Speech recognition and models
ASR
Automatic speech recognition, the technical term for turning speech into text automatically.
ASR stands for automatic speech recognition. It is the technical term for speech recognition, and the one you meet in technical documentation, research papers and API names.
You will also see STT, speech-to-text, which means the same thing. ASR is used mostly in research, STT mostly in products.
Why you meet the abbreviation
The field is English-speaking, and the abbreviation is the search term that works. If you want to read how good a model is on a given language, find an API, or compare two vendors technically, ASR is the word to search for. The equivalent term in your own language usually returns far fewer and less precise results.
What ASR does not cover
- Speaker recognition, that is, who is speaking. That is called speaker recognition, or diarisation when it is about splitting the recording by speaker.
- Speech synthesis, that is, making speech from text. That is TTS, text-to-speech, and runs the other way.
- Language understanding. ASR writes the words down. What they mean is handled by a language model afterwards.