Unsupervised Evaluation for Question Answering with Transformers

TL;DR

This paper introduces an unsupervised method to evaluate the correctness of answers in transformer-based question answering systems by analyzing hidden representations, achieving high accuracy without labeled data.

Contribution

It presents a novel unsupervised approach that leverages hidden representations in transformers to evaluate answer correctness across multiple datasets and domains.

Findings

Achieves 91.37% accuracy on SQuAD

Outperforms heuristic baselines

Works across diverse datasets and domains

Abstract

It is challenging to automatically evaluate the answer of a QA model at inference time. Although many models provide confidence scores, and simple heuristics can go a long way towards indicating answer correctness, such measures are heavily dataset-dependent and are unlikely to generalize. In this work, we begin by investigating the hidden representations of questions, answers, and contexts in transformer-based QA architectures. We observe a consistent pattern in the answer representations, which we show can be used to automatically evaluate whether or not a predicted answer span is correct. Our method does not require any labeled data and outperforms strong heuristic baselines, across 2 datasets and 7 domains. We are able to predict whether or not a model's answer is correct with 91.37% accuracy on SQuAD, and 80.7% accuracy on SubjQA. We expect that this method will have broad applications, e.g., in the semi-automatic development of QA datasets