Just cough into here, please: Cambridge-developed app slurps the sounds of COVID-19

A team at the UK's Cambridge University has created an application to collect the speaking, breathing and coughing sounds of participants in the hope of building predictive models to "contribute to the early diagnosis of COVID-19". Development has been led by Professor Cecilia Mascolo from the university's Department of …

[Anonymous Coward]

Rec-captcha

> Those good intentions are potentially undermined by the use of Google's reCAPTCHA service, used to deter bots from recording their coughs,

Coughing into speech recognition and then pasting that in as the response to the audio captcha probably works.

I got fed up with identifying traffic lights and wondering whether the pole counted or not so switched to using the audio instead. It's now a game to see just how unlike the sound a word can be and still get through. It's surprisingly lax.

[Icon -> because the need for captchas in the first place]

[Tim Anderson]

Re: Rec-captcha

problem is, reCaptcha 3 doesn't work like that. You don't type an answer or click an image, instead it looks at the data around the interaction - including, I suspect, what is entered into a form - and uses AI to score the result. So it grabs a lot of data.

[Anonymous Coward]

So..

..it's nothing to sneeze at ..

Oh come one, someone had to say it :)

My main concern is that Google is in the recording chain somewhere - all you need now is for them to try and curry favour with President Pandemic and say they can track foreign exposures. I'm not impressed with a data sniffer in the middle of it all. That said, I find it an interesting idea.

[Mage]

This is a data grab.

Self selecting?

Google's reCAPTCHA service should be illegal on any non-Google site.

Also don't you need a large selection of coughs that are NOT Covid-19 and proper verification of the data or else it's useless research.

Data gathering disguised as research?

[Jimmy2Cows]

Re: This is a data grab.

To be fair, the article does say users are asked to indicate whether or not they've tested positive for COVID-19 in the questionaire, as well as submitting audio samples.

To be equally fair, they could have made it much clearer that they'll need as many samples as possible from all cohorts - apparently healthy (no COVID-19 symptoms), potential (has symptoms but not tested) and definite (knowingly has / doesn't have COVID-19).

[Anonymous Coward]

Nuke to catch a fly?

These seems a much shotgun approach. I would assume it's far too unreliable to make it worth it vs just taking in any bad coughs (If its bad enough to be like coronavirus it's probably some life threatening chest infection). Right?

It could tell you "stop being a whiny child and get back to work" if the coughs are nothing. But as this covid virus mild for some, that seems a risky idea.