filler word

A filler word is the 'um,' 'uh,' or 'you know' that people say out of habit while thinking of what to say next or pausing briefly. These barely show up in writing, but if you listen to a recording of actual speech, you'll notice quite a few of them scattered between sentences.

Filler words used to be a headache for speech-to-text (transcription) features. If you transcribe speech exactly as spoken, the text ends up messy, full of things like 'um, you know, uh, this part.' So recent transcription technology has been moving toward automatically filtering out these sounds and cleaning up the text to keep only what matters. It's similar to a secretary who tidies up a transcript by cutting the noise and keeping just the key statements.

The article uses the phrase 'automatically filtering out filler words like um and uh.' It refers to Google's new transcription model automatically removing meaningless sounds that are just part of how people talk. It's worth remembering that filler words are simply a speaking habit, not errors, and that removing them isn't about changing what was said but about making the text easier to read.

1. Record a short voice memo and deliberately throw in a few 'um's and 'uh's while speaking.
2. Upload this recording to a service with a transcription (speech-to-text) feature.
3. Compare whether the resulting text still contains the 'um's and 'uh's, or whether they were automatically removed.
4. Try the same thing with a real conversation recording, like a meeting or an interview, and you'll see that text with filler words removed is much easier to read.