基于知识图谱的可解释性推荐系统(Explainable reasoning over knowledge graphs for recommendation)

此演讲使用中文 (This will be presented in Chinese)

Dingxian Wang (eBay)
14:0014:40 Friday, June 21, 2019
模型与方法 (Models and Methods)
Location: 多功能厅6A+B (Function Room 6A+B)

必要预备知识 (Prerequisite Knowledge)

  • Familiarity with knowledge graphs (KG) and embedding

您将学到什么 (What you'll learn)

  • Gain an understanding of knowledge graph reasoning and embedding

描述 (Description)

In recent years, there’s been increasing attention on incorporating knowledge graphs into recommender systems. By exploring the interlinks within a knowledge graph, you can discover the connectivity between users and items as paths, which provides rich and complementary information to user-item interactions. Such connectivity not only reveals the semantics of entities and relations but also helps to comprehend a user’s interest. However, existing efforts have not fully explored this connectivity to infer user preferences, especially in terms of modeling the sequential dependencies within and holistic semantics of a path.

Drawing on research undertaken for the paper “Explainable Reasoning over Knowledge Graphs for Recommendation,” Dingxian Wang outlines a new model, knowledge-aware path recurrent network (KPRN), for exploiting knowledge graphs for recommendation. KPRN can generate path representations by composing the semantics of both entities and relations. By leveraging the sequential dependencies within a path, effective reasoning on paths infers the underlying rationale of a user-item interaction. Dingxian also discusses the design for a new weighted pooling operation to discriminate the strengths of different paths in connecting a user with an item, endowing the model with a certain level of explainability.

Dingxian shares lessons learned from extensive experiments on two datasets about movies and music, demonstrating significant improvements over the state-of-the-art solutions, collaborative knowledge base embedding, and neural factorization machine. For example, a user may be connected to the song “I See Fire” since she likes “Shape of You,” sung by the same artist, Ed Sheeran. Such connectivity helps to reason about unseen user-item interactions (i.e., a potential recommendation) by synthesizing information from paths.

Running example: {Alice, Interact, Shape of You}^{Shape of You, SungBy, Ed Sheeran}^{Ed Sheeran, IsSingerOf, I See Fire}=>{Alice, Interact, I See Fire}.

The reasoning unveils the possible user intents behind an interaction, offering explanations behind a recommendation. How to model such connectivity in KGs, hence, is of critical importance to inject knowledge into a recommender systems. KPRN generates representations for paths by accounting for both entities and relations, and also performs reasoning based on paths to infer user preference. Specifically, Dingxian extracts qualified paths between a user-item pair from the KG, each of which consists of the related entities and relations, and adopts a long short-term memory (LSTM) network to model the sequential dependencies of entities and relations. A pooling operation aggregates the representations of paths to obtain prediction signals for the user-item pair. The pooling operation is capable of discriminating the contributions of different paths for a prediction, which functions as the attention mechanism. The model can propose path-wise explanations, such as “Castle on the Hill” recommended because the user listened to “Shape of You” sung and written by Ed Sheeran.

Dingxian Wang


Dingxian Wang is an applied researcher at eBay.

Leave a Comment or Question

Help us make this conference the best it can be for you. Have questions you'd like this speaker to address? Suggestions for issues that deserve extra attention? Feedback that you'd like to share with the speaker and other attendees?

Join the conversation here (requires login)