paper-with-me

홈 › Papers

HiTab: A Hierarchical Table Dataset for Question Answering and Natural Language Generation

2021-08-15 · ACL 2022 5 · Zhoujun Cheng, Haoyu Dong, Zhiruo Wang, Ran Jia, Jiaqi Guo, Yan Gao, Shi Han, Jian-Guang Lou, Dongmei Zhang

Tables are often created with hierarchies, but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables. Hierarchical tables challenge existing methods by hierarchical indexing, as well as implicit relationships of calculation and semantics. This work presents HiTab, a free and open dataset to study question answering (QA) and natural language generation (NLG) over hierarchical tables. HiTab is a cross-domain dataset constructed from a wealth of statistical reports (analyses) and Wikipedia pages, and has unique characteristics: (1) nearly all tables are hierarchical, and (2) both target sentences for NLG and questions for QA are revised from original, meaningful, and diverse descriptive sentences authored by analysts and professions of reports. (3) to reveal complex numerical reasoning in statistical analyses, we provide fine-grained annotations of entity and quantity alignment. HiTab provides 10,686 QA pairs and descriptive sentences with well-annotated quantity and entity alignment on 3,597 tables with broad coverage of table hierarchies and numerical reasoning types. Targeting hierarchical structure, we devise a novel hierarchy-aware logical form for symbolic reasoning over tables, which shows high effectiveness. Targeting complex numerical reasoning, we propose partially supervised training given annotations of entity and quantity alignment, which helps models to largely reduce spurious predictions in the QA task. In the NLG task, we find that entity and quantity alignment also helps NLG models to generate better results in a conditional generation setting. Experiment results of state-of-the-art baselines suggest that this dataset presents a strong challenge and a valuable benchmark for future research.

📄 PDF Abstract BibTeX arXiv:2108.06712

Code (1)

microsoft/hitab 공식 구현 pytorch

Tasks

DescriptiveEntity AlignmentQuestion AnsweringText Generation

Similar Papers 제목 키워드 기반

HiTab: A Hierarchical Table Dataset for Question Answering and Natural Language Generation

2021-10-16 · ACL ARR October 2021 10 · Anonymous

Tables are often created with hierarchies, but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables. Hierarchical tables challenge numerical reasoning by complex hierarchical inde…

Entity AlignmentQuestion AnsweringText Generation

Large Language Models are Complex Table Parsers

2023-12-13 · Bowen Zhao, Changkai Ji, Yuejie Zhang, Wen He 외

With the Generative Pre-trained Transformer 3.5 (GPT-3.5) exhibiting remarkable reasoning and comprehension abilities in Natural Language Processing (NLP), most Question Answering (QA) research has primarily centered aro…

Logical ReasoningQuestion Answering

General Table Question Answering via Answer-Formula Joint Generation

2025-03-16 · Zhongyuan Wang, Richong Zhang, Zhijie Nie

Advanced table question answering (TableQA) methods prompt large language models (LLMs) to generate answer text, SQL query, Python code, or custom operations, which impressively improve the complex reasoning problems in …

Question Answering

Orthogonal Hierarchical Decomposition for Structure-Aware Table Understanding with Large Language Models

2026-02-02 · Bin Cao, Huixian Lu, Chenwen Ma, Ting Wang 외 arxiv

Complex tables with multi-level headers, merged cells and heterogeneous layouts pose persistent challenges for LLMs in both understanding and reasoning. Existing approaches typically rely on table linearization or normal…

Question Answering

Towards Better Understanding Table Instruction Tuning: Decoupling the Effects from Data versus Models

2025-01-24 · Naihao Deng, Sheng Zhang, Henghui Zhu, Shuaichen Chang 외

Recent advances in natural language processing have leveraged instruction tuning to enhance Large Language Models (LLMs) for table-related tasks. However, previous works train different base models with different trainin…

Question Answering