Hacker News new | past | comments | ask | show | jobs | submit login

that must be qualified because there are a lot of ML applications that have basically been adopted in the mainstream, like voice recognition (which is pretty good, probably in the 80-90% accuracy range and better for specialized contextual tasks). I remember driving for lyft and being able to input destination addresses by voice (2 years ago; had a Moto X where this was cutting edge) was a godsend and improved my customer service ratings.



80% accuracy for voice is basically unusable. Low 90s will get someone to persist enough to optimise the voice model for them.

For continuous use, you need 95% at least. See http://anewdomain.net/2014/01/28/lamont-wood-windows-speech-...


> voice recognition ... probably in the 80-90% accuracy range

State of the art systems are far better than this. Microsoft recently published a paper with a 5.9% word error rate for conversational speech. Speech directed at computers/assistants is already in the high 90s, though I don't have a figure off the top of my head.


I gauged it off of my personal gut feeling, which includes a coefficient for "well I don't have a strong enough internet connection so I get google's spinner instead - and then it fails."

Occasionally, android gets the words right (as demonstrated by the onscreen text) and then flubs passing the correct intent because of "loss of connection", which is just about the most frustrating ML fail.

No doubt android's voice recognition is spectacular. I can prompt it in three different relatively orthogonal-sounding languages (English, French, Japanese), and it can figure out which language I'm using and usually get the transcription correct. Notably, I can't activate the italian/japanese pair and get useful results - which makes sense if you know both languages.

Google voice is horrible, however, at transcribing voicemail.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: