If you're a fluent English speaker, you likely have a skill that's as subtle as it is hard to program into a computer: you can look at a word you've never seen before (even one that's gibberish) and have a sense for how it's broken up into syllables and phonemes. Consider some nonsense I made up like "volumentesciently" and think about how your brain decided to group consonants and vowels, stressed and unstressed syllables, etc
- any / all
Making stuff to distract myself from existential dread
Art: @graham-illustrations
Dreams: @graham-dream-journal
Wizards: @make-up-a-wizard
Partner's Pottery: @kp-pottery
You must log in to comment.
in reply to @graham's post:
I'm getting a major parsing error on the "sci" (es-si-en, es-sai-en, esh-en) and a minor one on the "olu"(vo-loo, vol-yoo).
So, I've been given the impression that most anglophone schools teach "English" as a combination of linguistics and literature and most emphasis on the latter. Here in Latvia we have "Latvian" as the linguistics subject and "Literature" as, well, the thing it says. Which means among other things that we get taught some of the rules that govern the language, including syllable splitting – divide along prefixes and roots, then zoom in and seperate out vowel/dipthong sounds, split consonant sounds evenly between neighboring wovels/dipthongs, the middle one going to the following syllable. (technically these are hyphenation rules, which have a near, but not full overlap with syllables, but that's no longer middle school stuff)
English complicates that a little by having inconsistent orthography and you just gotta apply a statistical best guess (or a good understanding of etymology) to that, but the principle tends to work.
Ah, I didn't mean to make it difficult! 
That takes away from my point, haha. I was imagining the "volu" as the same as "volume" and the scient as starting with a "sh" or "s" sound like in "prescient".
For me, it might be pronounced something like "vol-yu-men-teh-shient-ly"
I definitely feel like I was told growing up that the best way to learn English was to read a ton. Even if I've found that to be true in my experience, it doesn't really solve the "how to pronounce stuff" unless there's also a heavy amount of listening to spoken English as well. But then again, even spoken English varies wildly based on regional accents. It's all a big mess!
It's interesting that applying a statistical best guess would correlate with being easier to program, yet I still think this problem in Natural Language Processing is one that hasn't been solved at the lay-engineer level yet