SS-BERT: A Semantic Information Selecting Approach for Open-Domain Question Answering

Fu, Xuan; Du, Jiangnan; Zheng, Hai-Tao; Li, Jianfeng; Hou, Cuiqin; Zhou, Qiyu; Kim, Hong-Gee

doi:10.3390/electronics12071692

Open AccessArticle

SS-BERT: A Semantic Information Selecting Approach for Open-Domain Question Answering

by

Xuan Fu

^1,†,

Jiangnan Du

²,

Hai-Tao Zheng

^3,*

,

Jianfeng Li

²,

Cuiqin Hou

²,

Qiyu Zhou

² and

Hong-Gee Kim

⁴

¹

Shenzhen International Graduate School, Tsinghua University, Shenzhen 518055, China

²

Ping An Technology, Shenzhen 518000, China

³

Pengcheng Laboratory, Shenzhen International Graduate School, Tsinghua University, Shenzhen 518055, China

⁴

Dental College, Seoul National University, Seoul 08826, Republic of Korea

^*

Author to whom correspondence should be addressed.

^†

Work conducted during the first author’s internship at Ping An Technology.

Electronics 2023, 12(7), 1692; https://doi.org/10.3390/electronics12071692

Submission received: 3 March 2023 / Revised: 21 March 2023 / Accepted: 27 March 2023 / Published: 3 April 2023

(This article belongs to the Special Issue Intelligent Big Data Analytics and Knowledge Management)

Download

Browse Figures

Versions Notes

Abstract

Open-Domain Question Answering (Open-Domain QA) aims to answer any factoid questions from users. Recent progress in Open-Domain QA adopts the “retriever-reader” structure, which has proven effective. Retriever methods are mainly categorized as sparse retrievers and dense retrievers. In recent work, the dense retriever showed a stronger semantic interpretation than the sparse retriever. When training a dual-encoder dense retriever for document retrieval and reranking, there are two challenges: negative selection and a lack of training data. In this study, we make three major contributions to this topic: negative selection by query generation, data augmentation from negatives, and a passage evaluation method. We prove that the model performs better by focusing on false negatives and data augmentation in the Open-Domain QA passage rerank task. Our model outperforms other single dual-encoder rerankers over BERT-base and BM25 by 0.7 in MRR@10, achieving the highest Recall@50 and the max Recall@1000, which is restricted by the BM25 retrieval results.

Keywords: open-domain question answering; passage rerank; data augmentation; negative selection; BERT

Share and Cite

MDPI and ACS Style

Fu, X.; Du, J.; Zheng, H.-T.; Li, J.; Hou, C.; Zhou, Q.; Kim, H.-G. SS-BERT: A Semantic Information Selecting Approach for Open-Domain Question Answering. Electronics 2023, 12, 1692. https://doi.org/10.3390/electronics12071692

AMA Style

Fu X, Du J, Zheng H-T, Li J, Hou C, Zhou Q, Kim H-G. SS-BERT: A Semantic Information Selecting Approach for Open-Domain Question Answering. Electronics. 2023; 12(7):1692. https://doi.org/10.3390/electronics12071692

Chicago/Turabian Style

Fu, Xuan, Jiangnan Du, Hai-Tao Zheng, Jianfeng Li, Cuiqin Hou, Qiyu Zhou, and Hong-Gee Kim. 2023. "SS-BERT: A Semantic Information Selecting Approach for Open-Domain Question Answering" Electronics 12, no. 7: 1692. https://doi.org/10.3390/electronics12071692

APA Style

Fu, X., Du, J., Zheng, H.-T., Li, J., Hou, C., Zhou, Q., & Kim, H.-G. (2023). SS-BERT: A Semantic Information Selecting Approach for Open-Domain Question Answering. Electronics, 12(7), 1692. https://doi.org/10.3390/electronics12071692

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Menu

SS-BERT: A Semantic Information Selecting Approach for Open-Domain Question Answering

Abstract

Share and Cite

Article Metrics

Article Access Statistics

Further Information

Guidelines

MDPI Initiatives

Follow MDPI