摄影领域评论情感词典构建方法  

Construction method of sentiment lexicon for photography reviews

作　　者：刘亚桥 陆向艳[1,2] 邓凯凯 阮开栋 刘峻 LIU Ya-qiao;LU Xiang-yan;DENG Kai-kai;RUAN Kai-dong;LIU Jun(School of Computer and Electronic Information,Guangxi University,Nanning 530004,China;Information Science Experiment Center of Guangxi,Guilin University of Electronic Technology,Guilin 541004,China)

机构地区：[1]广西大学计算机与电子信息学院,广西南宁530004 [2]桂林电子科技大学广西信息科学实验中心,广西桂林541004

出　　处：《计算机工程与设计》2019年第10期3037-3042,共6页Computer Engineering and Design

基　　金：广西软科学研究计划基金项目(桂科AB17205002);广西高校科研基金项目(KY2015YB008);广西大学科研基金项目(XJZ130355);广西研究生教育创新计划基金项目(JGY2015004)

摘　　要：目前少见摄影领域情感词典,且为改善Word2Vec情感词典构建方法词语领域区分度,提出一种改进Word2Vec的摄影领域情感词典构建方法。用TF-IDF训练结果词集及人工摄影情感基础词集对Word2Vec训练结果集进行领域重要程度修正,产生摄影领域情感词集,从大连理工大学情感词汇本体库中选取种子词和情感词集计算情感词的极性和强度,和HowNet情感词典合并形成摄影领域情感词典。实验结果表明,该方法在实际摄影评论数据集上取得了较好的效果。There is no sentiment lexicon in photography.The construction method of sentiment lexicon based on Word2Vec lacks domain discrimination of sentiment words.To solve these two problems,an improved construction method of sentiment lexicon in photography based on Word2Vec was proposed.TF-IDF training result words set and artificially labeled photographic basic affective words set were used to correct the domain importance of Word2Vec training results,to produce an sentiment words set in photography.Seed words were selected from the sentiment vocabulary library of DUTIR and were used to compute with sentiment words set produced,to generate the polarity and intensity of sentiment words.The sentiment words set in photography was combined with HowNet sentiment lexicon to generate sentiment lexicon in photography.Experimental results show that the proposed method is effective in the actual dataset of photography.

关 键 词：摄影评论 情感词典 Word2Vec TF-IDF 领域区分度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]