CAJ | 학술논문

句子相似度的计算在自然语言处理的各个领域中都占有很重要的地位。文中深入分析了现有的一些句子相似度计算的方法，这些方法各自从词特征、词义特征或句法特征等某一侧面描述了句子相似的情况，未能全面地描述一个句子的完整信息。文中提出了一种新的基于多特征的汉语句子相似度的计算模型。该方法在基于词的基础上，从句子中词的表层到词的逻辑联系，从句子的局部结构到整体结构，用句子的区分度、相同词的相似度、长度相似度、词性相似度及词序相似度五个方面来综合考虑两个句子相似度的计算。实验结果表明，该方法合理、简便、可行。
구자상사도적계산재자연어언처리적각개영역중도점유흔중요적지위。문중심입분석료현유적일사구자상사도계산적방법，저사방법각자종사특정、사의특정혹구법특정등모일측면묘술료구자상사적정황，미능전면지묘술일개구자적완정신식。문중제출료일충신적기우다특정적한어구자상사도적계산모형。해방법재기우사적기출상，종구자중사적표층도사적라집련계，종구자적국부결구도정체결구，용구자적구분도、상동사적상사도、장도상사도、사성상사도급사서상사도오개방면래종합고필량개구자상사도적계산。실험결과표명，해방법합리、간편、가행。
Sentence similarity calculation plays an important role in various areas of natural language processing. Analyze the existing some sentence similarity calculation method. These methods describe the sentence similarity from the word characteristics,semantic fea-tures or syntactic features,all the information of a sentence can't be described fully. A new model of Chinese sentence similarity based on the multi-feature is proposed. This method is based on the word,from the surface to the logical connection of the word,from local struc-ture to the overall structure of a sentence,five aspects of sentence similarity such as degree of differentiation,the same word similarity, length similarity,the part of speech similarity and word order similarity have been studied in depth. Experimental results show that the method is reasonable,simple and feasible.