针对现有基于用户正负偏好的电影推荐方法未充分考虑用户情感信息、用户正负偏好表示不够准确以及推荐效果不够理想的问题,提出一种利用评论中蕴含的用户情感改进其正负偏好,基于改进偏好实现电影推荐的方法.首先,构建电影领域情感词典,利用该词典实现对评论数据的挖掘与量化,得到用户评论的情感得分;然后,综合考虑用户的评分和评论情感得分计算用户对电影的喜好度,从而生成用户的正、负向偏好电影集合,并基于这2个集合构建用户的正、负向偏好特征向量,挖掘得到蕴含评论情感的用户正负偏好;最后,利用候选电影与用户正、负向偏好特征的相似度计算用户对候选电影的最终评分,实现电影推荐.基于豆瓣数据集的实验结果表明,本文方法的各项指标相对于传统方法有一定提升,其中F1、MAE和MAPE指标分别提升6.10%、3.32%、11.67%.
In the existing movie recommendation methods based on users positive and negative preferences. The users sentiment information was not fully considered, the users positive and negative preferences was not accurate enough, and the recommendation effect was not ideal. In this paper, a method is proposed to improve users positive and negative preferences by using users emotions contained in reviews, and the movie recommendation is realized based on the new preferences. Firstly, the sentiment dictionary in the field of movie was constructed, which is used to realize the mining and quantification of the review data and obtain the sentiment score of user reviews. Secondly, the users preference degree for movies was calculated in order to generate the users positive and negative preference movie sets by considering the users rating and the comment sentiment score comprehensively, then according to the two sets, the users positive and negative preference feature vectors were constructed, and the users positive and negative preferences containing user sentiment were mined. Finally, the similarity, which is between the candidate movie and the users positive and negative preference features, was used to calculate the users final rating for the candidate movie, and the movie recommendation was realized. Compared with traditional - DOI:10.3969/j.issn.1000-1565.2023.06.012基于蕴含情感要素用户正负偏好的电影推荐方法张彬,董雅倩,徐建民(河北大学 网络空间安全与计算机学院,河北 保定 071000)摘 要:针对现有基于用户正负偏好的电影推荐方法未充分考虑用户情感信息、用户正负偏好表示不够准确以及推荐效果不够理想的问题,提出一种利用评论中蕴含的用户情感改进其正负偏好,基于改进偏好实现电影推荐的方法.首先,构建电影领域情感词典,利用该词典实现对评论数据的挖掘与量化,得到用户评论的情感得分;然后,综合考虑用户的评分和评论情感得分计算用户对电影的喜好度,从而生成用户的正、负向偏好电影集合,并基于这2个集合构建用户的正、负向偏好特征向量,挖掘得到蕴含评论情感的用户正负偏好;最后,利用候选电影与用户正、负向偏好特征的相似度计算用户对候选电影的最终评分,实现电影推荐.基于豆瓣数据集的实验结果表明,本文方法的各项指标相对于传统方法有一定提升,其中F1、MAE和MAPE指标分别提升6.10%、3.32%、11.67%.关键词:电影推荐;正向偏好;负向偏好;评论情感中图分类号:TP391 文献标志码:A 文章编号:1000-1565(2023)06-0653-12A movie recommendation method based on the positive and negative preferences of users with emotional factorsZHANG Bin, DONG Yaqian, XU Jianmin(School of Cyber Security and Computer, Hebei University, Baoding 071000, China)Abstract: In the existing movie recommendation methods based on users positive and negative preferences. The users sentiment information was not fully considered, the users positive and negative preferences was not accurate enough, and the recommendation effect was not ideal. In this paper, a method is proposed to improve users positive and negative preferences by using users emotions contained in reviews, and the movie recommendation is realized based on the new preferences. Firstly, the sentiment dictionary in the field of movie was constructed, which is used to realize the mining and quantification of the review data and obtain the sentiment score of user reviews. Secondly, the users preference degree for movies was calculated in order to generate the users positive and negative preference movie sets by considering the users rating and the comment sentiment score comprehensively, then according to the two sets, the users positive and negative preference feature vectors were constructed, and the users positive and negative preferences containing user sentiment were mined. Finally, the similarity, which is between the candidate movie and the users positive and negative preference features, was used to calculate the users final rating for the candidate movie, and the movie recommendation was realized. Compared with traditional - 收稿日期:2023-02-23 基金项目:河北省社会科学基金资助项目(HB21TQ005) 第一作者:张彬(1980—),男,河北涿州人,河北大学高级实验师,博士,主要从事信息检索方向研究.E-mail:zb@hbu.edu.cn 通信作者:徐建民(1966—),男,河北馆陶人,河北大学教授,博士生导师,主要从事信息检索方向研究.E-mail:hbuxjm@hbu.edu.cn第6期张彬等:基于蕴含情感要素用户正负偏好的电影推荐方法methods, the experimental results based on Douban dataset show that the proposed method has certain improvements over the traditional methods, and the indexes of F1, MAE and MAPE are improved by 6.10%, 3.32% and 11.67%, respectively.
