kobotoolbox / kpi / 25368701766

82%

master: 76%

push

github

fix(nlp): update v1 speech recognition model to 'latest_long' DEV-2086 (#7000)

### 📣 Summary

Improved automatic transcription quality by specifying the correct
speech recognition model (v1 STT).

### 📖 Description

Audio transcriptions now use Google's `latest_long` model instead of the
unspecified default, which produced lower quality results.

7795 of 11991 branches covered (65.01%)

29616 of 36059 relevant lines covered (82.13%)

5.78 hits per line