Answers Unite! Unsupervised Metrics for Reinforced Summarization Models
This work addresses the need for more accurate and reference-free evaluation metrics in summarization, offering an incremental improvement for researchers and practitioners in natural language processing.
The paper tackled the problem of biased and suboptimal evaluation in abstractive summarization by proposing unsupervised metrics based on Question Answering, which improved over ROUGE in human evaluation and led to better performance when used as rewards in Reinforcement Learning models.
Abstractive summarization approaches based on Reinforcement Learning (RL) have recently been proposed to overcome classical likelihood maximization. RL enables to consider complex, possibly non-differentiable, metrics that globally assess the quality and relevance of the generated outputs. ROUGE, the most used summarization metric, is known to suffer from bias towards lexical similarity as well as from suboptimal accounting for fluency and readability of the generated abstracts. We thus explore and propose alternative evaluation measures: the reported human-evaluation analysis shows that the proposed metrics, based on Question Answering, favorably compares to ROUGE -- with the additional property of not requiring reference summaries. Training a RL-based model on these metrics leads to improvements (both in terms of human or automated metrics) over current approaches that use ROUGE as a reward.