Peer-reviewed research
Semantic Distance: A New Metric for ASR Performance Analysis Towards Spoken Language Understanding
Suyoun Kim et al., Proceedings of Interspeech 2021, ISCA, 2021
Read the originalRead on October 17, 2026
- What it is
- A 2021 study by Meta AI researchers that proposes semantic distance, a metric that compares the meaning of a recognized sentence with the reference, as an alternative to WER for downstream language tasks.
- How much weight it carries
- Peer-reviewed. The metric is evaluated on intent recognition, slot filling and related tasks, not on every kind of transcription, and it has not replaced WER as the standard.
Facts taken from this source
- Kim and colleagues write that WER takes into account only literal correctness, not semantic correctness, which matters for downstream language understanding.
“WER takes into consideration only literal correctness instead of semantic correctness” (Abstract)