Explainable Product Search with a Dynamic Relation Embedding Model

Abstract

USA Product search is one of the most popular methods for customers to discover products online. Most existing studies on product search focus on developing effective retrieval models that rank items by their likelihood to be purchased. They, however, ignore the problem that there is a gap between how systems and customers perceive the relevance of items. Without explanations, users may not understand why product search engines retrieve certain items for them, which consequentially leads to imperfect user experience and suboptimal system performance in practice. In this work, we tackle this problem by constructing explainable retrieval models for product search. Specifically, we propose to model the "search and purchase" behavior as a dynamic relation between users and items, and create a dynamic knowledge graph based on both the multi-relational product data and the context of the search session. Ranking is conducted based on the relationship between users and items in the latent space, and explanations are generated with logic inferences and entity soft matching on the knowledge graph. Empirical experiments show that our model, which we refer to as the Dynamic Relation Embedding Model (DREM), significantly outperforms the state-of-the-art baselines and has the ability to produce reasonable explanations for search results.

Explainable Recommendation: A Survey and New Perspectives

“…Besides, explanations can be provided by finding the shortest path from the user to the recommended item through the KG. By incorporating explicit user queries, the model can be further extended to conduct explainable search over knowledge graphs (Ai et al, 2019). Wang et al (2018a) proposed the Ripple Network, an end-to-end framework to incorporate KG into recommender systems.…”

“…Later, many explainable recommendation models are proposed for e-commerce recommendation. For instance, He et al (2015) introduced a tripartite graph ranking algorithm for explainable recommendation of electronics products; Chen et al (2016) proposed a learning to rank approach to cross-category explainable recommendation of the products; Seo et al (2017) and Wu et al (2019) conducted explainable recommendation for multiple product categories in Amazon, and highlighted important words in user reviews based on attention mechanism; Heckel et al (2017) adopted overlapping co-clustering to provide scalable and interpretable product recommendations; Chen et al (2019b) proposed a visually explainable recommendation model to provide visual explanations for fashion products; Hou et al (2018) used product aspects to conduct explainable video game recommendation in Amazon; Chen et al (2018a) leveraged neural attention regression based on reviews to conduct rating prediction on three Amazon product categories; Chen et al (2018c) adopted memory networks to provide explainable sequential recommendations in Amazon; Wang et al (2018b) leveraged multi-task learning with tensor factorization to learn textual explanations for Amazon product recommendation; By incorporating explicit queries, explainable recommendation can also be extended to explainable product search in e-commerce systems (Ai et al, 2019).…”

Explaining Documents' Relevance to Search Queries

“…Explainable recommendation has recently attracted considerable attention [84]. For instance, Ai et al [2] recently proposed a model based on dynamic relation embedding to produce explanation for recommendation in the context of e-commerce. Content-based models for explanation of recommender systems are closer to explanation of search results than those of collaborative filtering, yet structures in items and user-item interactions are not available in Web search.…”

“…Our model is thus trained with no manually labeled training data. To build weakly-labeled training data with a noise level that a supervised encoder-decoder model can tolerate and that learns a model that generalizes to open-domain input texts, we combine samples from two sources: (1) Wikipedia articles as more controlled edited content than the entire Web, and (2) anchor texts in a collection of general web pages.…”

Explainable recommendation based on knowledge graph and multi-objective optimization

“…The application of model-based explainable recommendation system is more extensive [21]. In the field of model-based explainable recommendation systems, there are many methods for modeling, including matrix factorization [38], factorization machines [3], deep learning [12], knowledge graph [2], etc.…”