Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Textual Summarisation of Large Sets: Towards a General Approach

Kittipitch Kuptavanich, Ehud Reiter, Kees Van Deemter, Advaith Siddharthan

TL;DR

This paper extends a rule-based natural language generation approach for summarising large sets from consumer products to bibliographical references in academic papers. It introduces the refSet algorithm, which omits the dominating citation-count attribute, uses quantifiers and ranges, and highlights top-per-venue publications and authors to support readers in decision-making tasks. An evaluation with human participants shows that refSet improves usefulness in scenario-based decisions (e.g., reviewing or browsing) and generally outperforms a semantic Scholar baseline, with strong statistical support in one scenario and partial support in another. The work demonstrates that set-summarisation techniques can generalise across domains, while also outlining future directions for domain-agnostic and data-analysis-driven summarisation when corpus data are unavailable.

Abstract

We are developing techniques to generate summary descriptions of sets of objects. In this paper, we present and evaluate a rule-based NLG technique for summarising sets of bibliographical references in academic papers. This extends our previous work on summarising sets of consumer products and shows how our model generalises across these two very different domains.

Textual Summarisation of Large Sets: Towards a General Approach

TL;DR

This paper extends a rule-based natural language generation approach for summarising large sets from consumer products to bibliographical references in academic papers. It introduces the refSet algorithm, which omits the dominating citation-count attribute, uses quantifiers and ranges, and highlights top-per-venue publications and authors to support readers in decision-making tasks. An evaluation with human participants shows that refSet improves usefulness in scenario-based decisions (e.g., reviewing or browsing) and generally outperforms a semantic Scholar baseline, with strong statistical support in one scenario and partial support in another. The work demonstrates that set-summarisation techniques can generalise across domains, while also outlining future directions for domain-agnostic and data-analysis-driven summarisation when corpus data are unavailable.

Abstract

We are developing techniques to generate summary descriptions of sets of objects. In this paper, we present and evaluate a rule-based NLG technique for summarising sets of bibliographical references in academic papers. This extends our previous work on summarising sets of consumer products and shows how our model generalises across these two very different domains.
Paper Structure (14 sections, 5 figures, 2 tables)

This paper contains 14 sections, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Text summarising relating to sets of items
  4. Summarising sets of consumer products
  5. Introductory paragraph,
  6. Listing of products' important features,
  7. Lessons from our previous work
  8. Summarising sets of references
  9. A Survey of reader preferences
  10. Algorithm
  11. Evaluation
  12. Method
  13. Results
  14. Conclusion

Figures (5)

  • Figure 1: An excerpt from a Summary Generated by [refSet]
  • Figure 2: An Excerpt from a Summary Generated by [prodSet]
  • Figure 3: An excerpt from a summary taken from [sem]
  • Figure 4: Ranking Counts in Scenario 1
  • Figure 5: Ranking Counts in Scenario 2