Fact · Languages and accents
10.31percent character error rate
The KoSpeech baseline reached a character error rate of 10.31 percent on the KsponSpeech corpus of spontaneous Korean, using the acoustic model only.
Checked against its source on October 17, 2026
In the source's own words
“By this metric, we achieved 10.31% CER only using acoustic models.”
KoSpeech: Open-Source Toolkit for End-to-End Korean Speech Recognition, Soohwan Kim et al., arXiv, 2020, Section 4, results.
Open the original- What it measures
- The score is the Levenshtein distance between the recognized and reference text, divided by the length of the reference script.
- What it does not show
- A 2020 baseline without a language model, so current systems score lower. The corpus is conversational and recorded in clean conditions, so it does not show accuracy in noisy rooms.