结合学科同义词与词向量的相似度评分算法  

Similarity scoring algorithm combining subject synonyms and word vectors

作　　者：付鹏斌 杨广越 杨惠荣 FU Peng-bin;YANG Guang-yue;YANG Hui-rong(Faculty of Information Technology,Beijing University of Technology,Beijing 100124,China)

机构地区：[1]北京工业大学信息学部,北京100124

出　　处：《计算机工程与设计》2020年第12期3390-3396,共7页Computer Engineering and Design

基　　金：北京市自然科学基金项目(4153058)。

摘　　要：为实现主观题自动评分,提出结合学科同义词与词向量的相似度评分算法。以地理学科为背景,通过提取学科知识信息,建立地理词典,将其引入Word2vec模型中训练词向量,构建地理语料库;针对同义词词林对学科同义词识别不准确的问题,建立地理同义词词库;基于词性提出一种关键词提取及权重分配算法,在文本相似度计算中融入学科知识背景,依据词语相似度建立语句相似度的可信值,实现相似度评分算法。实验结果表明,该方法与教师评分趋势基本一致,评分准确率达到了88.82%。To achieve automatic scoring of subjective questions,a similarity scoring algorithm combining subject synonyms and word vectors was proposed.Taking geography as the background,a geographic dictionary was established by extracting subject knowledge information,and it was introduced into the Word2vec model to train word vectors to build a geographic corpus.Aiming at the problem of inaccurate identification of subject synonyms by the Tongyici Cilin,a geographic synonym lexicon was established.Based on part-of-speech,a keyword extraction and weight allocation algorithm was proposed,the subject knowledge background was incorporated into the text similarity calculation,and according to the similarity of words,the confidence value of sentence similarity was established,a similarity scoring algorithm was implemented.Experimental results show that the performance of the proposed method is basically consistent with the trend of teacher scoring,and its accuracy of scoring reaches 88.82%.

关 键 词：主观题 自动评分 语料库 关键词提取 相似度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]