后生成Embedding并把它存储到Redis，再进行使用（从索引中搜索）。不过由于我们使用了工具，具体步骤会略微不同。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8121b238",
   "metadata": {},
   "source": [
    "&emsp;&emsp;索引的概念和数据库中的索引有点相似，就是要定义一组Schema，告诉Redis你的字段是什么，有哪些属性。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "id": "7325e474",
   "metadata": {},
   "outputs": [],
   "source": [
    "VECTOR_DIM = 1536 # 智谱AI需改为1024，因为智谱AI的Embedding维度是1024\n",
    "INDEX_NAME = \"faq\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "id": "d5f62ee9",
   "metadata": {},
   "outputs": [],
   "source": [
    "from redis.commands.search.query import Query\n",
    "from redis.commands.search.field import TextField, VectorField\n",
    "from redis.commands.search.indexDefinition import IndexDefinition"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "id": "f6431c3a",
   "metadata": {},
   "outputs": [],
   "source": [
    "# 建好要存字段的索引，针对不同属性字段，使用不同Field\n",
    "question = TextField(name=\"question\")\n",
    "answer = TextField(name=\"answer\")\n",
    "embedding = VectorField(\n",
    "    name=\"embedding\", \n",
    "    algorithm=\"HNSW\", \n",
    "    attributes={\n",
    "        \"TYPE\": \"FLOAT32\",\n",
    "        \"DIM\": VECTOR_DIM,\n",
    "        \"DISTANCE_METRIC\": \"COSINE\"\n",
    "    }\n",
    ")\n",
    "schema = (question, embedding, answer)\n",
    "index = r.ft(INDEX_NAME)\n",
    "try:\n",
    "    info = index.info()\n",
    "except:\n",
    "    index.create_index(schema, definition=IndexDefinition(prefix=[INDEX_NAME + \"-\"]))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fafe0622",
   "metadata": {},
   "source": [
    ">Hierarchical Navigable Small Worlds，HNSW，一种高效的相似搜索算法。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "id": "3c3f9eb0",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "b'OK'"
      ]
     },
     "execution_count": 91,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 如果需要删除已有文档的话，可以使用下面的命令\n",
    "# index.dropindex(delete_documents=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d1e740a9",
   "metadata": {},
   "source": [
    "&emsp;&emsp;接下来就是把数据存到Redis。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "id": "c9db8707",
   "metadata": {},
   "outputs": [],
   "source": [
    "for v in df.head().itertuples():\n",
    "    # 智谱AI\n",
    "    # emb = get_embedding_zhipu(v.Questions)\n",
    "    emb = get_embedding(v.Questions)\n",
    "    # 注意，redis要存储bytes或string\n",
    "    emb = np.array(emb, dtype=np.float32).tobytes()\n",
    "    im = {\n",
    "        \"question\": v.Questions,\n",
    "        \"embedding\": emb,\n",
    "        \"answer\": v.Link\n",
    "    }\n",
    "    # 重点是这句\n",
    "    r.hset(name=f\"{INDEX_NAME}-{v.Index}\", mapping=im)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7feaabcb",
   "metadata": {},
   "source": [
    "&emsp;&emsp;然后我们就可以进行搜索查询了，这一步构造查询输入稍微有一点麻烦。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "id": "667fede5",
   "metadata": {},
   "outputs": [],
   "source": [
    "# 构造查询输入\n",
    "query = \"kaggle alive?\"\n",
    "embed_query = get_embedding(query)\n",
    "# 智谱AI\n",
    "# embed_query = get_embedding_zhipu(query)\n",
    "params_dict = {\"query_embedding\": np.array(embed_query).astype(dtype=np.float32).tobytes()}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "id": "992cd90c",
   "metadata": {
    "code_folding": []
   },
   "outputs": [],
   "source": [
    "k = 3\n",
    "# {some filter query}=>[ KNN {num|$num} @vector_field $query_vec]\n",
    "base_query = f\"* => [KNN {k} @embedding $query_embedding AS score]\"\n",
    "return_fields = [\"question\", \"answer\", \"score\"]\n",
    "query = (\n",
    "    Query(base_query)\n",
    "     .return_fields(*return_fields)\n",
    "     .sort_by(\"score\")\n",
    "     .paging(0, k)\n",
    "     .dialect(2)\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4397a021",
   "metadata": {},
   "source": [
    "> KNN（K最近邻算法），简单来说就是对未知点，分别和已有的点算距离，挑距离最近的K个点。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "id": "978f7289",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "faq-1, Is Kaggle dead?, /Is-Kaggle-dead (Similarity: 0.929)\n",
