Personalized Neural Embeddings for Collaborative Filtering with Text

Proceedings of NAACL-HLT 2019, pages 2082–2088 Minneapolis, Minnesota, June 2 - June 7, 2019. c⃝2019 Association for Computational Linguistics 2082 Personalized Neural Embeddings for Collaborative Filtering with Text Guangneng Hu Department of Computer Science and Engineering Hong Kong University of Science and Technology, Hong Kong, China njuhgn@gmail.com Abstract Collaborative filtering (CF) is a core tech- nique for recommender systems. Tradition- al CF approaches exploit user-item relations (e.g., clicks, likes, and views) only and hence they suffer from the data sparsity issue. Item- s are usually associated with unstructured text such as article abstracts and product reviews. We develop a Personalized Neural Embedding (PNE) framework to exploit both interaction- s and words seamlessly. We learn such em- beddings of users, items, and words jointly, and predict user preferences on items based on these learned representations. PNE estimates the probability that a user will like an item by two terms—behavior factors and semantic fac- tors. On two real-world datasets, PNE shows better performance than four state-of-the-art baselines in terms of three metrics. We also show that PNE learns meaningful word embed- dings by visualization. 1 Introduction Recommender systems are widely used in e- commerce platforms, such as to help consumers buy products at Amazon, watch videos on Youtube, and read articles on Google News. They are useful to alleviate the information overload and improve user satisfaction. Given history records of con- sumers such as the product transactions and movie watching, collaborative filtering (CF) is among the most effective approaches based on the simple in- tuition that if users rated items similarly in the past then they are likely to rate items similarly in the future (Sarwar et al., 2001). History records include both implicit (e.g., pur- chase and clicks) and explicit (e.g., likes/dislikes and ratings) feedback which can be represented as a user-item interaction matrix. Typically, observed user-item interactions are incomplete with a large portion remaining not recorded. The goal of recom- mendation is to predict user preferences on these missing interactions. This setting requires to com- plete the partial observed rating matrix. Matrix Fac- torization (MF) techniques which can learn latent factors for users and items are the main cornerstone for CF (Mnih and Salakhutdinov, 2008; Koren, 2008; Koren et al., 2009). It is effective and flexible to integrate with additional data sources (Hu et al., 2015). Recently, neural networks like Multilayer Perceptron (MLP) are used to learn an interaction function from data with the power of learning high- ly nonlinear relationships between users and item- s (Dziugaite and Roy, 2015; Cheng et al., 2016; He et al., 2017; Hu et al., 2018b). MF and neural CF exploit user-item behavior interactions only and hence they both suffer from the data sparsity and cold-start issues. Items are usually associated with unstructured text, like news articles and product reviews. These additional sources are essential for recommenda- tion beyond user-item interactions since they con- tain independent and diverse information. Hence, they provide an opportunity to alleviate the da- ta sparsity issue (Ganu et al., …

Content truncated to 500 words