What is speech recognition?
Speech recognition is AI that turns speech into text, for example when someone dictates a note.
Modern speech recognition builds on deep learning and handles ordinary speech at a normal pace. Technical terms, dialects and background noise can still cause errors.
Together with a language model, a dictation can become more than a transcript. The text can be arranged after a template, for example with headings for background, assessment and action.
In healthcare it is used for draft records. Staff review the text and are responsible for what is documented.
How it works
The audio is split into short segments that a model interprets as sounds, words and sentences. The model has been trained on large amounts of recorded speech with matching text. It takes context into account, so a word that sounds unclear can still come out right when the rest of the sentence is clear. When several people speak, the system can also tell the voices apart.
Why it matters
A lot of work happens out loud. Meetings, calls and notes after a visit. Speech recognition makes what was said searchable and easy to work with, and together with a language model it can become minutes or a summary. The person speaking no longer has to type and can stay in the conversation.
What to keep in mind
Quality depends on the audio. A good microphone and a quiet room make a big difference. Word lists with names and technical terms help the system get them right.
Decide how long recordings are kept as well. It is often enough to keep the text and delete the audio once the transcript is done.
Models differ, especially in Swedish and with specialist vocabulary. Test with recordings from your own setting, with the voices and words that actually occur.
What it means for you
Speech can contain information as sensitive as text. Ask where the audio is processed and whether it is kept.
Common questions
How accurate is speech recognition?
It handles everyday speech well. Technical terms, names and accents can come out wrong, so read the text through before it is used.
Where does the audio go?
That depends on where the service runs. Recordings of patients and clients are often sensitive. Speech recognition can also run in your own environment.
Can speech recognition write clinical notes?
It can produce a draft from what is dictated. Responsibility for the content stays with the clinician who keeps the record.
Related terms
Read more
All terms Updated