Combining Deep Learning and Topic Modeling for Review Understanding in Context-Aware Recommendation

Proceedings of NAACL-HLT 2018, pages 1605–1614 New Orleans, Louisiana, June 1 - 6, 2018. c⃝2018 Association for Computational Linguistics Combining Deep Learning and Topic Modeling for Review Understanding in Context-Aware Recommendation Mingmin Jin1 Xin Luo1 Huiling Zhu1,2 Hankz Hankui Zhuo1,2∗ 1School of Data and Computer Science, Sun Yat-Sen University 2Guangdong Key Laboratory of Big Data Analysis and Processing Guangzhou, China {jinmm1,luox351,zhuhling62,zhuohank2}@{mail21,mail2}.sysu.edu.cn Abstract With the rise of e-commerce, people are ac- customed to writing their reviews after receiv- ing the goods. These comments are so impor- tant that a bad review can have a direct impact on others buying. Besides, the abundant in- formation within user reviews is very useful for extracting user preferences and item prop- erties. In this paper, we investigate the ap- proach to effectively utilize review informa- tion for recommender systems. The proposed model is named LSTM-Topic matrix factor- ization (LTMF) which integrates both LSTM and Topic Modeling for review understand- ing. In the experiments on popular review dataset Amazon , our LTMF model outper- forms previous proposed HFT model and Con- vMF model in rating prediction. Furthermore, LTMF shows the better ability on making topic clustering than traditional topic model based method, which implies integrating the infor- mation from deep learning and topic modeling is a meaningful approach to make a better un- derstanding of reviews. 1 Introduction Recommender systems (RSs) are widely used in the field of electronic commerce to provide per- sonalized recommendation services for customers. Most popular RSs are based on Collaborative Fil- tering (CF), which makes use of users’ explicit ratings or implicit behaviour for recommendations (Koren, 2008). But CF models suffer from data sparsity, which is also called “cold-start” prob- lem. Models perform poorly when there is few available data. To alleviate this problem, utiliz- ing user reviews can be a good approach because user reviews can directly reflect users’ preferences and items’ properties and exactly correspond to the user latent factors and item latent factors in CF models. ∗corresponding author To understand user reviews, previous ap- proaches are mainly based on topic modeling, a suite of algorithms that aim to discover the the- matic information among documents (Blei, 2012). The simplest and commonly used topic model is latent dirichlet allocation(LDA). Recently, as deep learning shows great performance in com- puter vision (Krizhevsky et al., 2017) and NLP (Kim, 2014), some approaches combining deep learning with CF are proposed to capture latent context features from reviews. However, we find there are some limitations in existing models. First, the LDA algorithm used in previous models like Hidden Factors as Top- ics (HFT) (McAuley and Leskovec, 2013) ignores contextual information. If a user writes “I pre- fer apple than banana when choosing fruits” in a review, we can clearly know the user’s prefer- ence and recommend items including apple. But LDA ignores the structural information and con- siders the two words as the same since they both appear once in the sentence. Compared with topic modeling, deep learning methods such as Convolutional Neural …

Content truncated to 500 words