Que2Search: Fast and Accurate Query and Document Understanding for Search at Facebook
Authors:
Open Access
- 14 August 2021
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 3376-3384
Abstract
In this paper, we present Que2Search, a deployed query and product understanding system for search. Que2Search leverages multi-task and multi-modal learning approaches to train query and product representations. We achieve over 5% absolute offline relevance improvement and over 4% online engagement gain over state-of-the-art Facebook product understanding system by combining the latest multilingual natural language understanding architectures like XLM and XLM-R with multi-modal fusion techniques. In this paper, we describe how we deploy XLM-based search query understanding model that runs <1.5ms @P99 on CPU at Facebook scale, which has been a significant challenge in the industry. We also describe what model optimizations worked (and what did not) based on numerous offline and online A/B experiments. We deploy Que2Search to Facebook Marketplace Search and share our deployment experience to production and tuning tricks to achieve higher efficiency in online A/B experiments. Que2Search has demonstrated gains in production applications and operates at Facebook scale.Keywords
This publication has 10 references indexed in Scilit:
- DeText2020
- UNITER: UNiversal Image-TExt Representation Learning2020
- GrokNet2020
- Embedding-based Retrieval in Facebook Search2020
- Improving Ad Click Prediction by Considering Non-displayed Events2019
- ArcFace: Additive Angular Margin Loss for Deep Face Recognition2019
- Exploring the Limits of Weakly Supervised Pretraining2018
- Wide & Deep Learning for Recommender Systems2016
- UnicornProceedings of the VLDB Endowment, 2013
- SIGNATURE VERIFICATION USING A “SIAMESE” TIME DELAY NEURAL NETWORK1994