Hacker News Reader

Comment by Causality1 | Hacker News Reader

Will Facebook license this tech out? I can't be the only one who's noticed Google's speech recognition has gotten significantly less accurate the last couple of years, probably the result of some cost-cutting strategy.

View on HN

They have an open source repo linked from the article here: https://github.com/pytorch/fairseq/tree/master/examples/wav2...

There was a time a few years back where Google's speech recognition was virtually flawless (at least for my voice) but nowadays it verges on useless. I've had similar experiences with Google Maps (specifically navigation) and Google Translate. My understanding is that the implementation of these products is now mostly ML driven whereas in previous iterations ML was just one aspect of it. Curious to know how much truth there is to that or if I'm just viewing things with rose tinted glasses.

Indeed. Growing up playing with voice recognition on Windows I knew how to talk to computers. You speak clearly, enunciate your consonants, and keep a consistent pace and even tone. When I used my "computer voice" on Android I could carry on a text or IM conversation with my phone sitting in my pocket. Nowadays it hasn't gotten any better at understanding my natural drawl but the "computer voice" fills the sentences with bizarre punctuation and randomly-capitalized words that make me look like a lunatic.

I really don't know just what the hell my phone wants from me anymore.

I wonder if it's less accurate for specific people, but more accurate generally. In other words speech recognition was more accurate in the beginning on english-speaking men, and maybe now it's better for women, kids, people with accents, other languages, etc...

They train this on examples of speaking and maybe it's more broad now.

What in the world are you talking about?

He's saying that it might not be getting worse due to a cost-cutting strategy, but rather an attempt to optimize the value for the most possible people.

That's possible, but it wouldn't explain bizarre behavior like capitalizing random words in the middle of sentences or the absolute refusal to type the word "o'clock".

Reply on news.ycombinator.com