Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Semantic-Enhanced Relational Metric Learning for Recommender Systems

Mingming Li, Fuqing Zhu, Feng Yuan, Songlin Hu

TL;DR

SERML addresses the lack of semantic information in relational metric learning for recommender systems by injecting semantic cues derived from target reviews into the relation induction process. The framework combines three components—textual representation learning via HLSTM with attention, memory-based relation induction, and relational metric learning—into an end-to-end model with a regression-guided supervision signal. The model optimizes a joint objective that blends sentiment-based text supervision, relational regression, and margin-based ranking, achieving superior item ranking and rating prediction across four public datasets. By leveraging semantic signals, SERML reduces overfitting due to co-occurrence and improves discrimination between items with similar histories, enhancing practical recommendation quality.

Abstract

Recently, relational metric learning methods have been received great attention in recommendation community, which is inspired by the translation mechanism in knowledge graph. Different from the knowledge graph where the entity-to-entity relations are given in advance, historical interactions lack explicit relations between users and items in recommender systems. Currently, many researchers have succeeded in constructing the implicit relations to remit this issue. However, in previous work, the learning process of the induction function only depends on a single source of data (i.e., user-item interaction) in a supervised manner, resulting in the co-occurrence relation that is free of any semantic information. In this paper, to tackle the above problem in recommender systems, we propose a joint Semantic-Enhanced Relational Metric Learning (SERML) framework that incorporates the semantic information. Specifically, the semantic signal is first extracted from the target reviews containing abundant item features and personalized user preferences. A novel regression model is then designed via leveraging the extracted semantic signal to improve the discriminative ability of original relation-based training process. On four widely-used public datasets, experimental results demonstrate that SERML produces a competitive performance compared with several state-of-the-art methods in recommender systems.

Semantic-Enhanced Relational Metric Learning for Recommender Systems

TL;DR

SERML addresses the lack of semantic information in relational metric learning for recommender systems by injecting semantic cues derived from target reviews into the relation induction process. The framework combines three components—textual representation learning via HLSTM with attention, memory-based relation induction, and relational metric learning—into an end-to-end model with a regression-guided supervision signal. The model optimizes a joint objective that blends sentiment-based text supervision, relational regression, and margin-based ranking, achieving superior item ranking and rating prediction across four public datasets. By leveraging semantic signals, SERML reduces overfitting due to co-occurrence and improves discrimination between items with similar histories, enhancing practical recommendation quality.

Abstract

Recently, relational metric learning methods have been received great attention in recommendation community, which is inspired by the translation mechanism in knowledge graph. Different from the knowledge graph where the entity-to-entity relations are given in advance, historical interactions lack explicit relations between users and items in recommender systems. Currently, many researchers have succeeded in constructing the implicit relations to remit this issue. However, in previous work, the learning process of the induction function only depends on a single source of data (i.e., user-item interaction) in a supervised manner, resulting in the co-occurrence relation that is free of any semantic information. In this paper, to tackle the above problem in recommender systems, we propose a joint Semantic-Enhanced Relational Metric Learning (SERML) framework that incorporates the semantic information. Specifically, the semantic signal is first extracted from the target reviews containing abundant item features and personalized user preferences. A novel regression model is then designed via leveraging the extracted semantic signal to improve the discriminative ability of original relation-based training process. On four widely-used public datasets, experimental results demonstrate that SERML produces a competitive performance compared with several state-of-the-art methods in recommender systems.
Paper Structure (30 sections, 15 equations, 4 figures, 6 tables)

This paper contains 30 sections, 15 equations, 4 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Problem Formulation
  5. Framework Overview
  6. Textual Representation Learning
  7. Relation Induction
  8. Memory-based Network.
  9. Relation Regression.
  10. Relational Metric Learning
  11. Joint Learning
  12. Prediction
  13. Experiments
  14. Item Ranking
  15. Datasets
  16. ...and 15 more sections

Figures (4)

  • Figure 1: An illustrative example (left: LRML, right: the ideal case). For example, $v_{1}$ and $v_{3}$ are similar items; $v_{2}$ and $v_{4}$ are similar items.
  • Figure 2: Framework illustration of SERML.
  • Figure 3: Experimental results of different induced relations: $MLP^{2}$, $MLP^{4}$, Element-wise, and Memory-based.
  • Figure 4: The effect of dimension on Instant Video and Yelp13.