G-STO: Sequential Main Shopping Intention Detection via Graph-Regularized Stochastic Transformer

By Y. Zhuang et al

Read the original document by opening this link in a new tab.

Table of Contents

1. Introduction
2. Related Works
3. Method
3.1 Problem Definition
3.2 Stochastic Embedding Layers
3.3 Intention Relation Graph Regularizer

Summary

The document discusses the importance of detecting main shopping intentions in sequential recommendation systems. It introduces a novel approach, G-STO, that utilizes stochastic embeddings and a graph regularizer to model shopping intentions as Gaussian distributions and improve relevance between intentions. The method involves mean and covariance transformers to encode sequential information and Wasserstein distance for model training. The proposed G-STO outperforms existing baselines significantly in identifying main shopping intentions.