Automatic verification of data summaries

Rezgui, Rayhane; Saeed, Mohammed; Papotti, Paolo
INLG 2021, 14th International Conference on Natural Language Generation, 20-24 September 2021, Aberdeen, United Kingdom (Virtual Event)

We present a generic method to compute the factual accuracy of a generated data summary with minimal user effort. We look at the problem as a fact-checking task to verify the numerical claims in the text. The verification algorithm assumes that the data used to generate the text is available. In this paper, we describe how the proposed solution has been used to identify incorrect claims about basketball textual summaries in the context of the Accuracy Shared Task at INLG 2021.

Type:
Poster / Demo
City:
Aberdeen
Date:
2021-09-20
Department:
Data Science
Eurecom Ref:
6631
Copyright:
Copyright ACL. Personal use of this material is permitted. The definitive version of this paper was published in INLG 2021, 14th International Conference on Natural Language Generation, 20-24 September 2021, Aberdeen, United Kingdom (Virtual Event) and is available at :

PERMALINK : https://www.eurecom.fr/publication/6631