      "faq-3, What are some alternatives to Kaggle?, /What-are-some-alternatives-to-Kaggle (Similarity: 0.844)\n",
      "faq-2, How should a beginner get started on Kaggle?, /How-should-a-beginner-get-started-on-Kaggle (Similarity: 0.833)\n"
     ]
    }
   ],
   "source": [
    "# 查询\n",
    "res = index.search(query, params_dict)\n",
    "for i,doc in enumerate(res.docs):\n",
    "    similarity = 1 - float(doc.score)\n",
    "    print(f\"{doc.id}, {doc.question}, {doc.answer} (Similarity: {round(similarity ,3) })\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c906354d",
   "metadata": {},
   "source": [
    "&emsp;&emsp;上面，我们通过几种不同的方法为大家介绍了如何使用Embedding进行QA任务。简单回顾一下，要做QA任务首先咱们得有一个QA库，这些QA就是我们的仓库，每当一个新的问题过来时，我们就用这个问题去和咱们仓库里的每一个Q去匹配，然后找到最相似的那个，接下来就把该问题的Answer当做新问题的Answer交给用户。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "76eadfc8",
   "metadata": {},
   "source": [
    "&emsp;&emsp;这个任务的核心就是如何找到这个最相似的，涉及两个知识点：如何表示一个Question，以及如何查找到相似的Question。对于第一点，我们用API提供的Embedding表示，我们可以把它当做一个黑盒子，输入任意长度的文本，输出一个向量。查找相似问题则主要是用到相似度算法，语义相似度一般用cosine距离来衡量。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "633912db",
   "metadata": {},
   "source": [
    "&emsp;&emsp;当然实际中可能会更加复杂一些，比如我们可能除了使用语义匹配，还会使用字词匹配（经典的做法）。而且，一般都会找到topN个相似的，然后对这topN个结果进行排序，选出最可能的那个。不过，这个前面我们举过例子了，完全可以通过ChatGPT来解决，让它帮你选出最好的那个。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0720ffa2",
   "metadata": {},
   "source": [
    "### 3.2 聚类"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bd3e4298",
   "metadata": {},
   "source": [
    "&emsp;&emsp;聚类的意思是把彼此相近的样本聚集在一起，本质也是在使用一种表示和相似度衡量来处理文本。比如我们有大量的未分类文本，如果事先能知道有几种类别，就可以用聚类的方法先将样本大致分一下。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dad8280e",
   "metadata": {},
   "source": [
    "&emsp;&emsp;我们使用Kaggle的DBPedia数据集：[DBPedia Classes | Kaggle](https://www.kaggle.com/datasets/danofer/dbpedia-classes?select=DBPEDIA_val.csv)。\n",
    "\n",
    "&emsp;&emsp;这个数据集对一段文本会给出三个不同层次级别的分类标签，我们这里用第一层的类别。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 211,
   "id": "385936c7",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 212,
   "id": "9df5883b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(36003, 4)"
      ]
     },
     "execution_count": 212,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv(\"./dataset/DBPEDIA_val.csv\")\n",
    "df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 213,
   "id": "3e2affca",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>text</th>\n",
       "      <th>l1</th>\n",
       "      <th>l2</th>\n",
       "      <th>l3</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Li Curt is a station on the Bernina Railway li...</td>\n",
       "      <td>Place</td>\n",
       "      <td>Station</td>\n",
       "      <td>RailwayStation</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Grafton State Hospital was a psychiatric hospi...</td>\n",
       "      <td>Place</td>\n",
       "      <td>Building</td>\n",
       "      <td>Hospital</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>The Democratic Patriotic Alliance of Kurdistan...</td>\n",
       "      <td>Agent</td>\n",
       "      <td>Organisation</td>\n",
       "      <td>PoliticalParty</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Ira Rakatansky (October 3, 1919 – March 4, 201...</td>\n",
       "      <td>Agent</td>\n",
       "      <td>Person</td>\n",
       "      <td>Architect</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Universitatea Reșița is a women handball club ...</td>\n",
       "      <td>Agent</td>\n",
       "      <td>SportsTeam</td>\n",
       "      <td>HandballTeam</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                text     l1            l2  \\\n",
       "0  Li Curt is a station on the Bernina Railway li...  Place       Station   \n",
       "1  Grafton State Hospital was a psychiatric hospi...  Place      Building   \n",
       "2  The Democratic Patriotic Alliance of Kurdistan...  Agent  Organisation   \n",
       "3  Ira Rakatansky (October 3, 1919 – March 4, 201...  Agent        Person   \n",
       "4  Universitatea Reșița is a women handball club ...  Agent    SportsTeam   \n",
       "\n",
