Create chatgpt

chatgpt

@@ -0,0 +1,81 @@
+#!/usr/bin/env python3
+import openai
+import configparser
+import arxiv
+import random
+import argparse
+
+config = configparser.ConfigParser()
+config.read('.config')
+
+openai.api_key = config.get('open_api_key', 'key')
+
+
+def summarize_paper(paper):
+ system = """
+ 論文を以下の制約に従って要約して出力してください。
+
+ [制約]
+ タイトルは日本語で書く
+ 要点は3つにまとめる
+
+
+ [出力]
+ タイトルの日本語訳
+
+ ・要点1
+ ・要点2
+ ・要点3
+ """
+
+ text = f"title: {paper.title}\nbody: {paper.summary}"
+ response = openai.ChatCompletion.create(
+ model="gpt-3.5-turbo",
+ messages=[
+ {'role': 'system', 'content': system},
+ {'role': 'user', 'content': text}
+ ],
+ temperature=0.2,
+ )
+
+ summary = response['choices'][0]['message']['content']
+ date_str = paper.published.strftime("%Y-%m-%d %H:%M:%S")
+ message = f"発行日: {date_str}\n{paper.entry_id}\n{paper.title}\n{summary}\n"
+ return message
+
+def get_arxiv(query: str, paper_all_numb: int = 5, paper_select_numb: int = 3):
+ # search arxiv paper
+ result = arxiv.Search(
+ query=query,
+ max_results=paper_all_numb,
+ sort_by=arxiv.SortCriterion.SubmittedDate,
+ sort_order=arxiv.SortOrder.Descending,
+ ).results()
+
+ return random.sample(list(result), k=paper_select_numb)
+
+
+def get_specific_paper_from_arxiv(arxiv_id: str):
+ result = arxiv.Search(id_list=[arxiv_id]).results()
+
+ return list(result)
+
+
+if __name__ == "__main__":
+ parser = argparse.ArgumentParser(description='--paper_id is arxiv paper id')
+
+ parser.add_argument("-i", "--paper_id", type=str, default="None", help="arxiv paper id")
+
+ args = parser.parse_args()
+
+ if args.paper_id != "None":
+ paper = get_specific_paper_from_arxiv(arxiv_id=args.paper_id)
+ print(summarize_paper(paper[0]))
+ else:
+ paper_list = get_arxiv(query='deep learning', paper_all_numb=100, paper_select_numb=3)
+ for i, paper in enumerate(paper_list):
+ try:
+ print(str(i+1) + '本目の論文')
+ print(summarize_paper(paper))
+ except:
+ print('error')