Sentence-Level Fluency Evaluation: References Help, But Can Be Spared!
This work addresses the need for efficient and accurate fluency evaluation in natural language processing, offering both referenceless and reference-based metrics that improve upon existing methods, though it is incremental in nature.
The paper tackles the problem of evaluating sentence-level fluency in natural language generation by proposing SLOR and WPSLOR as referenceless metrics, which achieve significantly higher correlation with human fluency scores than word-overlap metrics like ROUGE on a benchmark dataset. It also introduces ROUGE-LM, a reference-based metric that outperforms all baselines, including WPSLOR, in correlation with human judgments.
Motivated by recent findings on the probabilistic modeling of acceptability judgments, we propose syntactic log-odds ratio (SLOR), a normalized language model score, as a metric for referenceless fluency evaluation of natural language generation output at the sentence level. We further introduce WPSLOR, a novel WordPiece-based version, which harnesses a more compact language model. Even though word-overlap metrics like ROUGE are computed with the help of hand-written references, our referenceless methods obtain a significantly higher correlation with human fluency scores on a benchmark dataset of compressed sentences. Finally, we present ROUGE-LM, a reference-based metric which is a natural extension of WPSLOR to the case of available references. We show that ROUGE-LM yields a significantly higher correlation with human judgments than all baseline metrics, including WPSLOR on its own.