File size: 5,049 Bytes
f1bf1b2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "\n",
    "with open('../data/task4.json') as f:\n",
    "    data = json.load(f)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# If use Together API\n",
    "from together import Together\n",
    "\n",
    "your_model_name = 'xxxxxxxxxxxxxxxxxxxxxxxx'\n",
    "your_api_key = 'xxxxxxxxxxxxxxxxxxxxxxxx'\n",
    "client = Together(api_key=your_api_key)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "for draft in data:\n",
    "    statements = draft['Statements']\n",
    "    for statement in statements:\n",
    "        country = statement['country']\n",
    "        system_prompt = f\"\"\"\n",
    "        Assume you are the representative of {country} in UNSC, given the content of a UNSC draft resolution, generate a statement that you would make in the meeting.\n",
    "        \"\"\"\n",
    "        user_prompt = f\"\"\"\n",
    "        Draft resolution: {draft['Content']}\n",
    "        Your statement:\n",
    "        \"\"\"\n",
    "        response = client.chat.completions.create(\n",
    "            model=your_model_name,\n",
    "            temperature=0.0,\n",
    "            messages=[\n",
    "                {\"role\": \"system\", \"content\": system_prompt},\n",
    "                {\"role\": \"user\", \"content\": user_prompt}\n",
    "            ]\n",
    "        )\n",
    "        statement['generation'] = response.choices[0].message.content\n",
    "    draft['Statements'] = statements"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "statements = []\n",
    "generations = []\n",
    "\n",
    "for draft in data:\n",
    "    for statement in draft['Statements']:\n",
    "        statements.append(statement['statement'])\n",
    "        generations.append(statement['generation'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# average ROUGE\n",
    "from rouge_score import rouge_scorer\n",
    "scorer = rouge_scorer.RougeScorer(['rougeL'], use_stemmer=True)\n",
    "\n",
    "rouge_scores = []\n",
    "for i in range(len(statements)):\n",
    "    scores = scorer.score(statements[i], generations[i])\n",
    "    rouge_scores.append(scores['rougeL'].fmeasure)\n",
    "\n",
    "print('Average ROUGE-L:', sum(rouge_scores) / len(rouge_scores))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# average Jaccard similarity\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "from sklearn.metrics import jaccard_score\n",
    "\n",
    "jaccard_scores = []\n",
    "vectorizer = CountVectorizer(binary=True)\n",
    "\n",
    "for i in range(len(statements)):\n",
    "    X = vectorizer.fit_transform([statements[i], generations[i]]).toarray()\n",
    "    jaccard_scores.append(jaccard_score(X[0], X[1]))\n",
    "\n",
    "print('Average Jaccard similarity:', sum(jaccard_scores) / len(jaccard_scores))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# average Cosing Similarity (TF-IDF)\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.metrics.pairwise import cosine_similarity\n",
    "\n",
    "tf_cosine_scores = []\n",
    "for i in range(len(statements)):\n",
    "    tfidf = TfidfVectorizer().fit_transform([statements[i], generations[i]])\n",
    "    tf_cosine_scores.append(cosine_similarity(tfidf[0], tfidf[1])[0][0])\n",
    "\n",
    "print('Average Cosine Similarity (TF-IDF):', sum(tf_cosine_scores) / len(tf_cosine_scores))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sentence_transformers import SentenceTransformer, util\n",
    "\n",
    "device = 'cuda:1'\n",
    "model = SentenceTransformer('stsb-roberta-large', device=device)\n",
    "cosine_scores = []\n",
    "for i in range(len(statements)):\n",
    "    embeddings = model.encode([statements[i], generations[i]])\n",
    "    cosine_scores.append(util.pytorch_cos_sim(embeddings[0], embeddings[1]).item())\n",
    "\n",
    "print('Average Cosine Similarity (BERT):', sum(cosine_scores) / len(cosine_scores))\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "llm",
   "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": "3.9.19"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}