NoaLingua for Chrome
Add to Chrome — Free Add to Chrome

Glossary

Speech recognition (ASR) — the NoaLingua glossary

Automatic speech recognition converts spoken audio into text.

For language learners it has two uses: transcribing material that has no subtitles, and checking whether what you said matches what you meant to say.

Both are less reliable than they appear. Recognition accuracy drops on accented speech — which is precisely what a learner produces — and on proper nouns, which is precisely what subtitles are most useful for. A low pronunciation score can mean the recogniser misheard rather than that you mispronounced.

In NoaLingua Say It Back writes out what it hears in real time and shows which words did not pass, so you can tell a mispronunciation from a misrecognition. It keeps the score and the missed words — never a recording of your voice.

Sources

Research and official documentation used for this definition.

  1. Research The impact of non-native English speakers’ phonological and prosodic features on ASR accuracy Speech Communication, 2024

    Why browser speech-recognition output can reflect recogniser limitations as well as a learner’s pronunciation.

  2. First-hand product evidence Shadowing and pronunciation practice: product evidence and limits NoaLingua feature documentation

    Plain captures of the speaking turn and missed-word result, plus the boundary on what the score can mean.