U3E: Unsupervised and Erasure-based Evidence Extraction for Machine Reading Comprehension

Suzhe He; Shumin Shi; Chenghao Wu; Heyan Huang

doi:10.1109/CCIS57298.2022.10016420

U3E: Unsupervised and Erasure-based Evidence Extraction for Machine Reading Comprehension

Suzhe He, Shumin Shi^*, Chenghao Wu, Heyan Huang

^*Corresponding author for this work

School of Computer Science and Technology

Beijing Institute of Technology

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

More tasks in Machine Reading Comprehension (MRC) require predicting the answer and extracting supported evidence sentences simultaneously. However, the annotation of supporting evidence sentences is usually time-consuming and labor-intensive. In this paper, to address this issue and consider that most of the existing extraction methods are semi-supervised, we propose an unsupervised and erasure-based evidence extraction method named U3E, which takes the changes after sentence-level feature erasure in the document as input, simulating the decline in problem-solving ability caused by human memory decline. In order to make selections based on fully understanding the semantics of the context, we also propose metrics to quickly select the optimal memory model for these input changes. To compare U3E with typical evidence extraction methods and investigate its effectiveness in evidence, we conduct experiments on different datasets. Experimental results show that U3E is simple but effective, also improves model performance.

Original language	English
Title of host publication	Proceedings of 2022 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022
Editors	Fuji Ren, Witold Pedrycz, Zhiquan Luo, Dan Yang, Tianrui Li, Mengqi Zhou, Weining Wang, Aijing Li, Dandan Dandan, Liu Yaru Zou, Yanna Liu
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	585-589
Number of pages	5
ISBN (Electronic)	9781665477352
DOIs	https://doi.org/10.1109/CCIS57298.2022.10016420
Publication status	Published - 2022
Event	8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022 - Chengdu, China Duration: 26 Nov 2022 → 28 Nov 2022

Publication series

Name	Proceedings of 2022 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022

Conference

Conference	8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022
Country/Territory	China
City	Chengdu
Period	26/11/22 → 28/11/22

Keywords

Machine Reading Comprehension
evidence extraction
feature erasure
unsupervised learning

Access to Document

10.1109/CCIS57298.2022.10016420

Cite this

He, S., Shi, S., Wu, C., & Huang, H. (2022). U3E: Unsupervised and Erasure-based Evidence Extraction for Machine Reading Comprehension. In F. Ren, W. Pedrycz, Z. Luo, D. Yang, T. Li, M. Zhou, W. Wang, A. Li, D. Dandan, L. Y. Zou, & Y. Liu (Eds.), Proceedings of 2022 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022 (pp. 585-589). (Proceedings of 2022 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/CCIS57298.2022.10016420

He, Suzhe ; Shi, Shumin ; Wu, Chenghao et al. / U3E : Unsupervised and Erasure-based Evidence Extraction for Machine Reading Comprehension. Proceedings of 2022 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022. editor / Fuji Ren ; Witold Pedrycz ; Zhiquan Luo ; Dan Yang ; Tianrui Li ; Mengqi Zhou ; Weining Wang ; Aijing Li ; Dandan Dandan ; Liu Yaru Zou ; Yanna Liu. Institute of Electrical and Electronics Engineers Inc., 2022. pp. 585-589 (Proceedings of 2022 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022).

@inproceedings{0673763aae364ac0a69d9b42f8979db5,

title = "U3E: Unsupervised and Erasure-based Evidence Extraction for Machine Reading Comprehension",

abstract = "More tasks in Machine Reading Comprehension (MRC) require predicting the answer and extracting supported evidence sentences simultaneously. However, the annotation of supporting evidence sentences is usually time-consuming and labor-intensive. In this paper, to address this issue and consider that most of the existing extraction methods are semi-supervised, we propose an unsupervised and erasure-based evidence extraction method named U3E, which takes the changes after sentence-level feature erasure in the document as input, simulating the decline in problem-solving ability caused by human memory decline. In order to make selections based on fully understanding the semantics of the context, we also propose metrics to quickly select the optimal memory model for these input changes. To compare U3E with typical evidence extraction methods and investigate its effectiveness in evidence, we conduct experiments on different datasets. Experimental results show that U3E is simple but effective, also improves model performance.",

keywords = "Machine Reading Comprehension, evidence extraction, feature erasure, unsupervised learning",

author = "Suzhe He and Shumin Shi and Chenghao Wu and Heyan Huang",

note = "Publisher Copyright: {\textcopyright} 2022 IEEE.; 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022 ; Conference date: 26-11-2022 Through 28-11-2022",

year = "2022",

doi = "10.1109/CCIS57298.2022.10016420",

language = "English",

series = "Proceedings of 2022 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "585--589",

