Where do Recaptcha audio alternatives come from?
January 19, 2019 12:04 PM   Subscribe

I'm totally blind, and so end up solving a lot of Recaptchas from GOogle, AKA the "I'm not a robot," checkbox. THeir audio challenges are weird, and I wonder where they come from?

Google seems to take excerpts from… Lectures, or perhaps videos, as far as I understand it. They play a few words of what's clearly a larger bit of audio, and usually I can type the answer in correctly and move on. The topics are sometimes hard to distinguish since they're so brief, but I wonder where they got such eclectic material.

Any insight into how the audio captcha works would be awesome. I presume they have audio transcriptions of some sort to match against the user's input.

I'm not interested in any sort of malicious use for this info, call this idle curiosity :)
posted by Alensin to Computers & Internet (2 answers total) 3 users marked this as a favorite
 
Youtube, probably?
posted by zamboni at 12:45 PM on January 19 [2 favorites]


That's interesting. I know that a lot of the visual recaptchas are actually used in AI training. The two main versions are either typing in a distorted looking collection of letters helps the AI learn to scan old printed material correctly. And identifying pictures of road signs is used to train self driving cars.

So I would guess that the audio recaptchas are to help develop better transcription/translation software? (Not that that helps with where they came from, I know)
posted by Caravantea at 2:47 PM on January 19 [4 favorites]


« Older If I wanted to make the best chicken soup ever...   |   Ankle/foot hurts and feels numb all of a sudden.... Newer »

You are not logged in, either login or create an account to post comments