       "               l3  \n",
       "0  RailwayStation  \n",
       "1        Hospital  \n",
       "2  PoliticalParty  \n",
       "3       Architect  \n",
       "4    HandballTeam  "
      ]
     },
     "execution_count": 213,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "96fd25f0",
   "metadata": {},
   "source": [
    "查看一下类别数量："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "id": "4601f84e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "l1\n",
       "Agent             18647\n",
       "Place              6855\n",
       "Species            3210\n",
       "Work               3141\n",
       "Event              2854\n",
       "SportsSeason        879\n",
       "UnitOfWork          263\n",
       "TopicalConcept      117\n",
       "Device               37\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 123,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.l1.value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5e7a6d9a",
   "metadata": {},
   "source": [
    "数量有点多，我们随机采样200条："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "id": "e3f248f3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "l1\n",
       "Agent             107\n",
       "Place              36\n",
       "Species            21\n",
       "Work               14\n",
       "Event              11\n",
       "SportsSeason        8\n",
       "TopicalConcept      2\n",
       "UnitOfWork          1\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 117,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sdf = df.sample(200)\n",
    "sdf.l1.value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8b439fbd",
   "metadata": {},
   "source": [
    "&emsp;&emsp;为了便于观察，我们只保留3个数量差不多的类别：Place、Work和Species。（类型太多时，样本点会混在一块难以观察，大家可以自己尝试。）"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "id": "dcc48fa6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(71, 4)"
      ]
     },
     "execution_count": 124,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cdf = sdf[\n",
    "    (sdf.l1 == \"Place\") | (sdf.l1 == \"Work\") | (sdf.l1 == \"Species\")\n",
    "]\n",
    "cdf.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e3c7d3d3",
   "metadata": {},
   "source": [
    "&emsp;&emsp;接下来先把文本变成向量："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "id": "e280440d",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/xt/mpcnl_9151dbq43hptdlv2jr0000gn/T/ipykernel_69847/6636503.py:1: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  cdf[\"embedding\"] = cdf.text.apply(lambda x: get_embedding(x))\n"
     ]
    }
   ],
   "source": [
    "# 智谱AI更换为get_embedding_zhipu(x)\n",
    "cdf[\"embedding\"] = cdf.text.apply(lambda x: get_embedding(x))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "537df9e9",
   "metadata": {},
   "source": [
    "&emsp;&emsp;接下来用PCA（主成分分析）给降维，将原来的向量从1536维降到3维，便于显示。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "id": "b2e07d6f",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/xt/mpcnl_9151dbq43hptdlv2jr0000gn/T/ipykernel_69847/3198218742.py:5: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  cdf[\"embed_vis\"] = vis_dims.tolist()\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "((71, 1536), (71, 3))"
      ]
     },
     "execution_count": 129,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.decomposition import PCA\n",
    "arr = np.array(cdf.embedding.to_list())\n",
    "pca = PCA(n_components=3)\n",
    "vis_dims = pca.fit_transform(arr)\n",
    "cdf[\"embed_vis\"] = vis_dims.tolist()\n",
    "arr.shape, vis_dims.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "id": "15c8c8e8",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x800 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "\n",
    "fig, ax = plt.subplots(subplot_kw={\"projection\": \"3d\"}, figsize=(8, 8))\n",
    "cmap = plt.get_cmap(\"tab20\")\n",
    "categories = sorted(cdf.l1.unique())\n",
    "\n",
    "# 分别绘制每个类别\n",
    "for i, cat in enumerate(categories):\n",
    "    sub_matrix = np.array(cdf[cdf.l1 == cat][\"embed_vis\"].to_list())\n",
    "    x=sub_matrix[:, 0]\n",
    "    y=sub_matrix[:, 1]\n",
    "    z=sub_matrix[:, 2]\n",
