Semantic Distance: A New Metric for ASR Performance Analysis Towards Spoken Language Understanding
This addresses the need for better ASR evaluation metrics in spoken language understanding systems, though it is incremental as it builds on existing embedding methods.
The authors tackled the problem that Word Error Rate (WER) is a poor indicator for ASR performance in downstream NLU tasks by proposing Semantic Distance (SemDist), a new metric based on sentence embeddings from RoBERTa, which they demonstrated effectively evaluates ASR on tasks like intent recognition and semantic parsing.
Word Error Rate (WER) has been the predominant metric used to evaluate the performance of automatic speech recognition (ASR) systems. However, WER is sometimes not a good indicator for downstream Natural Language Understanding (NLU) tasks, such as intent recognition, slot filling, and semantic parsing in task-oriented dialog systems. This is because WER takes into consideration only literal correctness instead of semantic correctness, the latter of which is typically more important for these downstream tasks. In this study, we propose a novel Semantic Distance (SemDist) measure as an alternative evaluation metric for ASR systems to address this issue. We define SemDist as the distance between a reference and hypothesis pair in a sentence-level embedding space. To represent the reference and hypothesis as a sentence embedding, we exploit RoBERTa, a state-of-the-art pre-trained deep contextualized language model based on the transformer architecture. We demonstrate the effectiveness of our proposed metric on various downstream tasks, including intent recognition, semantic parsing, and named entity recognition.