02_zero_shot_classification.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "3860afdb",
   "metadata": {},
   "source": [
    "## 大语言模型 - 10行代码的情感分析解决方案"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "843b09f3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "好评例子的评分 : 0.040873\n",
      "差评例子的评分 : -0.016688\n"
     ]
    }
   ],
   "source": [
    "from openai import OpenAI\n",
    "import numpy as np\n",
    "import os\n",
    "\n",
    "client = OpenAI(api_key=os.environ['OPENAI_API_KEY'])\n",
    "\n",
    "EMBEDDING_MODEL = \"text-embedding-ada-002\"\n",
    "\n",
    "def get_embedding(text, model=EMBEDDING_MODEL):\n",
    "   text = text.replace(\"\\n\", \" \")\n",
    "   return client.embeddings.create(input = [text], model=model).data[0].embedding\n",
    "\n",
    "def cosine_similarity(vector_a, vector_b):\n",
    "  dot_product = np.dot(vector_a, vector_b)\n",
    "  norm_a = np.linalg.norm(vector_a)\n",
    "  norm_b = np.linalg.norm(vector_b)\n",
    "  epsilon = 1e-10\n",
    "  cosine_similarity = dot_product / (norm_a * norm_b + epsilon)\n",
    "  return cosine_similarity\n",
    "\n",
    "positive_review = get_embedding(\"好评\")\n",
    "negative_review = get_embedding(\"差评\")\n",
    "\n",
    "positive_example = get_embedding(\"买的银色版真的很好看，一天就到了，晚上就开始拿起来完系统很丝滑流畅，做工扎实，手感细腻，很精致哦苹果一如既往的好品质\")\n",
    "negative_example = get_embedding(\"随意降价，不予价保，服务态度差\")\n",
    "\n",
    "def get_score(sample_embedding):\n",
    "  return cosine_similarity(sample_embedding, positive_review) - cosine_similarity(sample_embedding, negative_review)\n",
    "\n",
    "positive_score = get_score(positive_example)\n",
    "negative_score = get_score(negative_example)\n",
    "\n",
    "print(\"好评例子的评分 : %f\" % (positive_score))\n",
    "print(\"差评例子的评分 : %f\" % (negative_score))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8b19192a",
   "metadata": {},
   "source": [
    "## 语序问题"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "90a515e9",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "好评餐馆的评分 : 0.051292\n",
      "差评餐馆的评分 : -0.006322\n"
     ]
    }
   ],
   "source": [
    "good_restraurant = get_embedding(\"这家餐馆太好吃了，一点都不糟糕\")\n",
    "bad_restraurant = get_embedding(\"这家餐馆太糟糕了，一点都不好吃\")\n",
    "\n",
    "good_score = get_score(good_restraurant)\n",
    "bad_score = get_score(bad_restraurant)\n",
    "print(\"好评餐馆的评分 : %f\" % (good_score))\n",
    "print(\"差评餐馆的评分 : %f\" % (bad_score))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1b611338",
   "metadata": {},
   "source": [
    "## 在更大的数据集上的真实案例"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "97dac5a2",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "from sklearn.metrics import classification_report\n",
    "\n",
    "datafile_path = os.environ.get(\"JUPYTER_HOME\") + \"/data/fine_food_reviews_with_embeddings_1k.csv\"\n",
    "\n",
    "df = pd.read_csv(datafile_path)\n",
    "df[\"embedding\"] = df.embedding.apply(eval).apply(np.array)\n",
    "\n",
    "# convert 5-star rating to binary sentiment\n",
    "df = df[df.Score != 3]\n",
    "df[\"sentiment\"] = df.Score.replace({1: \"negative\", 2: \"negative\", 4: \"positive\", 5: \"positive\"})"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7a8346aa",
   "metadata": {},
   "source": [
    "## 生成评测结果"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "38513ac7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "    negative       0.98      0.73      0.84       136\n",
      "    positive       0.96      1.00      0.98       789\n",
      "\n",
      "    accuracy                           0.96       925\n",
      "   macro avg       0.97      0.86      0.91       925\n",
      "weighted avg       0.96      0.96      0.96       925\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.metrics import PrecisionRecallDisplay\n",
    "\n",
    "def evaluate_embeddings_approach(\n",
    "    labels = ['negative', 'positive'], \n",
    "    model = EMBEDDING_MODEL,\n",
    "):\n",
    "    label_embeddings = [get_embedding(label, model=model) for label in labels]\n",
    "\n",
    "    def label_score(review_embedding, label_embeddings):\n",
    "        return cosine_similarity(review_embedding, label_embeddings[1]) - cosine_similarity(review_embedding, label_embeddings[0])\n",
    "\n",
    "    probas = df[\"embedding\"].apply(lambda x: label_score(x, label_embeddings))\n",
    "    preds = probas.apply(lambda x: 'positive' if x>0 else 'negative')\n",
    "\n",
    "    report = classification_report(df.sentiment, preds)\n",
    "    print(report)\n",
    "\n",
    "    display = PrecisionRecallDisplay.from_predictions(df.sentiment, probas, pos_label='positive')\n",
    "    _ = display.ax_.set_title(\"2-class Precision-Recall curve\")\n",
    "\n",
    "evaluate_embeddings_approach(labels=['An Amazon review with a negative sentiment.', 'An Amazon review with a positive sentiment.'])\n"
   ]
  }
 ],
 "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": "3.10.1"
  },
  "vscode": {
   "interpreter": {
    "hash": "8114e84f04cf14e493992e1b725447accf84073d5ec18e7063d492738bf032cb"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}