Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DETQUS: Decomposition-Enhanced Transformers for QUery-focused Summarization

Yasir Khan, Xinlei Wu, Sangpil Youm, Justin Ho, Aryaan Shaikh, Jairo Garciga, Rohan Sharma, Bonnie J. Dorr

TL;DR

DETQUS tackles query-focused tabular summarization by applying LLM-guided tabular decomposition to reduce input size while preserving query-relevant information. The framework couples decomposition with fine-tuned encoder-decoder models (T5, Flan-T5, BART, OmniTab) and demonstrates strong performance on the QTSUMM dataset, achieving a new state-of-the-art ROUGE-L of 0.4437 that surpasses REFACTOR. The study provides a comprehensive evaluation, including human judgments and error analysis, showing that decomposed inputs generally improve quality and offering insights into model strengths such as fact extraction versus holistic summarization. It also discusses ethical considerations and limitations, proposing future directions like chain-of-thought reasoning, ensemble methods, hierarchical decomposition, and alternative metrics to further advance scalable, accurate query-focused table summarization.

Abstract

Query-focused tabular summarization is an emerging task in table-to-text generation that synthesizes a summary response from tabular data based on user queries. Traditional transformer-based approaches face challenges due to token limitations and the complexity of reasoning over large tables. To address these challenges, we introduce DETQUS (Decomposition-Enhanced Transformers for QUery-focused Summarization), a system designed to improve summarization accuracy by leveraging tabular decomposition alongside a fine-tuned encoder-decoder model. DETQUS employs a large language model to selectively reduce table size, retaining only query-relevant columns while preserving essential information. This strategy enables more efficient processing of large tables and enhances summary quality. Our approach, equipped with table-based QA model Omnitab, achieves a ROUGE-L score of 0.4437, outperforming the previous state-of-the-art REFACTOR model (ROUGE-L: 0.422). These results highlight DETQUS as a scalable and effective solution for query-focused tabular summarization, offering a structured alternative to more complex architectures.

DETQUS: Decomposition-Enhanced Transformers for QUery-focused Summarization

TL;DR

DETQUS tackles query-focused tabular summarization by applying LLM-guided tabular decomposition to reduce input size while preserving query-relevant information. The framework couples decomposition with fine-tuned encoder-decoder models (T5, Flan-T5, BART, OmniTab) and demonstrates strong performance on the QTSUMM dataset, achieving a new state-of-the-art ROUGE-L of 0.4437 that surpasses REFACTOR. The study provides a comprehensive evaluation, including human judgments and error analysis, showing that decomposed inputs generally improve quality and offering insights into model strengths such as fact extraction versus holistic summarization. It also discusses ethical considerations and limitations, proposing future directions like chain-of-thought reasoning, ensemble methods, hierarchical decomposition, and alternative metrics to further advance scalable, accurate query-focused table summarization.

Abstract

Query-focused tabular summarization is an emerging task in table-to-text generation that synthesizes a summary response from tabular data based on user queries. Traditional transformer-based approaches face challenges due to token limitations and the complexity of reasoning over large tables. To address these challenges, we introduce DETQUS (Decomposition-Enhanced Transformers for QUery-focused Summarization), a system designed to improve summarization accuracy by leveraging tabular decomposition alongside a fine-tuned encoder-decoder model. DETQUS employs a large language model to selectively reduce table size, retaining only query-relevant columns while preserving essential information. This strategy enables more efficient processing of large tables and enhances summary quality. Our approach, equipped with table-based QA model Omnitab, achieves a ROUGE-L score of 0.4437, outperforming the previous state-of-the-art REFACTOR model (ROUGE-L: 0.422). These results highlight DETQUS as a scalable and effective solution for query-focused tabular summarization, offering a structured alternative to more complex architectures.

Paper Structure

This paper contains 24 sections, 2 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Explanation of the Task
  4. Prior Approaches to Tabular QA and Summarization
  5. Advances in Query-focused Text Summarization
  6. Advances in Table Summarization Techniques
  7. QTSUMM Dataset
  8. Our Methodological Approaches Derived from Previous Works
  9. System Architecture for Tabular Summarization
  10. Table Decomposition
  11. Algorithms for Table Decomposition
  12. Data and Experiments
  13. Dataset
  14. Preprocessing
  15. Fine-tuned Models
  16. ...and 9 more sections

Figures (2)

  • Figure 1: Query-focused table summarization with QTSUMM, generating a summary from the query: "What is the latest ... intelligence?"
  • Figure 2: Prompt for converting the table to markdown format for LLM (Llama3-70b)