Re:English has been my pain for 15 years (2013)
by retrac on Wednesday December 22, 2021 @10:05AM
(#29650245)
Yes, I was thinking of speech synthesizers when I wrote my comment :) Closing the gap is devillish and requires increasingly complex grammatical analysis and more and more rules. One of the big barriers for early speech synthesizers was enough RAM/ROM to cram all the text-to-phonetic transformation rules.
I believe that modern approaches mostly use deep learning. And definitely still aren't perfect. I don't use speech synthesizers regularly personally as a tool, but synthesized speech is common enough that I do hear obvious stress errors in synthesized speech every day. Probably because English orthography is most opaque about stress, which basically doesn't exist as far as the writing system is concerned.
Re:English has been my pain for 15 years (2013)
by aasasd on Wednesday December 22, 2021 @12:06PM
(#29652073)
> synthesized speech is common enough
I'm curious: where do you encounter much synthesized speech? In my ‘fairly general’ usage (if there is such a thing), I only hear it in YouTube tutorials when the author is totally reluctant to both use their voice and do on-screen text.
Re:English has been my pain for 15 years (2013)
by retrac on Thursday December 23, 2021 @09:54AM
(#29662377)
Bus stop announcements on transit, those annoying interactive phone menus, people asking Siri math questions they should be able to do in their head, stuff like that.
Re:English has been my pain for 15 years (2013)
by formerly_proven on Wednesday December 22, 2021 @10:27AM
(#29650531)
> Words like read (present vs. past tense) or Nice (the adjective vs. the French city) can be pronounced in two different ways. Same with an address / to address, a present / to present etc.
lead is somehow even worse than read.
Words like read (present vs. past tense) or Nice (the adjective vs. the French city) can be pronounced in two different ways. Same with an address / to address, a present / to present etc.
Determining which way to use is impossible without understanding the text. You can get pretty far by using some simple heuristics. For example, when the word read is preceded by "have", you almost always want to use the past tense pronunciation, but if it's preceded by "to", you want the present tense. At some point, though, you will hit the "time flies like an arrow" vs. "fruit flies like a banana" problem, and simple sentence analysis will not be enough.
Another issue with English is that it likes to mash words together, and most pronunciation rules are not ready for that. While "Black Lives Matter" might be easy to pronounce, #blacklivesmatter is not. Getting this wrong sometimes leads to hilarious results[1].
I'm blind and I use a speech synthesizer daily, so I'm familiar with most of these issues. I haven't found a single piece of software that actually solved them well. This is still much better than languages like Hebrew, though, as they require you to manually insert vowels in the appropriate places, a task modern computers are not ready for.
[1] https://www.boredpanda.com/worst-domain-names/?utm_source=go...