editor = "Fuji Ren and Witold Pedrycz and Zhiquan Luo and Dan Yang and Tianrui Li and Mengqi Zhou and Weining Wang and Aijing Li and Dandan Dandan and Zou, {Liu Yaru} and Yanna Liu",

booktitle = "Proceedings of 2022 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022",

address = "United States",

}

He, S, Shi, S, Wu, C & Huang, H 2022, U3E: Unsupervised and Erasure-based Evidence Extraction for Machine Reading Comprehension. in F Ren, W Pedrycz, Z Luo, D Yang, T Li, M Zhou, W Wang, A Li, D Dandan, LY Zou & Y Liu (eds), Proceedings of 2022 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022. Proceedings of 2022 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022, Institute of Electrical and Electronics Engineers Inc., pp. 585-589, 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022, Chengdu, China, 26/11/22. https://doi.org/10.1109/CCIS57298.2022.10016420

U3E: Unsupervised and Erasure-based Evidence Extraction for Machine Reading Comprehension. / He, Suzhe; Shi, Shumin; Wu, Chenghao et al.
Proceedings of 2022 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022. ed. / Fuji Ren; Witold Pedrycz; Zhiquan Luo; Dan Yang; Tianrui Li; Mengqi Zhou; Weining Wang; Aijing Li; Dandan Dandan; Liu Yaru Zou; Yanna Liu. Institute of Electrical and Electronics Engineers Inc., 2022. p. 585-589 (Proceedings of 2022 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - U3E

T2 - 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022

AU - He, Suzhe

AU - Shi, Shumin

AU - Wu, Chenghao

AU - Huang, Heyan

PY - 2022

Y1 - 2022

N2 - More tasks in Machine Reading Comprehension (MRC) require predicting the answer and extracting supported evidence sentences simultaneously. However, the annotation of supporting evidence sentences is usually time-consuming and labor-intensive. In this paper, to address this issue and consider that most of the existing extraction methods are semi-supervised, we propose an unsupervised and erasure-based evidence extraction method named U3E, which takes the changes after sentence-level feature erasure in the document as input, simulating the decline in problem-solving ability caused by human memory decline. In order to make selections based on fully understanding the semantics of the context, we also propose metrics to quickly select the optimal memory model for these input changes. To compare U3E with typical evidence extraction methods and investigate its effectiveness in evidence, we conduct experiments on different datasets. Experimental results show that U3E is simple but effective, also improves model performance.

AB - More tasks in Machine Reading Comprehension (MRC) require predicting the answer and extracting supported evidence sentences simultaneously. However, the annotation of supporting evidence sentences is usually time-consuming and labor-intensive. In this paper, to address this issue and consider that most of the existing extraction methods are semi-supervised, we propose an unsupervised and erasure-based evidence extraction method named U3E, which takes the changes after sentence-level feature erasure in the document as input, simulating the decline in problem-solving ability caused by human memory decline. In order to make selections based on fully understanding the semantics of the context, we also propose metrics to quickly select the optimal memory model for these input changes. To compare U3E with typical evidence extraction methods and investigate its effectiveness in evidence, we conduct experiments on different datasets. Experimental results show that U3E is simple but effective, also improves model performance.

KW - Machine Reading Comprehension

KW - evidence extraction

KW - feature erasure

KW - unsupervised learning

UR - http://www.scopus.com/inward/record.url?scp=85147672160&partnerID=8YFLogxK

U2 - 10.1109/CCIS57298.2022.10016420

DO - 10.1109/CCIS57298.2022.10016420

M3 - Conference contribution

AN - SCOPUS:85147672160

T3 - Proceedings of 2022 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022

SP - 585

EP - 589

BT - Proceedings of 2022 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022

A2 - Ren, Fuji

A2 - Pedrycz, Witold

A2 - Luo, Zhiquan

A2 - Yang, Dan

A2 - Li, Tianrui

A2 - Zhou, Mengqi

A2 - Wang, Weining

A2 - Li, Aijing

A2 - Dandan, Dandan

A2 - Zou, Liu Yaru

A2 - Liu, Yanna

PB - Institute of Electrical and Electronics Engineers Inc.

Y2 - 26 November 2022 through 28 November 2022

ER -

He S, Shi S, Wu C, Huang H. U3E: Unsupervised and Erasure-based Evidence Extraction for Machine Reading Comprehension. In Ren F, Pedrycz W, Luo Z, Yang D, Li T, Zhou M, Wang W, Li A, Dandan D, Zou LY, Liu Y, editors, Proceedings of 2022 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022. Institute of Electrical and Electronics Engineers Inc. 2022. p. 585-589. (Proceedings of 2022 8th IEEE International Conference on Cloud Computing and Intelligence Systems, CCIS 2022). doi: 10.1109/CCIS57298.2022.10016420

U3E: Unsupervised and Erasure-based Evidence Extraction for Machine Reading Comprehension

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this