Pre-training Tasks for User Intent Detection and Embedding Retrieval in E-commerce Search

BERT-style models pre-trained on the general corpus (e.g., Wikipedia) and fine-tuned on specific task corpus, have recently emerged as breakthrough techniques in many NLP tasks: question answering, text classification, sequence labeling and so on. However, this technique may not always work, especia...

Full description

Saved in:
Bibliographic Details
Published inarXiv.org
Main Authors Qiu, Yiming, Zhao, Chenyu, Zhang, Han, Zhuo, Jingwei, Li, Tianhao, Zhang, Xiaowei, Wang, Songlin, Xu, Sulong, Long, Bo, Wen-Yun, Yang
Format Paper
LanguageEnglish
Published Ithaca Cornell University Library, arXiv.org 22.08.2022
Subjects
Customization
Electronic commerce
Embedding
Encyclopedias
Retrieval
Searching
Training
Online AccessGet full text

Cover

More Information
Summary:BERT-style models pre-trained on the general corpus (e.g., Wikipedia) and fine-tuned on specific task corpus, have recently emerged as breakthrough techniques in many NLP tasks: question answering, text classification, sequence labeling and so on. However, this technique may not always work, especially for two scenarios: a corpus that contains very different text from the general corpus Wikipedia, or a task that learns embedding spacial distribution for a specific purpose (e.g., approximate nearest neighbor search). In this paper, to tackle the above two scenarios that we have encountered in an industrial e-commerce search system, we propose customized and novel pre-training tasks for two critical modules: user intent detection and semantic embedding retrieval. The customized pre-trained models after fine-tuning, being less than 10% of BERT-base's size in order to be feasible for cost-efficient CPU serving, significantly improve the other baseline models: 1) no pre-training model and 2) fine-tuned model from the official pre-trained BERT using general corpus, on both offline datasets and online system. We have open sourced our datasets for the sake of reproducibility and future works.
ISSN:2331-8422