vix.ing · top · new · best · stats · spec

A Gold Standard Methodology for Evaluating Accuracy in Data-To-Text\n Systems

2020/11/08 by Craig Thomson, Ehud Reiter, Thomson, Craig +1 · 1 citation
Computer Science · #Natural Language Processing Techniques #Semantic Web and Ontologies #Topic Modeling

paper · pdf · doi:10.48550/arxiv.2011.03992

Abstract

Most Natural Language Generation systems need to produce accurate texts. We\npropose a methodology for high-quality human evaluation of the accuracy of\ngenerated texts, which is intended to serve as a gold-standard for accuracy\nevaluations of data-to-text systems. We use our methodology to evaluate the\naccuracy of computer generated basketball summaries. We then show how our gold\nstandard evaluation can be used to validate automated metrics\n

Cited by

Related