    "    colors = [cmap(i/len(categories))] * len(sub_matrix)\n",
    "    ax.scatter(x, y, z, c=colors, label=cat)\n",
    "\n",
    "ax.legend(bbox_to_anchor=(1.2, 1))\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "84c0a00e",
   "metadata": {},
   "source": [
    "&emsp;&emsp;可以比较明显的看出，三个不同类型的数据分别在不同的位置。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5601da46",
   "metadata": {},
   "source": [
    "### 3.3 推荐"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3d202c15",
   "metadata": {},
   "source": [
    "&emsp;&emsp;我们在很多APP或网站上都能看到推荐功能。比如在购物网站，每当你登陆或者选购一件商品后，系统就会给你推荐一些相关的产品。在这一小节中，我们就来做一个类似的应用，不过我们推荐的不是商品，而是文本，比如帖子、文章、新闻等。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2e29531d",
   "metadata": {},
   "source": [
    "&emsp;&emsp;我们以新闻为例，先说一下基本逻辑：\n",
    "\n",
    "- 首先要有一个基础的文章库，可能包括标题、内容、标签等。\n",
    "- 计算已有文章的Embedding并存储。\n",
    "- 根据用户浏览记录，推荐和浏览记录最相似的文章。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d58913fc",
   "metadata": {},
   "source": [
    "&emsp;&emsp;使用如下的数据集：[AG News Classification Dataset | Kaggle](https://www.kaggle.com/datasets/amananandrai/ag-news-classification-dataset?select=train.csv)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "81e61909",
   "metadata": {},
   "source": [
    "&emsp;&emsp;看起来好像和前面的QA差不多，事实也的确如此，因为它们本质上都是相似匹配问题。只不过QA使用的是用户的Question去匹配已有知识库，而推荐是使用用户的浏览记录去匹配。但是很明显，推荐相比QA要更复杂一些，主要包括以下几个方面：\n",
    "\n",
    "- 刚开始用户没有记录时的推荐（一般行业称为冷启动问题）。\n",
    "- 除了相似还有其他要考虑的因素：比如热门内容、新内容、内容多样性、随时间变化的兴趣变化等等。\n",
    "- 编码（Embedding输入）问题：我们应该取标题呢，还是文章，还是简要描述或者摘要，还是都要计算。\n",
    "- 规模问题：推荐面临的量级一般会远超QA，除了横向扩展机器，是否能从流程和算法设计上提升效率。\n",
    "- 用户反馈对推荐系统的影响问题：用户反感或喜欢与文章本身并没有直接关系，比如用户喜欢体育新闻但讨厌中国足球。\n",
    "- 线上实时更新问题。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "567ba044",
   "metadata": {},
   "source": [
    "&emsp;&emsp;当然，一个完整的线上系统要考虑的因素可能更多。我们给大家列出这些只是希望读者在设计一个方案时能够充分调研和考虑，同时结合实际情况进行。反过来说，可能并不需要考虑上面的每个因素。所以，大家在实际操作时一定要活学活用，充分理解需求后再动手实施。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cedc2983",
   "metadata": {},
   "source": [
    "&emsp;&emsp;我们这里综合考虑上面的因素给大家一个比较简单的方案，但务必注意，其中每个模块的方案都不是唯一的。整体设计如下：\n",
    "\n",
    "- 用户注册登录时，让其选择感兴趣的类型（如体育、音乐、时尚等），我们通过这一步将用户框在一个大的范围内，同时用来顺道解决冷启动问题。\n",
    "- 给用户推荐内容时，在知道类别（用户注册时选择+浏览记录）后，应依次考虑时效性、热门程度、多样性等。\n",
    "- 考虑到性能问题，可以编码「标题+摘要」。\n",
    "- 对大类别进一步细分，只在细分类别里进行相似度计算。\n",
    "- 记录用户实时行为（如浏览Item、浏览时长、评论、收藏、点赞、转发等）。\n",
    "- 动态更新内容库，更新用户行为库。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8044d739",
   "metadata": {},
   "source": [
    "&emsp;&emsp;在具体实施时，我们使用最常用的流程线方案：召回+排序。\n",
    "\n",
    "- 召回：通过各种不同属性或特征（如用户偏好、热点、行为等）先找到一批要推荐列表。\n",
    "- 排序：根据多样性、时效性、用户反馈、热门程度等属性对结果进行排序。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 237,
   "id": "63249fae",
   "metadata": {},
   "outputs": [],
   "source": [
    "from dataclasses import dataclass\n",
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 238,
   "id": "3cc23305",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(120000, 3)"
      ]
     },
     "execution_count": 238,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv(\"./dataset/AG_News.csv\")\n",
    "df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 239,
   "id": "d491f791",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Class Index</th>\n",
       "      <th>Title</th>\n",
       "      <th>Description</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3</td>\n",
       "      <td>Wall St. Bears Claw Back Into the Black (Reuters)</td>\n",
       "      <td>Reuters - Short-sellers, Wall Street's dwindli...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3</td>\n",
       "      <td>Carlyle Looks Toward Commercial Aerospace (Reu...</td>\n",
       "      <td>Reuters - Private investment firm Carlyle Grou...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>Oil and Economy Cloud Stocks' Outlook (Reuters)</td>\n",
       "      <td>Reuters - Soaring crude prices plus worries\\ab...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>Iraq Halts Oil Exports from Main Southern Pipe...</td>\n",
       "      <td>Reuters - Authorities have halted oil export\\f...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3</td>\n",
