Author Topic: Speech transcription software  (Read 813 times)

Speech transcription software
« on: 03 June, 2024, 11:53:56 am »
Is there such a thing as free software that can transcribe speech more or less as it is happening? Does anyone have any experience of this?
I am aware that his kind of thing is very much not perfect.

Context: my father has been persuaded to be on the committee of his Bridge club, even though he has repeatedly said that he is old ( 80) and deaf. They seem to be quite desperate, though (I think they may need a quorum). The chair person has suggested that running speech transcription software might help Dad to hear during committee meetings. Dad is not convinced. Nor am I, but I said I would do some research, and I think there are people on here who have used this kind of thing.
Dad reckons his latest hearing aids are better and might be good enough in combination with his "mini loop system".

Cudzoziemiec

  • Ride adventurously and stop for a brew.
Re: Speech transcription software
« Reply #1 on: 03 June, 2024, 12:08:32 pm »
A deaf friend has an app on her phone which does this. I'll ask her what it's called. It's certainly not perfect but the mistakes are understandable in context (and sometimes very amusing), but it is a huge battery eater, so your father might need a powerbank for his phone to keep it going if the meetings are longer than about half an hour.
Riding a concrete path through the nebulous and chaotic future.

Cudzoziemiec

  • Ride adventurously and stop for a brew.
Re: Speech transcription software
« Reply #2 on: 03 June, 2024, 12:23:22 pm »
Apparently it's simply "Live translate" free download on Android. Not available for iPhones.
Riding a concrete path through the nebulous and chaotic future.

Kim

  • Timelord
    • Fediverse
Re: Speech transcription software
« Reply #3 on: 03 June, 2024, 12:31:30 pm »
There's an Otter thing that barakta sometimes uses.

The rule of thumb here is that speech recognition software struggles with context in almost exactly the same way that a deaf person does.  So it will mangle critical nouns, drop 'not' in the worst possible places, and otherwise generate grammatically correct nonsense.  Which makes it more useful as a memory aid than it does for understanding what's going on in real time.  If two people are having a conversation, and the speaker can watch the transcript and clarify when it goes wrong, it's a bit more useful than it is for someone giving a presentation.

The gold standard is STTR, with a human operator (preferably in the same room, but they can work remotely).  This is sufficiently expensive that it's in the realm of conference budgets or Access To Work funding, rather than something a Bridge Club is likely to be able to afford.

And regardless of whether you're using speech recognition or a remote human operator, microphone technique is a complicated subject with a need for more than one highlighter pen.  If you're feeding the listener room audio from a fondleslab's internal mic, they're going to struggle.

barakta

  • Bastard lovechild of Yomiko Readman and Johnny 5
Re: Speech transcription software
« Reply #4 on: 03 June, 2024, 12:41:59 pm »
There's a few built-in options on phones and Otter.ai is free for 40 min chunks and not-expensive for slightly longer slots.

For guides on free built in options see https://mcmw.abilitynet.org.uk/?Adjustment=Captions and select your device and operating system.

In practice the speech to text is only as good as the sound quality into the device it runs on. The org can help by strong chairing, one person speaking at a time, agenda and sticking to topic and so on. If your dad has gadgets with his hearing aid and it is a like a mic, then ask people to only speak with the mic, the chair can coordinate and the natural pauses are a good break for everyone.

Cudzoziemiec

  • Ride adventurously and stop for a brew.
Re: Speech transcription software
« Reply #5 on: 03 June, 2024, 01:05:38 pm »
Even (or especially) human transcription fails when the source sound is a conference telephone call with everyone talking over each other, ambient noise, and no speaker introduction or identification.
Riding a concrete path through the nebulous and chaotic future.

Re: Speech transcription software
« Reply #6 on: 03 June, 2024, 01:14:24 pm »
There's a few built-in options on phones and Otter.ai is free for 40 min chunks and not-expensive for slightly longer slots.

For guides on free built in options see https://mcmw.abilitynet.org.uk/?Adjustment=Captions and select your device and operating system.

In practice the speech to text is only as good as the sound quality into the device it runs on. The org can help by strong chairing, one person speaking at a time, agenda and sticking to topic and so on. If your dad has gadgets with his hearing aid and it is a like a mic, then ask people to only speak with the mic, the chair can coordinate and the natural pauses are a good break for everyone.

I've used Otter to transcribe meetings I've had at work, and it's pretty good.  :thumbsup:

Re: Speech transcription software
« Reply #7 on: 03 June, 2024, 02:08:06 pm »
Thank you all. That gives me things to try. I hadn't thought of there being apps, but of course there are.
(I should have added that Dad has standard-issue Old Person's Hearing Loss, nothing more complicated.)

barakta

  • Bastard lovechild of Yomiko Readman and Johnny 5
Re: Speech transcription software
« Reply #8 on: 03 June, 2024, 03:26:40 pm »
Even standard old people deafness will affect the higher frequencies of speech which are most needed in English for distinguishing words from one another. He may also lose his ability to locate sounds in space even with clever hearing aid features.

I hope a mix of good chairing and tech enables him to enjoy bridge committeeing. Do come back to us if he's still stuck as we can always try and think of other ideas.

fruitcake

  • some kind of fruitcake
Re: Speech transcription software
« Reply #9 on: 14 June, 2024, 07:31:38 am »
Automated subtitling is odd in the words it 'mishears'. As others have said, it's not context aware, nor is it aware of the way spoken language is generally used.

I'm reminded that lots of social media video content is auto-subtitled so that people can consume video when they're on the bus or wherever: the subtitling errors can be bizarre, sometimes comical, and that can be distracting; the lack of any punctuation can make for an exhausting read too.

I imagine auto transcription apps on smartphones would use similar software with a similar capability level, and throw up similar issues. It would be least useful when two people speak together, which happens a lot in real life, and that might be when transcription support is most needed.

fruitcake

  • some kind of fruitcake
Re: Speech transcription software
« Reply #10 on: 14 June, 2024, 08:36:41 am »
I think in this situation I'd want good minutes, and a capable minute taker, so I could read up on anything I'd missed. I've been in meetings where minutes are typed on a laptop as the meeting is taking place and then shared (via a data projector) and approved at the close of the meeting.

As technology to facilitate this, XVGA data projectors can be bought cheaply (£50 secondhand), now that offices and consumers are upgrading to the latest and greatest higher definitions.

I've also taken part in meetings where a PC monitor connected to the minute taker's laptop is placed on the meeting table, so that one person can read the minutes as they are being typed. I can imagine this set up being applied to transcription. I'm imagining a human doing the typing there, rather than auto-transcription.

This approach - a human doing the typing along with, I expect, some summarising and rephrasing of messy sections of spoken word - is the approach used to capture training sessions and lectures that I've been to.

I should say that I'm not suggesting this instead of the user's hearing aids / strong Chairing / lip-reading / other tactics required by the individual.