Knowledge Enhanced Multi-intent Transformer Network for Recommendation

Ding Zou,Wei Wei,Feida Zhu,Chuanyu Xu,Tao Zhang,Chengfu Huo

DOI: https://doi.org/10.1145/3589335.3648296

2024-05-31

Abstract:Incorporating Knowledge Graphs into Recommendation has attracted growing attention in industry, due to the great potential of KG in providing abundant supplementary information and interpretability for the underlying models. However, simply integrating KG into recommendation usually brings in negative feedback in industry, due to the ignorance of the following two factors: i) users' multiple intents, which involve diverse nodes in KG. For example, in e-commerce scenarios, users may exhibit preferences for specific styles, brands, or colors. ii) knowledge noise, which is a prevalent issue in Knowledge Enhanced Recommendation (KGR) and even more severe in industry scenarios. The irrelevant knowledge properties of items may result in inferior model performance compared to approaches that do not incorporate knowledge. To tackle these challenges, we propose a novel approach named Knowledge Enhanced Multi-intent Transformer Network for Recommendation (KGTN), comprising two primary modules: Global Intents Modeling with Graph Transformer, and Knowledge Contrastive Denoising under Intents. Specifically, Global Intents with Graph Transformer focuses on capturing learnable user intents, by incorporating global signals from user-item-relation-entity interactions with a graph transformer, meanwhile learning intent-aware user/item representations. Knowledge Contrastive Denoising under Intents is dedicated to learning precise and robust representations. It leverages intent-aware representations to sample relevant knowledge, and proposes a local-global contrastive mechanism to enhance noise-irrelevant representation learning. Extensive experiments conducted on benchmark datasets show the superior performance of our proposed method over the state-of-the-arts. And online A/B testing results on Alibaba large-scale industrial recommendation platform also indicate the real-scenario effectiveness of KGTN.

Information Retrieval

What problem does this paper attempt to address?

This paper attempts to solve two main problems encountered when integrating Knowledge Graphs (KGs) into recommendation systems in large - scale industrial scenarios:

1. **Multiple intents of users**: Users may have a variety of different intents when interacting with items. For example, in e - commerce scenarios, users may show preferences for specific styles, brands or colors. These diverse intents are related to different nodes in the KG, but existing methods often overlook this, resulting in poor recommendation effects.

2. **Knowledge noise**: In Knowledge Enhanced Recommendation (KGR), knowledge noise is a common problem, especially more serious in industrial scenarios. Irrelevant knowledge attributes may cause the model performance to be lower than methods without using knowledge.

To address these problems, the paper proposes a new method - Knowledge Enhanced Multi - intent Transformer Network (KGTN), which contains two main modules:

- **Global Intents Modeling with Graph Transformer**: By combining the global signals of user - item - relation - entity, use a well - designed graph transformer to capture learnable user intents and generate intent - aware user / item representations.
- **Knowledge Contrastive Denoising under Intents**: Use intent - aware user / item representations to sample relevant knowledge, and propose a local - global contrast mechanism to enhance noise - independent representation learning.

Through these two modules, KGTN aims to learn user / item representations more finely and comprehensively, and effectively remove irrelevant knowledge information, thereby improving the performance of the recommendation system. Experimental results show that KGTN outperforms existing methods on three benchmark datasets, and also shows the effectiveness in actual scenarios on Alibaba's large - scale online recommendation platform.