mottox/ebr-papers

Helpful papers on embedding-based retrieval (EBR)

EBR-papers

Introduction

Embedding-based retrieval (EBR) is commonly applied in search, recommendation, natural language processing and computer vision. This repository lists some representative papers on EBR.

Papers on EBR

personalization

• Embedding-based Product Retrieval in Taobao Search [pdf]
• Towards Personalized and Semantic Retrieval: An End-to-End Solution for E-commerce Search via Embedding Learning [pdf]

diversity

• Multi-Interest Network with Dynamic Routing for Recommendation at Tmall [pdf]
• Controllable Multi-Interest Framework for Recommendation [pdf]

multitask

• MOBIUS: Towards the Next Generation of Query-Ad Matching in Baidu’s Sponsored Search [pdf]

sampling&correction:

• Embedding-based Retrieval in Facebook Search [pdf]
• Sampling-Bias-Corrected Neural Modeling for Large Corpus Item Recommendations [pdf]
• Mixed Negative Sampling for Learning Two-tower Neural Networks in Recommendations [pdf]

hard negative mining:

• Embedding-based Retrieval in Facebook Search [pdf]
• Learning Dense Representations for Entity Retrieval [pdf]
• Dense Passage Retrieval for Open-Domain Question Answering [pdf]

long-tail item recommendation:

• Self-supervised Learning for Large-scale Item Recommendations [pdf]
• A Model of Two Tales: Dual Transfer Learning Framework for Improved Long-tail Item Recommendation [pdf]

Useful Code:

• Tensorflow Recommenders [pdf]