Fact · Languages and accents
7.62percent character error rate
In the AISHELL-1 baseline study, the best of three 2017 models, LFMMI, reached a character error rate of 7.62 percent on the test set of Mandarin read speech.
Checked against its source on October 17, 2026
In the source's own words
“Table 5. Baseline results Model CER of dev CER of test MLLT+SAT 10.43% 12.23% TDNN-HMM 7.23% 8.42% LFMMI 6.44% 7.62%”
AISHELL-1: An Open-Source Mandarin Speech Corpus and A Speech Recognition Baseline, Hui Bu et al., arXiv (Oriental COCOSDA 2017), 2017, Table 5, baseline results.
Open the original- What it measures
- AISHELL-1 is a corpus of Mandarin read speech. The paper reports results only as character error rate, which is the usual way to score Mandarin because words are not separated by spaces.
- What it does not show
- The models are from 2017 and the speech is read aloud, not spontaneous. Newer systems score much lower on this corpus, so the number is a historical baseline, not a measure of current Mandarin dictation.