       "      <td>Oil prices soar to all-time record, posing new...</td>\n",
       "      <td>AFP - Tearaway world oil prices, toppling reco...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Class Index                                              Title  \\\n",
       "0            3  Wall St. Bears Claw Back Into the Black (Reuters)   \n",
       "1            3  Carlyle Looks Toward Commercial Aerospace (Reu...   \n",
       "2            3    Oil and Economy Cloud Stocks' Outlook (Reuters)   \n",
       "3            3  Iraq Halts Oil Exports from Main Southern Pipe...   \n",
       "4            3  Oil prices soar to all-time record, posing new...   \n",
       "\n",
       "                                         Description  \n",
       "0  Reuters - Short-sellers, Wall Street's dwindli...  \n",
       "1  Reuters - Private investment firm Carlyle Grou...  \n",
       "2  Reuters - Soaring crude prices plus worries\\ab...  \n",
       "3  Reuters - Authorities have halted oil export\\f...  \n",
       "4  AFP - Tearaway world oil prices, toppling reco...  "
      ]
     },
     "execution_count": 239,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 240,
   "id": "3fb1a9d9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Class Index\n",
       "3    30000\n",
       "4    30000\n",
       "2    30000\n",
       "1    30000\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 240,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[\"Class Index\"].value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d9e0a957",
   "metadata": {},
   "source": [
    "&emsp;&emsp;根据数据集介绍（上面的链接），四个类型分别是：1-World, 2-Sports, 3-Business, 4-Sci/Tech，每个类型有30000条数据，共12万条。接下来，我们将使用上面已经介绍的知识来做一个简单的流水线系统。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "500930f9",
   "metadata": {},
   "source": [
    "&emsp;&emsp;为了便于运行，依然取100条sample作为示例："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 284,
   "id": "5c4b22d6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Class Index\n",
       "3    27\n",
       "2    26\n",
       "1    24\n",
       "4    23\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 284,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 指定种子，不然得不到下面一样的结果\n",
    "sdf = df.sample(100, random_state=26187)\n",
    "\n",
    "sdf[\"Class Index\"].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 285,
   "id": "1256f00f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Class Index                                                    2\n",
       "Title                 Swimming: Shibata Joins Japanese Gold Rush\n",
       "Description     ATHENS (Reuters) - Ai Shibata wore down Frenc...\n",
       "Name: 5303, dtype: object"
      ]
     },
     "execution_count": 285,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sdf.iloc[1]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6b60672a",
   "metadata": {},
   "source": [
    "&emsp;&emsp;首先维护一个用户偏好和行为记录："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 286,
   "id": "9926e533",
   "metadata": {},
   "outputs": [],
   "source": [
    "from typing import List\n",
    "@dataclass\n",
    "class User:\n",
    "    \n",
    "    user_name: str\n",
    "\n",
    "@dataclass\n",
    "class UserPrefer:\n",
    "    \n",
    "    user_name: str\n",
    "    prefers: List[int]\n",
    "\n",
    "\n",
    "@dataclass\n",
    "class Item:\n",
    "    \n",
    "    item_id: str\n",
    "    item_props: dict\n",
    "\n",
    "\n",
    "@dataclass\n",
    "class Action:\n",
    "    \n",
    "    action_type: str\n",
    "    action_props: dict\n",
    "\n",
    "\n",
    "@dataclass\n",
    "class UserAction:\n",
    "    \n",
    "    user: User\n",
    "    item: Item\n",
    "    action: Action\n",
    "    action_time: str"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 287,
   "id": "40790df7",
   "metadata": {},
   "outputs": [],
   "source": [
    "u1 = User(\"u1\")\n",
    "up1 = UserPrefer(\"u1\", [1, 2])\n",
    "# sdf.iloc[1] 正好是sport（类别为2）\n",
    "i1 = Item(\"i1\", {\n",
    "    \"id\": 1, \n",
    "    \"catetory\": \"sport\",\n",
    "    \"title\": \"Swimming: Shibata Joins Japanese Gold Rush\", \n",
    "    \"description\": \"\\\n",
