from sentence_transformers import SentenceTransformer
from sklearn.metrics.pairwise import cosine_similarity
import numpy as np

class TextSimilarityScorer:
    def __init__(self, model_name='bert-base-chinese'):
        # 初始化 BERT 模型，这里使用中文预训练模型
        self.model = SentenceTransformer(model_name)
        
    def calculate_similarity(self, standard_answer, student_answer):
        # 获取文本嵌入
        standard_embedding = self.model.encode([standard_answer])
        student_embedding = self.model.encode([student_answer])
        
        # 计算余弦相似度
        similarity = cosine_similarity(standard_embedding, student_embedding)[0][0]
        return similarity
    
    def score(self, standard_answer, student_answer, max_score=100):
        # 计算相似度
        similarity = self.calculate_similarity(standard_answer, student_answer)
        
        # 将相似度转换为分数
        score = similarity * max_score
        
        # 四舍五入到整数
        return round(score)

# 使用示例
def main():
    # 初始化评分器
    scorer = TextSimilarityScorer()
    
    # 示例标准答案和学生答案
    standard_answer = "机器学习是人工智能的一个子领域，它使用统计学方法让计算机系统能够从数据中学习和改进。"
    student_answers = [
        "机器学习是AI的分支，通过统计方法让计算机从数据中学习。",  # 相似但较简短
        "哈哈哈哈哈哈。",  # 比较相似
        "人工智能是计算机科学的重要领域。"  # 不太相关
    ]
    
    # 对每个学生答案进行评分
    for i, student_answer in enumerate(student_answers, 1):
        similarity = scorer.calculate_similarity(standard_answer, student_answer)
        score = scorer.score(standard_answer, student_answer)
        print(f"\n学生答案 {i}:")
        print(f"答案: {student_answer}")
        print(f"相似度: {similarity:.2f}")
        print(f"得分: {score}")

if __name__ == "__main__":
    main()