基于语句结构及语义相似度计算主观题评分算法的研究

Research on assessment algorithm of subjective issue based on sentence structure and semantic similarity computation

文字类主观题的自动评分是实现远程教育中在线考试系统的一个关键技术,由于其自动评判具有相当难度,使自动评分系统中在对语句结构、关键字匹配、词性、词义以及语义方面的判断还存在很多问题。通过对已有的算法分析,提出了一种方法,采用浅层次句法结构分析和深层次语义分析相结合的算法计算相似度,该方法可以提高主观题自动评分的效率和准确度,具有一定的实用价值。

Automated assessment technology of the subject issue based on free texts is one key techniques of online test in the distance teaching system. There is so much difficulty to realize the automated scoring system of subjective issue. There are many problems about sentence structure, keyword matching, gender, acceptation and semantic analysis and so on. This paper presents a new algorithm, which adopted low level sentence and deep semantic similarity computation. This method has the practical value to a certain degree because it can improve the efficiency and accuracy of automatic check about the subject issue.