    "    ATHENS (Reuters) - Ai Shibata wore down French teen-ager  Laure Manaudou to win the women's 800 meters \\\n",
    "    freestyle gold  medal at the Athens Olympics Friday and provide Japan with  their first female swimming \\\n",
    "    champion in 12 years.\", \n",
    "    \"content\": \"content\"\n",
    "})\n",
    "a1 = Action(\"浏览\", {\n",
    "    \"open_time\": \"2023-04-01 12:00:00\", \n",
    "    \"leave_time\": \"2023-04-01 14:00:00\",\n",
    "    \"type\": \"close\",\n",
    "    \"duration\": \"2hour\"\n",
    "})\n",
    "ua1 = UserAction(u1, i1, a1, \"2023-04-01 12:00:00\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "868d61e8",
   "metadata": {},
   "source": [
    "&emsp;&emsp;计算所有文本的Embedding，这一步和之前一样："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 288,
   "id": "33743282",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics.pairwise import cosine_similarity\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "id": "39aedd21",
   "metadata": {},
   "outputs": [],
   "source": [
    "sdf[\"embedding\"] = sdf.apply(lambda x: get_embedding(x.Title + x.Description), axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 292,
   "id": "4a4b5967",
   "metadata": {},
   "outputs": [],
   "source": [
    "# 智谱AI更改为 get_embedding_zhipu\n",
    "sdf[\"embedding\"] = sdf.apply(lambda x: get_embedding_zhipu(x.Title + x.Description), axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "66d5a6e5",
   "metadata": {},
   "source": [
    "&emsp;&emsp;我们先处理一下召回："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 293,
   "id": "a035bc3c",
   "metadata": {},
   "outputs": [],
   "source": [
    "import random\n",
    "class Recall:\n",
    "    \n",
    "    def __init__(self, df: pd.DataFrame):\n",
    "        self.data = df\n",
    "    \n",
    "    def user_prefer_recall(self, user, n):\n",
    "        up = self.get_user_prefers(user)\n",
    "        idx = random.randrange(0, len(up.prefers))\n",
    "        return self.pick_by_idx(idx, n)\n",
    "    \n",
    "    def hot_recall(self, n):\n",
    "        # 随机进行示例\n",
    "        df = self.data.sample(n)\n",
    "        return df\n",
    "    \n",
    "    def user_action_recall(self, user, n):\n",
    "        actions = self.get_user_actions(user)\n",
    "        interest = self.get_most_interested_item(actions)\n",
    "        recoms = self.recommend_by_interest(interest, n)\n",
    "        return recoms\n",
    "    \n",
    "    def get_most_interested_item(self, user_action):\n",
    "        \"\"\"\n",
    "        可以选近一段时间内用户交互时间、次数、评论（相关属性）过的Item\n",
    "        \"\"\"\n",
    "        # 就是sdf的第2行，idx为1的那条作为最喜欢（假设）\n",
    "        # 是一条游泳相关的Item\n",
    "        idx = user_action.item.item_props[\"id\"]\n",
    "        im = self.data.iloc[idx]\n",
    "        return im\n",
    "    \n",
    "    def recommend_by_interest(self, interest, n):\n",
    "        cate_id = interest[\"Class Index\"]\n",
    "        q_emb = interest[\"embedding\"]\n",
    "        # 确定类别\n",
    "        base = self.data[self.data[\"Class Index\"] == cate_id]\n",
    "        # 此处可以复用QA那一段代码，用给定embedding计算base中embedding的相似度\n",
    "        base_arr = np.array(\n",
    "            [v.embedding for v in base.itertuples()]\n",
    "        )\n",
    "        q_arr = np.expand_dims(q_emb, 0)\n",
    "        sims = cosine_similarity(base_arr, q_arr)\n",
    "        # 排除掉自己\n",
    "        idxes = sims.argsort(0).squeeze()[-(n+1):-1]\n",
    "        return base.iloc[reversed(idxes.tolist())]\n",
    "    \n",
    "    def pick_by_idx(self, category, n):\n",
    "        df = self.data[self.data[\"Class Index\"] == category]\n",
    "        return df.sample(n)\n",
    "    \n",
    "    def get_user_actions(self, user):\n",
    "        dct = {\"u1\": ua1}\n",
    "        return dct[user.user_name]\n",
    "    \n",
    "    def get_user_prefers(self, user):\n",
    "        dct = {\"u1\": up1}\n",
    "        return dct[user.user_name]\n",
    "    \n",
    "    def run(self, user):\n",
    "        ur = self.user_action_recall(user, 5)\n",
    "        if len(ur) == 0:\n",
    "            ur = self.user_prefer_recall(user, 5)\n",
