基于领域本体和依存句法分析的主观题自动评分方法

Automatic Scoring Method for Subjective Questions Based on Domain Ontology and Dependency Parsing

现有主观题自动评分方法大多无法识别电力系统领域的专业术语,且在评分过程中易遗漏文本间的语义关系,进而导致其评分结果与人工评分结果偏差较大,无法满足实际考试的要求。针对这些问题,本文提出一种基于领域本体和依存句法分析的主观题自动评分方法。该方法综合了节点距离相似度、节点信息相似度、通用词语相似度和依存句法搭配词对相似度等因素,并将电力系统领域本体引入到评分过程中,进而提高评分结果的有效性。实验结果表明:与现有方法相比,本文评分方法在主观题自动评分中更贴近人工评分结果,在MAE、RMSE和SMAPE指标上优于其他方法。

Most of the existing automatic scoring methods of subjective questions can not identify the professional words in the field of power system,and the semantic relationship between texts is omitted in the process of scoring,which leads to the deviation between the scoring results and the manual scoring results and can not meet the requirements of the actual examination.To solve these problems,this paper proposes an automatic scoring method for subjective questions based on domain ontology and dependency parsing.This method integrates factors such as node distance similarity,node information similarity,common word similarity and dependent syntactic collocation word pair similarity,and introduces the domain ontology of power system into the rating process,thereby improving the effectiveness of the scoring results.The experimental results show that,compared with the existing methods,the scoring method in this paper is closer to the manual scoring results in the automatic scoring of subjective questions,and is better than other methods on the MAE,RMSE and SMAPE indicators.