基于情感分析和概念词典的图书推荐方法  

Book Recommendation Method Based on Sentiment Analysis and Concept Dictionary

作　　者：杨群峰 王忠群 皇苏斌 YANG Qunfeng;WANG Zhongqun;HUANG Subin(School of Computer and Information,Anhui Polytechnic University,Wuhu 241000,China)

机构地区：[1]安徽工程大学计算机与信息学院,安徽芜湖241000

出　　处：《安徽工程大学学报》2022年第5期59-65,共7页Journal of Anhui Polytechnic University

基　　金：国家自然科学基金资助项目(71371012);教育部人文社会科学研究规划基金资助项目(18YJA630114)。

摘　　要：传统基于协同过滤的图书推荐算法根据读者的历史信息进行推荐,结果具有不可解释性且未考虑读者对图书内容的认知。针对此问题,本文提出一种基于情感分析和概念词典的图书推荐方法。首先,基于读者所需的概念词,通过Word2vec寻找概念内涵词汇,初步构建概念词典,再利用概念外延对词典进行扩展并赋予权值,确保推荐的准确性;然后,利用BEET方法对当当网图书评论进行情感分类,使推荐结果与读者所需的图书内容相关;最后整体分析得出图书概念系数以实施推荐并评估推荐结果。结果表明,本研究能合理解释推荐结果,且推荐准确性较高,具有较好的可行性。The traditional collaborative filtering-based book recommendation algorithm makes recommendations based on readers'historical information,and the results are uninterpretable and do not take into account readers'cognition of the book content.To solve this problem,a book recommendation method based on sentiment analysis and concept dictionary is proposed.Firstly,based on the reader's desired concept words,the concept dictionary is initially constructed by finding concept connotation words through word vectors(Word2vec),and then the dictionary is extended and given weights using concept extents to ensure the accuracy of the recommendation.Then,this method uses bidirectional encoder representation from transformers(BERT)to classify the sentiment of Dangdang.com book reviews,so that the recommendation results are related to the readers'desired book contents.Finally,the overall analysis brings about a book concept factor to implement the recommendation and evaluate the recommendation results.The results show that this method can reasonably explain the recommendation results and the recommendation accuracy is high,which has good feasibility.

关 键 词：图书推荐 情感分析 概念词典 词向量 

分 类 号：G252.62[文化科学—图书馆学]