    "        hr = self.hot_recall(3)\n",
    "        return pd.concat([ur, hr], axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 294,
   "id": "cb256c91",
   "metadata": {},
   "outputs": [],
   "source": [
    "r = Recall(sdf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 295,
   "id": "c7c3f152",
   "metadata": {},
   "outputs": [],
   "source": [
    "rd = r.run(u1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 296,
   "id": "69d1baa3",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Class Index</th>\n",
       "      <th>Title</th>\n",
       "      <th>Description</th>\n",
       "      <th>embedding</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>15201</th>\n",
       "      <td>2</td>\n",
       "      <td>Mauresmo Starts US Open With a Win</td>\n",
       "      <td>tarting Grand Slam tournaments is usually not ...</td>\n",
       "      <td>[-0.039115481078624725, 0.023394467309117317, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>901</th>\n",
       "      <td>2</td>\n",
       "      <td>American Aphrodite</td>\n",
       "      <td>Brooklyn-native Yvette Jarvis is an 'every wom...</td>\n",
       "      <td>[-0.004616602323949337, 0.06016676872968674, -...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45406</th>\n",
       "      <td>2</td>\n",
       "      <td>Greek hero Charisteas the man for the big occa...</td>\n",
       "      <td>Greek international Angelos Charisteas seems t...</td>\n",
       "      <td>[0.007043285761028528, 0.0375998318195343, -0....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37606</th>\n",
       "      <td>2</td>\n",
       "      <td>AL Wrap: Texas Completes Sweep of Oakland (Reu...</td>\n",
       "      <td>Reuters - David Dellucci's two-run double in t...</td>\n",
       "      <td>[-0.012884793803095818, 0.04332660138607025, -...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>109735</th>\n",
       "      <td>2</td>\n",
       "      <td>Washington list shrinks</td>\n",
       "      <td>Boston College football coach Tom O'Brien yest...</td>\n",
       "      <td>[0.020562877878546715, 0.027914173901081085, 0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>46082</th>\n",
       "      <td>2</td>\n",
       "      <td>IT Analysts Pick Key Contests</td>\n",
       "      <td>Two of the biggest games of the season are on ...</td>\n",
       "      <td>[-0.0059912255965173244, 0.06189167872071266, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>76848</th>\n",
       "      <td>1</td>\n",
       "      <td>Deutsche Bank third quarter at top end of expe...</td>\n",
       "      <td>AFP - Deutsche Bank said that its bottom-line ...</td>\n",
       "      <td>[-0.04670709744095802, 0.023904236033558846, 0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11525</th>\n",
       "      <td>3</td>\n",
       "      <td>Vodafone Plans Joint Australia 3G Rollout</td>\n",
       "      <td>Vodafone, which has some 330,000 customers in ...</td>\n",
       "      <td>[-0.01319542620331049, -0.02437293902039528, -...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        Class Index                                              Title  \\\n",
       "15201             2                 Mauresmo Starts US Open With a Win   \n",
       "901               2                                 American Aphrodite   \n",
       "45406             2  Greek hero Charisteas the man for the big occa...   \n",
       "37606             2  AL Wrap: Texas Completes Sweep of Oakland (Reu...   \n",
       "109735            2                            Washington list shrinks   \n",
       "46082             2                      IT Analysts Pick Key Contests   \n",
       "76848             1  Deutsche Bank third quarter at top end of expe...   \n",
       "11525             3          Vodafone Plans Joint Australia 3G Rollout   \n",
       "\n",
       "                                              Description  \\\n",
       "15201   tarting Grand Slam tournaments is usually not ...   \n",
       "901     Brooklyn-native Yvette Jarvis is an 'every wom...   \n",
