Clozer: Adaptable Data Augmentation for Cloze-style Reading Comprehension

被引：0

作者：

Lovenia, Holy ^{[1
]}

Wilie, Bryan ^{[1
]}

Chung, Willy ^{[1
]}

Zeng, Min ^{[1
]}

Cahyawijaya, Samuel ^{[1
]}

Dan, Su ^{[1
]}

Fung, Pascale ^{[1
]}

机构：

[1] Hong Kong Univ Sci & Technol, Ctr Artificial Intelligence Res CAiRE, Hong Kong, Peoples R China

来源：

PROCEEDINGS OF THE 7TH WORKSHOP ON REPRESENTATION LEARNING FOR NLP | 2022年

关键词：

D O I：

暂无

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

Task-adaptive pre-training (TAPT) alleviates the lack of labelled data and provides performance lift by adapting unlabelled data to downstream task. Unfortunately, existing adaptations mainly involve deterministic rules that cannot generalize well. Here, we propose Clozer, a sequence-tagging based doze answer extraction method used in TAPT that is extendable for adaptation on any doze-style machine reading comprehension (MRC) downstream tasks. We experiment on multiple-choice dozestyle MRC tasks, and show that Clozer performs significantly better compared to the oracle and state-of-the-art in escalating TAPT effectiveness in lifting model performance, and prove that Clozer is able to recognize the gold answers independently of any heuristics.

引用

页码：60 / 66

页数：7

共 50 条

[41] MRCAug: Data Augmentation via Machine Reading Comprehension for Document-Level Event Argument Extraction [J].

Liu, Jian ;

Chen, Yufeng ;

Xu, Jinan .

IEEE/ACM Transactions on Audio Speech and Language Processing, 2022, 30 :3160-3172

[42] Speculation and negation identification via unified Machine Reading Comprehension frameworks with lexical and syntactic data augmentation [J].

Qian, Zhong ;

Zou, Ting ;

Zhang, Zihao ;

Li, Peifeng ;

Zhu, Qiaoming ;

Zhou, Guodong .

ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE, 2024, 131

[43] Extract, Integrate, Compete: Towards Verification Style Reading Comprehension [J].

Zhang, Chen ;

Lai, Yuxuan ;

Feng, Yansong ;

Zhao, Dongyan .

FINDINGS OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS, EMNLP 2021, 2021, :2976-2986

[44] RELATIONSHIPS BETWEEN MEASURES OF LEARNING STYLE AND READING-COMPREHENSION [J].

SCHMECK, RR .

PERCEPTUAL AND MOTOR SKILLS, 1980, 50 (02) :461-462

[45] Examination-Style Reading Comprehension with Neural augmented Retrieval [J].

Zhang, Yiqing ;

Zhao, Hai ;

Zhang, Zhuosheng .

PROCEEDINGS OF THE 2019 INTERNATIONAL CONFERENCE ON ASIAN LANGUAGE PROCESSING (IALP), 2019, :182-187

[46] CLOZE PROCEDURE AS A MEASURE OF READING-COMPREHENSION - THE EFFECTS OF VARIATIONS IN SUBJECT-MATTER ON TEXTUAL REDUNDANCY [J].

BAILEY, M ;

HARRISON, C .

HUMAN LEARNING, 1984, 3 (03) :185-201

[47] Modeling Local Item Dependence in Cloze and Reading Comprehension Test Items Using Testlet Response Theory [J].

Baghaei, Purya ;

Ravand, Hamdollah .

PSICOLOGICA, 2016, 37 (01) :85-104

[48] GNN-MRC: Machine Reading Comprehension Based on GNN Augmentation [J].

Zhang, Sen ;

Li, Baokui .

ARTIFICIAL NEURAL NETWORKS AND MACHINE LEARNING, ICANN 2023, PT IV, 2023, 14257 :398-409

[49] Thought-Path Contrastive Learning via Premise-Oriented Data Augmentation for Logical Reading Comprehension [J].

Wang, Chenxu ;

Jian, Ping ;

Yang, Zhen .

THIRTY-NINTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, AAAI-25, VOL 39 NO 24, 2025, :25345-25352

[50] Parallel Data Augmentation for Formality Style Transfer [J].

Zhang, Yi ;

Ge, Tao ;

Sun, Xu .

58TH ANNUAL MEETING OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS (ACL 2020), 2020, :3221-3228

← 1 2 3 4 5 →