CAJ | 학술논문

简单命题的语义识别是进行自然命题逻辑推理的重要基础之一，知网在中文词汇和句子的相似度计算方面扮演着越来越重要的角色。在对领域命题进行了预处理后，针对简单命题的句型结构和构成成分的特点，通过改进义原相似度、词语相似度、句子相似度的计算方法，提出了一种基于知网的、新的句子相似度计算方法。实验表明：该方法可以简单而有效地计算词语相似度；在句子相似度计算时，可以减少助动词、修饰词的影响，更适合判断两个命题是否高度相似。
간단명제적어의식별시진행자연명제라집추리적중요기출지일，지망재중문사회화구자적상사도계산방면분연착월래월중요적각색。재대영역명제진행료예처리후，침대간단명제적구형결구화구성성분적특점，통과개진의원상사도、사어상사도、구자상사도적계산방법，제출료일충기우지망적、신적구자상사도계산방법。실험표명：해방법가이간단이유효지계산사어상사도；재구자상사도계산시，가이감소조동사、수식사적영향，경괄합판단량개명제시부고도상사。
Semantic recognition for simple propositions is an important foundation of logical reasoning, and Hownet plays an increasingly important role in Chinese word and sentence similarity computing.After prepro-cessing for domain propositions,aiming at the characteristic of sentence structure and constituent components a-bout simple propositions,by improving sememe similarity,word similarity and sentence similarity calculation methods,a new sentence similarity calculation approach based on Hownet is proposed.Experiments show that the method can compute word similarity easily and efficiently.When considering sentence similarity calcula-tion,it can reduce the effects of auxiliary verbs and modifiers,and is more suitable to determine whether the two propositions are highly similar.