How to Evaluate a Summarizer: Study Design and Statistical Analysis for Manual Linguistic Quality Evaluation

TL;DR

This paper investigates evaluation methods for automatic text summarization, highlighting inconsistencies in current practices, and demonstrates how evaluation design and statistical analysis significantly impact the reliability of linguistic quality assessments.

Contribution

It provides empirical evidence on the effects of annotation methods, study parameters, and statistical analysis choices on summarization evaluation reliability.

Findings

Likelihood and ranking annotations vary in effectiveness across aspects.

Study parameters like annotator count influence statistical power.

Current analysis methods can significantly inflate error rates.

Abstract

Manual evaluation is essential to judge progress on automatic text summarization. However, we conduct a survey on recent summarization system papers that reveals little agreement on how to perform such evaluation studies. We conduct two evaluation experiments on two aspects of summaries' linguistic quality (coherence and repetitiveness) to compare Likert-type and ranking annotations and show that best choice of evaluation method can vary from one aspect to another. In our survey, we also find that study parameters such as the overall number of annotators and distribution of annotators to annotation items are often not fully reported and that subsequent statistical analysis ignores grouping factors arising from one annotator judging multiple summaries. Using our evaluation experiments, we show that the total number of annotators can have a strong impact on study power and that current statistical analysis methods can inflate type I error rates up to eight-fold. In addition, we highlight that for the purpose of system comparison the current practice of eliciting multiple judgements per summary leads to less powerful and reliable annotations given a fixed study budget.