Explainable Recommendation via Multi-Task Learning in Opinionated Text Data

By Nan Wang, Hongning Wang, Yiling Jia, Yue Yin. 2018

237 citations recorded by the source corpus.

Explaining automatically generated recommendations allows users to make more informed and accurate decisions about which results to utilize, and therefore improves their satisfaction. In this work, we develop a multi-task learning solution for explainable recommendation. Two companion learning tasks of user preference modeling for recommendation and opinionated content modeling for explanation are integrated via a joint tensor factorization. As a result, the algorithm predicts not only a user's preference over a list of items, i.e., recommendation, but also how the user would appreciate a particular item at the feature level, i.e., opinionated textual explanation. Extensive experiments on two large collections of Amazon and Yelp reviews confirmed the effectiveness of our solution in both recommendation and explanation tasks, compared with several existing recommendation algorithms. And our extensive user study clearly demonstrates the practical value of the explainable recommendations generated by our algorithm.

Subfields: Computational Mathematics, Information Systems, Management Science and Operations Research

How researchers see this paper

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

In Computational Mathematics

In Information Systems

In Management Science and Operations Research

Citations in the corpus

Explainable Recommendation: A Survey and New Perspectives (2018, methodology): “Though the term explainable recommendation was formally introduced in the recent years (Zhang et al. [2014a]), the basic concept, however, dated back to some of the most early works in personalized recommendation research.”
Incorporating Interpretability into Latent Factor Models via Fast Influence Analysis (2018, background): “The first category is to leverage external data sources to express the semantics of latent dimensions, and provide feature-based explanations for the recommendation results [9, 11, 36, 39].”
Dynamic Explainable Recommendation Based on Neural Attentive Models (2019, methodology): “LRPPM (Chen et al. 2016b) and MTER (Wang et al. 2018) extended EFM for more accurate user-item-feature explanations based on tensor factorization techniques.”
Explainable Product Search with a Dynamic Relation Embedding Model (2019, methodology): “Specifically, we follow the methodology proposed by Wang et al. [60] and create three questions for the annotations of search result explanations in our laboratory study: • Q1: Item Information .”
A Capsule Network for Recommendation and Explaining What You Like and Dislike (2019, methodology): “ries. The first category resorts to using the external NLP tools to extract aspects and sentiments from reviews. For example, EFM and MTER generate a sentiment lexicon through a phrase-level NLP tool [31, 38]. Similarly, TriRank utilizes the extracted aspects to construct a tripartite graph over use”
Recommending what video to watch next: a multitask ranking system (2019, background): “On the other hand, many existing multi-objective ranking systems are designed for specifc types of features and applications, such as text [38] and vision [13].”
Generate Natural Language Explanations for Recommendation (2021, methodology): “Template-based models, such as [8, 38, 45], dene one or more explanation sentence templates, and then ll dierent words into the templates according to the corresponding recommendation so as to generate dierent explanations.”
Transparent, Scrutable and Explainable User Models for Personalized Recommendation (2019, background): “Note that this work, along with many others [9, 11, 27, 30, 43, 52], generates explanations with the help of reviews written for items, which may not always be available.”
Progressive Layered Extraction (PLE): A Novel Multi-Task Learning (MTL) Model for Personalized Recommendations (2020, background): “[23] impose regularization on latent representations learned for the recommendation task and explanation task to optimize them jointly.”
Explainable Outfit Recommendation with Joint Outfit Matching and Comment Generation (2018, background): “[19] develop a multi-task learning solution which uses a joint tensor factorization to model user preference for recommendation and opinionated content for explanation; the algorithm can generate explanations by projecting the features and opinionated phrases onto the space jointly spanned by the us”
The FacT: Taming Latent Factor Models for Explainability with Factorization Trees (2019, methodology): “Following [34], we use the relative normalized weight of BPR to quantify the influence of BPR pairs in Session 3B: Interpretatibility and Explainability SIGIR ’19, July 21–25, 2019, Paris, France”