       "45406   Greek international Angelos Charisteas seems t...   \n",
       "37606   Reuters - David Dellucci's two-run double in t...   \n",
       "109735  Boston College football coach Tom O'Brien yest...   \n",
       "46082   Two of the biggest games of the season are on ...   \n",
       "76848   AFP - Deutsche Bank said that its bottom-line ...   \n",
       "11525   Vodafone, which has some 330,000 customers in ...   \n",
       "\n",
       "                                                embedding  \n",
       "15201   [-0.039115481078624725, 0.023394467309117317, ...  \n",
       "901     [-0.004616602323949337, 0.06016676872968674, -...  \n",
       "45406   [0.007043285761028528, 0.0375998318195343, -0....  \n",
       "37606   [-0.012884793803095818, 0.04332660138607025, -...  \n",
       "109735  [0.020562877878546715, 0.027914173901081085, 0...  \n",
       "46082   [-0.0059912255965173244, 0.06189167872071266, ...  \n",
       "76848   [-0.04670709744095802, 0.023904236033558846, 0...  \n",
       "11525   [-0.01319542620331049, -0.02437293902039528, -...  "
      ]
     },
     "execution_count": 296,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 共8个，5个用户行为推荐、3个热门\n",
    "rd"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ff07b532",
   "metadata": {},
   "source": [
    "&emsp;&emsp;需要再次说明的是，上面的只是一个大致的流程，实际中有很多细节或优化点需要注意，比如：\n",
    "\n",
    "- 建数据库表（上面的`get_`其实都是查表）\n",
    "- 将Item、User和Action也进行Embedding，全部使用Embedding后再做召回\n",
    "- 对『感兴趣`get_most_interested_item`』更多的优化，考虑更多行为和反馈，召回更多不同类型条目\n",
    "- 性能和自动更新数据的考虑\n",
    "- 线上评测，A/B等"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "971608ef",
   "metadata": {},
   "source": [
    "&emsp;&emsp;可以发现，我们虽然只做了召回一步，但其中涉及到的内容已经远远不止之前QA那一点了，QA用到的东西可能只是其中一小部分。不过事无绝对，即便是QA任务也可能根据实际情况不同需要做很多优化，比如召回+排序。但总体来说，类似推荐这样比较综合的系统相对来说会更加复杂一些。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4231395b",
   "metadata": {},
   "source": [
    "&emsp;&emsp;后面就是排序了，这一步需要区分不同的应用场景，可以做或不做，做的话（就是对刚刚得到的列表进行排序）也可以简单或复杂。比如简单地按发布时间，复杂的综合考虑多样性、时效性、用户反馈、热门程度等多种属性。具体操作时，可以直接按相关属性排序，也可以用模型排序。这里就不再继续深入了。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9b0ca8fd",
   "metadata": {},
   "source": [
    "## 相关文献\n",
    "\n",
    "- 【1】[浅析文本分类 —— 情感分析与自然语言处理 | Yam](https://yam.gift/2021/10/27/NLP/2021-10-27-Senta/)\n",
    "- 【2】[句子表征综述 | Yam](https://yam.gift/2022/03/27/NLP/2022-03-27-Sentence-Representation-Summarization/)\n",
    "- 【3】[NLP 表征的历史与未来 | Yam](https://yam.gift/2020/12/12/NLP/2020-12-12-NLP-Representation-History-Future/)\n",
    "\n",
    "\n",
    "memo：\n",
    "\n",
    "- [openai-cookbook/Semantic_text_search_using_embeddings.ipynb at main · openai/openai-cookbook](https://github.com/openai/openai-cookbook/blob/main/examples/Semantic_text_search_using_embeddings.ipynb)\n",
    "- [openai-cookbook/getting-started-with-redis-and-openai.ipynb at main · openai/openai-cookbook](https://github.com/openai/openai-cookbook/blob/main/examples/vector_databases/redis/getting-started-with-redis-and-openai.ipynb)\n",
    "- [openai-cookbook/Visualizing_embeddings_in_3D.ipynb at main · openai/openai-cookbook](https://github.com/openai/openai-cookbook/blob/main/examples/Visualizing_embeddings_in_3D.ipynb)\n",
    "\n",
    "注①：感谢DataWhale成员于沼懿指出此处不严谨描述。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "dd85cefd",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.18"
  },
  "latex_envs": {
   "LaTeX_envs_menu_present": true,
   "autoclose": false,
   "autocomplete": true,
   "bibliofile": "biblio.bib",
   "cite_by": "apalike",
   "current_citInitial": 1,
   "eqLabelWithNumbers": true,
   "eqNumInitial": 1,
   "hotkeys": {
    "equation": "Ctrl-E",
    "itemize": "Ctrl-I"
   },
   "labels_anchors": false,
   "latex_user_defs": false,
   "report_style_numbering": false,
   "user_envs_cfg": false
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": false,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": true,
   "toc_position": {
    "height": "calc(100% - 180px)",
    "left": "10px",
    "top": "150px",
    "width": "175px"
   },
   "toc_section_display": true,
   "toc_window_display": true
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}