Explainable Recommendation via Interpretable Feature Mapping and Evaluation of Explainability
This work addresses the problem of balancing explainability and accuracy in recommender systems for users and developers, offering a novel method that eliminates the need for metadata, though it builds incrementally on existing latent factor collaborative filtering techniques.
The paper tackles the trade-off between explainability and performance in recommender systems by proposing a feature mapping approach that transforms uninterpretable general features into interpretable aspect features, achieving both satisfactory accuracy and explainability without needing metadata. Experimental results demonstrate strong performance in both recommendation and explanation, with specific metrics showing improvements over baseline methods.
Latent factor collaborative filtering (CF) has been a widely used technique for recommender system by learning the semantic representations of users and items. Recently, explainable recommendation has attracted much attention from research community. However, trade-off exists between explainability and performance of the recommendation where metadata is often needed to alleviate the dilemma. We present a novel feature mapping approach that maps the uninterpretable general features onto the interpretable aspect features, achieving both satisfactory accuracy and explainability in the recommendations by simultaneous minimization of rating prediction loss and interpretation loss. To evaluate the explainability, we propose two new evaluation metrics specifically designed for aspect-level explanation using surrogate ground truth. Experimental results demonstrate a strong performance in both recommendation and explaining explanation, eliminating the need for metadata. Code is available from https://github.com/pd90506/AMCF.