71 lines
2.7 KiB
Python
71 lines
2.7 KiB
Python
import os
|
||
|
||
os.environ['HF_HUB_DISABLE_SYMLINKS_WARNING'] = '1'
|
||
|
||
from sentence_transformers import SentenceTransformer
|
||
from sklearn.metrics.pairwise import cosine_similarity
|
||
import numpy as np
|
||
|
||
|
||
class TextSimilarityScorer:
|
||
def __init__(self, model_name='shibing624/text2vec-base-chinese'):
|
||
# 使用专门针对中文优化的模型
|
||
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):
|
||
# 基础相似度
|
||
similarity = self.calculate_similarity(standard_answer, student_answer)
|
||
|
||
# 长度比例
|
||
len_ratio = min(len(student_answer) / len(standard_answer), 1.0)
|
||
|
||
# 关键词覆盖度（简单实现）
|
||
standard_keywords = set(standard_answer.split())
|
||
student_keywords = set(student_answer.split())
|
||
keyword_coverage = len(student_keywords.intersection(standard_keywords)) / len(standard_keywords)
|
||
|
||
# 综合评分
|
||
final_score = (
|
||
similarity * 0.6 + # 语义相似度权重
|
||
len_ratio * 0.2 + # 长度比例权重
|
||
keyword_coverage * 0.2 # 关键词覆盖权重
|
||
) * max_score
|
||
|
||
return round(final_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
|
||
"机器学习是计算机科学的一个领域，使用统计方法从数据中学习模式。", # 比较相似
|
||
"人工智能是计算机科学的重要领域。",
|
||
"哈哈哈哈哈哈哈哈。"# 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() |