There was an interesting study (https://aclanthology.org/2021.sigtyp-1.1/) where they evaluated phonemicity of various language orthographies by training a neural net and then seeing how accurately it could predict things. Of two dozen languages they had there, the only ones that scored worse for writing are French and Chinese, but most notably, English is the only one that scored below 50% accuracy for reading, and with a significant gap at that. This is very unfortunate for an international language, since reading is kind of the most basic practical thing you can usually do with a second language.
- words spoken by toddlers: what's the spelling of a word that doesn't exist outside of a kid's brain ? In particular parents can accept it as a word without ever setting an associated writing.
- written words that don't have a pronounciation: typically Latin is dead and how any of it is pronounced is up to how we feel about it.
That's without going into words with phonems unrelated to their written form (XIV as fourteen for instance) and I assume there will be words that exchange spelling and pronounciation with others.
Languages are plenty weird, we should embrace their weirdity IMHO.
How words were pronounced can be deduced from poetry.
Of course. But it's not like we know nothing about it.
I hate that the past tense of "stay" is "stayed", but "say" is "said" and "pay" is "paid", which is often misspelled as "payed", which IS a word, but is unrelated to transferring money from one person to another.
Then you got all the ways "-ough" is pronounced. Thorough, enough, cough, through, thought, dough, drought..."-ough" is now looking like a completely nonsense letter sequence.
There's always a relation between a spoken word and its written representation, they're the same thing in different mediums.