layer2 commit

layer2/process/step4.py

@@ -0,0 +1,78 @@
+"""
+对821道英文问题进行处理
+1. 判断是否包含多个子问题，将问题拆分为完整子问题(去掉推理过程，只保留最后结果)
+2. 判断题目类型
+3. 将题目做成选择题
+   对计算题，在数值附近随机生成三个相似答案作为错误选项
+   对简答题，与标准答案最相近的其他问题的答案作为三个错误选项
+4. 将正确和错误选项随机打乱，生成ABCD选择题的模型
+5. 添加prompt，并将选择题包裹在[ANSWER]<answer>[/ANSWER]标签中
+6. 模型打分
+"""
+import json
+import random
+from typing import List, Dict
+
+def process_json_file(file_path: str) -> List[Dict]:
+    with open(file_path, 'r', encoding='utf-8') as f:
+        data = json.load(f)
+
+    for item in data:
+        # 收集所有选项
+        options = [
+            item['answer'],
+            item.get('wrong_answers_1', ''),
+            item.get('wrong_answers_2', ''),
+            item.get('wrong_answers_3', '')
+        ]
+
+        # 过滤掉空选项
+        options = [opt for opt in options if opt]
+
+        # 打乱选项
+        random.shuffle(options)
+
+        # 找出正确答案的位置
+        correct_answer_index = options.index(item['answer'])
+        correct_answer_letter = chr(65 + correct_answer_index)  # A, B, C, or D
+
+        # 构建选项文本
+        options_text = ""
+        for i, option in enumerate(options):
+            letter = chr(65 + i)  # A, B, C, or D
+            options_text += f"({letter}){option}"
+            if i < len(options) - 1:
+                options_text += " "
+
+        # 更新问题和答案
+        item['question'] = f"{"The following is a question about Fundamentals of Materials Science"}{item['question']} {options_text}{"You MUST include the letter(s) of the correct answer (separated by comma if there are many) within the following tags: [ANSWER] and [/ANSWER].\nFor example, '[ANSWER]<answer>[/ANSWER]', where <answer> is comma- or space-separated list of the correct letters. Always answer in exactly this format of comma-separated letters between the two tags, even if you are unsure. We require this because we use automatic parsing."}"
+        item['answer'] = f"[ANSWER]{correct_answer_letter}[/ANSWER]"
+
+        # 删除原始的错误选项
+        if 'wrong_answers_1' in item:
+            del item['wrong_answers_1']
+        if 'wrong_answers_2' in item:
+            del item['wrong_answers_2']
+        if 'wrong_answers_3' in item:
+            del item['wrong_answers_3']
+
+    return data
+
+def save_processed_data(data: List[Dict], output_path: str) -> None:
+    """
+    保存处理后的数据到新的JSON文件
+    """
+    with open(output_path, 'w', encoding='utf-8') as f:
+        json.dump(data, f, ensure_ascii=False, indent=2)
+
+# 使用示例
+if __name__ == "__main__":
+    input_file = "/home/ubuntu/50T/fsy/5_1.json"  # 替换为你的输入文件路径
+    output_file = "output.json"  # 替换为你想要的输出文件路径
+
+    try:
+        processed_data = process_json_file(input_file)
+        save_processed_data(processed_data, output_file)
+        print(f"处理完成！结果已保存到 {output_file}")
+    except Exception as e:
+        print(f"处理过程中出现错误: {e}")