[1] LIU Y, MIYAZAKI J. Knowledge-aware attentional neural network for review-based movie recommendation with explanations[J]. Neural Comput & Applic, 2022, 35(3): 2717-2735. DOI: 10.1007/s00521-022-07689-1.
[2] 潘华莉,谢珺,高婧,等.融合多模态特征的深度强化学习推荐模型[J/OL].数据分析与知识发现,2022:1-18. [2023-03-20]. http://kns.cnki.net/kcms/detail/10.1478.G2.20220907.1507.008.html.
[3] 张建伟,李月琳,李东东.网络学术资源平台个性化推荐服务特征研究[J].情报资料工作,2021,42(5):76-83. DOI: 10.12154/j.qbzlgz.2021.05.009
[4] 冉从敬,宋凯.基于混合方法的高校专利个性化推荐模型构建[J].情报理论与实践,2020,43(10):93-98. DOI: 10.16353/j.cnki.1000-7490.2020.10.017.
[5] BOBADILLA J, ORTEGA F, HERNANDO A, et al. Recommender systems survey[J]. Knowledge-Based Systems, 2013, 46: 109-132. DOI: 10.1016/j.knosys.2013.03.012.
[6] 潘涛涛,文锋,刘勤让.基于矩阵填充和物品可预测性的协同过滤算法[J].自动化学报,2017,43(9):1597-1606. DOI: 10.16383/j.aas.2017.c160644.
[7] 马鑫,王芳.融合类目偏好和数据场聚类的协同过滤推荐算法研究[J].现代情报,2023 43(1):6-18.
[8] 王洪伟,郑丽娟,尹裴,等.在线评论的情感极性分类研究综述[J].情报科学,2012,30(8):1263-1271,1276. DOI: 10.13833/j.cnki.is.2012.08.008.
[9] 庞庆华,董显蔚,周斌,等.基于情感分析与TextRank的负面在线评论关键词抽取[J].情报科学,2022,40(5):111-117. DOI: 10.13833/j.issn.1007-7634.2022.05.015.
[10] 徐怡,唐一民,王冉.基于正相关和负相关最近邻居的协同过滤算法[J].工程科学与技术,2018,50(5):189-195. DOI: 10.15961/j.jsuese.201700764.
[11] 张宇,王文剑,赵胜男.基于正负反馈的SVM协同过滤Top-N推荐算法[J].小型微型计算机系统,2017,38(5):961-966.
[12] WALEK B, FOJTIK V. A hybrid recommender system for recommending relevant movies using an expert system[J]. Expert Systems with Applications, 2020, 158, 113452: 1-18. DOI: 10.1016/j.eswa.2020.113452.
[13] 丁来旭,刘洪娟.复杂网络上基于多维特征表示学习的推荐算法[J].东北大学学报(自然科学版),2022,43(3):359-367. DOI: 10.12068/j.issn.1005-3026.2022.03.008.
[14] CHEN Y L, YEH Y H, MA M R. A movie recommendation method based on users positive and negative profiles[J]. Inf Process Manag, 2021, 58(3), 102531: 1-13. DOI: 10.1016/j.ipm.2021.102531.
