基于HNC理论的问答社区中答案摘要生成研究  

Research on Answer Summary Generation in Q&A Community Based on the HNC Theory

作　　者：王宇[1] 王芳[1] WANG Yu;WANG Fang(School of Economics and Management,DUT,Dalian 116024,China)

机构地区：[1]大连理工大学经济管理学院,辽宁大连116024

出　　处：《信息系统学报》2021年第2期1-12,共12页China Journal of Information Systems

摘　　要：针对问答社区中意见型问题答案文本情感突出的情况,首先提出一种补全未收录词HNC符号的方式,然后利用HNC理论的相关规律进行情感词的定位和情感值计算,进而利用情感值对答案句子进行聚类,生成积极、消极和中性三类情感句子类簇,最后提出基于最佳答案模板的相似度排序算法,抽取出3类情感答案摘要。实验证明,本文提出的方法在实例中的应用与人工生成的最佳摘要更为接近,尤其在摘要可读性、逻辑性方面比其他方法更具优势。In order to solve the problem of the esoteric questions in the Q&A community,the paper proposes a method to supplement the HNC symbols of unregistered words,then uses the HNC theory to locate the emotional words and calculate the emotional values.The emotional similarity is used to cluster the answer sentences to generate positive,negative and neutral emotional sentence clusters.Finally,a similarity ranking algorithm based on the best answer template is proposed,and three types of emotional answer summaries are extracted.The experimental results show that the application of the proposed method is closer to the best summary generated by the artificial,especially in terms of the readability and logic of the abstract.

关 键 词：问答社区 答案摘要 HNC理论 情感值计算 最佳答案模板 

分 类 号：G203[文化科学—传播学]