Benchmarking Answer Verification Methods for Question Answering-Based Summarization Evaluation Metrics

TL;DR

This paper benchmarks various answer verification methods for question answering-based summarization metrics, revealing that improved verification does not always enhance overall metric performance due to dataset properties.

Contribution

It systematically compares lexical, BERTScore, and LERC verification methods, highlighting their strengths and limitations in QA-based summarization evaluation.

Findings

LERC outperforms lexical methods in some settings

Improved verification does not always improve overall metric quality

Dataset properties significantly influence verification impact

Abstract

Question answering-based summarization evaluation metrics must automatically determine whether the QA model's prediction is correct or not, a task known as answer verification. In this work, we benchmark the lexical answer verification methods which have been used by current QA-based metrics as well as two more sophisticated text comparison methods, BERTScore and LERC. We find that LERC out-performs the other methods in some settings while remaining statistically indistinguishable from lexical overlap in others. However, our experiments reveal that improved verification performance does not necessarily translate to overall QA-based metric quality: In some scenarios, using a worse verification method -- or using none at all -- has comparable performance to using the best verification method, a result that we attribute to properties of the datasets.