[15] ZENG Z, SHI Y, PIEPTEA L F, et al. Using latent features for building an interpretable recommendation system[J]. The Electronic Library, 2021, 39(2): 281-295. DOI:10.1108/EL-06-2020-0154.
[16] 李浩君,吕韵,汪旭辉,等.融入情感分析的多层交互深度推荐模型研究[J].数据分析与知识发现,2023,7(3):43-57.
[17] 蒋翠清,郭轶博,刘尧.基于中文社交媒体文本的领域情感词典构建方法研究[J].数据分析与知识发现,2019,3(2):98-107. DOI: 10.11925/infotech.2096-3467.2018.0578.
[18] 徐琳宏,林鸿飞,潘宇,等.情感词汇本体的构造[J].情报学报,2008,27(2):180-185. DOI: 10.3969/j.issn.1000-0135.2008.02.004.
[19] ZHU Y L, MIN J, ZHOU Y Q, et al. Semantic orientation computing based on HowNet[J]. Journal of Chinese Information Processing, 2006,20(1):14-20.
[20] KU L W, CHEN H H. Mining opinions from the web: Beyond relevance retrieval[J]. Journal of the American Society for Information Science and Technology, 2010, 58(12): 1838-1850. DOI: 10.1002/asi.20630.
[21] 夏建勋,吴非,谢长生.应用数据填充缓解稀疏问题实现个性化推荐[J].计算机工程与科学,2013,35(5):15-19. DOI: 10.3969/j.issn.1007-130X.2013.05.003.
[22] ABDELWAHAB A, SEKIYA H, MATSUBA I, et al. Collaborative filtering based on an iterative prediction method to alleviate the sparsity problem[C] //Proceedings of the 11th International Conference on Information Integration and Web-based Applications & Services, December 14-16, 2009, Kuala Lumpur, Malaysia, New York: ACM, 2009: 375-379. DOI: 10.1145/1806338.1806406.
[23] SON J, KIM S B. Content-based filtering for recommendation systems using multiattribute networks[J]. Expert Syst Appl, 2017, 89: 404-412. DOI: 10.1016/j.eswa.2017.08.008.
[24] 姜霖,张麒麟.基于评论情感分析的个性化推荐策略研究:以豆瓣影评为例[J].情报理论与实践,2017,40(8):99-104.DOI: 10.16353/j.cnki.1000-7490.2017.08.018.
[25] 许智宏,于子琪,董永峰,等.影评情感分析知识图谱构建研究[J].计算机仿真,2020,37(8):424-430.
[26] 顾亦然,张远之,杨海根.基于电影属性和交互信息的电影推荐算法[J].南京理工大学学报,2022,46(2):177-184. DOI: 10.14177/j.cnki.32-1397n.2022.46.02.008.
[27] 曹宗胜,许倩倩,李朝鹏,等.基于对偶四元数的协同知识图谱推荐模型[J].计算机学报,2022,45(10):2221-2242. DOI: 10.11897/SP.J.1016.2022.02221.
[28] 梁浩宏,古天龙,宾辰忠,等.联合学习用户端和项目端知识图谱的个性化推荐[J].计算机科学,2021,48(5):109-116. DOI: 10.11896/jsjkx.200600115.
[29] 袁泉,成振华,江洋.基于知识图谱和协同过滤的电影推荐算法研究[J].计算机工程与科学,2020,42(4):714-721. DOI: 10.3969/j.issn.1007-130X.2020.04.019.
[30] ZHEN W, ZHANG J, FENG J, et al. Knowledge graph embedding by translating on hyperplanes[C]. National Conference on Artificial Intelligence AAAI Press, 2014: 1112-1119.
[31] 牛秀娟.基于用户情感分析和信任关系的推荐算法研究[D].西安:西北大学,2020.
[32] 金丹,张娇娇,李依玲,等.一种改进的协同过滤算法研究:以电影推荐系统为例[J].国际商务,2020(1):128-141. DOI: 10.13509/j.cnki.ib.2020.01.010. (
