Learning Distributed Representations from Reviews for Collaborative Filtering

By Amjad Almahairi, Kyle Kastner, Kyunghyun Cho, Aaron C. Courville. 2015

114 citations recorded by the source corpus.

Recent work has shown that collaborative filter-based recommender systems can be improved by incorporating side information, such as natural language reviews, as a way of regularizing the derived product representations. Motivated by the success of this approach, we introduce two different models of reviews and study their effect on collaborative filtering performance. While the previous state-of-the-art approach is based on a latent Dirichlet allocation (LDA) model of reviews, the models we explore are neural network based: a bag-of-words product-of-experts model and a recurrent neural network. We demonstrate that the increased flexibility offered by the product-of-experts model allowed it to achieve state-of-the-art performance on the Amazon review dataset, outperforming the LDA-based approach. However, interestingly, the greater modeling power offered by the recurrent neural network appears to undermine the model's ability to act as a regularizer of the product representations.

Subfields: Artificial Intelligence, Information Systems, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Information Systems

In Natural Language Processing

Citations in the corpus

Learning Distributed Representations of Sentences from Unlabelled Data (2016, methodology): “In the SkipThought, S(D)AE and NMT models, the cost is computed based on a non-linear decoding of the internal sentence representations, so, as also observed by (Almahairi et al., 2015), the informative geometry of the representation space may not be reflected in a simple cosine distance.”
Variational Autoencoders for Collaborative Filtering (2018, neutral): “Besides being applied in session-based sequential recommendation (see Section 2.1), various approaches [2, 28, 47, 48] have applied neural networks to incorporate side information into collaborative filtering models to better handle the cold-start problem.”
Ask the GRU: Multi-task Learning for Deep Text Recommendations (2016, methodology): “Existing hybrid methods have used unsupervised learning objectives on text content to regularize the parameters of the recommendation model [4, 23, 24].”
Joint Deep Modeling of Users and Items Using Reviews for Recommendation (2017, methodology): “To the best of our knowledge, the only work which has utilized deep learning techniques to use review text to improve recommendation is presented in [1].”
Explainable recommendation with fusion of aspect information (2018, methodology): “The existing methods for explainable recommendation [1, 21, 42] often suffer from two defects.”
Generate Natural Language Explanations for Recommendation (2021, background): “Since user textual reviews are informative and be er re ect user preferences, a lot of research explored the possibility of incorporating user reviews to improve the recommendation quality [3, 20, 22, 43, 49] and recommendation explainability [4, 6, 18, 29, 39, 47], which helps to enhance the e ecti”
Automatic Generation of Natural Language Explanations (2017, unknown): “The work of [2] is among the first where a recommender sys-tem is utilising the review text as side information to improve the performance of recommender system and the solutions are rooted in recurrent neural networks.”
Recurrent Recommender Networks (2017, methodology): “Most often, these reviews have been transformed into a bag-of-words-model and used as a sort of regularization for the rating predictions (McAuley & Leskovec, 2013; Diao et al., 2014; Almahairi et al., 2015; Wu et al., 2016b).”
Factorization Meets the Item Embedding: Regularizing Matrix Factorization with Item Co-occurrence (2016, background): “It is also standard to further regularize the factors with side information [1, 2, 7, 12], which comprises user, item, and/or user-item covariates (e.”
Joint Representation Learning for Top-N Recommendation with Heterogeneous Information Sources (2017, methodology): “One trend is to augment numerical ratings with textual reviews for recommendation [8], which includes topic modeling [3, 26, 27, 34], sentiment analysis [9, 13, 39, 42–44], and (deep) neural network [2, 36, 41, 45] approaches for review modeling.”
Neural Rating Regression with Abstractive Tips Generation for Recommendation (2017, background): “Recently, some approaches also consider text information in addition to the rating data [1, 21, 23, 26, 40, 49].”
Sequence-Aware Recommender Systems (2018, methodology): “ applied in particular in the domain of Natural Language Processing [82] and, in the context of recommender systems, as an alternative method for representing textual data in content-based approaches [7]. While the applications discussed in this section indicate that using such representations can b”