Word-Level ASR Quality Estimation for Efficient Corpus Sampling and Post-Editing through Analyzing Attentions of a Reference-Free Metric

• DOI: 10.48550/arxiv.2401.11268
• Publication Date: 2024-01-01
• Authors: Golara Javadi; Kamer Ali Yüksel; Yunsu Kim; Thiago Castro Fer...; Mohamed Al-Badras...

In the realm of automatic speech recognition (ASR), quest for models that not only perform with high accuracy but also offer transparency in their decision-making processes is crucial. The potential quality estimation (QE) metrics introduced and evaluated as a novel tool to enhance explainable artificial intelligence (XAI) ASR systems. Through experiments analyses, capabilities NoRefER (No Reference Error Rate) metric are explored identifying word-level errors aid post-editors refining hypotheses. investigation extends utility corpus-building process, demonstrating its effectiveness augmenting datasets insightful annotations. diagnostic aspects examined, revealing ability provide valuable insights into model behaviors decision patterns. This has proven beneficial prioritizing hypotheses post-editing workflows fine-tuning models. findings suggest merely error detection comprehensive framework enhancing systems' transparency, efficiency, effectiveness. To ensure reproducibility results, all source codes this study made publicly available.