Fact · Dictation for work and study
The 2022 study reports that adding an in-domain language model and common legal phrases to a generic speech model can improve its word error rate on court hearings.
Checked against its source on October 16, 2026
In the source's own words
“fine-tuning a generic off-the-shelf pre-trained Automatic Speech Recognition (ASR) system with an in-domain language model as well as infusing common phrases extracted with a collocation detection model can improve not only the Word Error Rate (WER)”
Better Transcription of UK Supreme Court Hearings, Hadeel Saadany et al., arXiv (Cornell University), 2022, Abstract.
Open the original- What it measures
- The result is for transcripts of court hearings. It shows that domain adaptation helps, which is why legal products ship with legal vocabularies.
- What it does not show
- A preprint. The abstract gives no numeric improvement, and the study is about courtroom audio rather than a lawyer dictating alone.