SearchGCN: Powering Embedding Retrieval by Graph Convolution Networks for E-Commerce Search

TL;DR

SearchGCN leverages graph convolution networks to improve embedding-based candidate retrieval in large-scale e-commerce search, especially enhancing long tail query and item performance, and has been successfully deployed in JD.com's search engine.

Contribution

This paper introduces SearchGCN, a novel application of GCNs for industrial-scale search engine retrieval, demonstrating improved embeddings over existing methods.

Findings

Better embedding representations for long tail queries and items

Successful deployment in JD.com's search engine since July 2020

Empirical evidence of improved retrieval performance

Abstract

Graph convolution networks (GCN), which recently becomes new state-of-the-art method for graph node classification, recommendation and other applications, has not been successfully applied to industrial-scale search engine yet. In this proposal, we introduce our approach, namely SearchGCN, for embedding-based candidate retrieval in one of the largest e-commerce search engine in the world. Empirical studies demonstrate that SearchGCN learns better embedding representations than existing methods, especially for long tail queries and items. Thus, SearchGCN has been deployed into JD.com's search production since July 2020.