Speech recognition and models
Code-switching
Shifting between two or more languages in the same conversation, often within a single sentence.
Code-switching is shifting between languages in the same conversation, and often within the same sentence. “Vi må deploye før vi merger” is Norwegian sentence structure with English verbs.
In Nordic work meetings this is the norm, not the exception. IT, consulting, finance and academia are full of English jargon inside local sentences.
Why it is a problem
Most transcription systems pick one language for the whole recording. If the pick is the local language, the English words get pressed into local spelling: deploie, meile, shære.
The worst part of this error type is where it lands. The jargon is precisely the set of words that carried the content of the meeting. Ordinary words come out right, and the ones that mattered come out wrong.
What you can do
- Choose a tool that handles it. Some systems detect language continuously instead of locking one in for the whole file.
- Enter the jargon in advance if the tool supports a custom dictionary. That is often the simplest and most effective step.
- Split the recording at clear language switches, if the meeting genuinely moves into another language for longer stretches.
- Expect to edit. Build a find-and-replace list of the twenty jargon terms your organisation uses. The same list works for all your meetings.