english orthography is extremely regular at the word level too; there are only a handful of words like 'desert' which have multiple pronunciations for the same spelling. but, thus considered, the number of rules is extremely large, perhaps several thousand: a hundred or so for the words that do have regular orthography at the letter level, plus one for each exception: 'would', 'one', 'who', 'police', etc. there's a pretty good list in espeak-ng-1.51+dfsg/dictsource/en_list and espeak-ng-1.51+dfsg/dictsource/en_rules. en_list has about 5000 rules in it, things like
Tycho taIkoU
aperitif a#pEr@t'i:f
(wouldn't have to) ,wUd@-ntavt@5 $verbf $strend2
while the 7000 or so rules in en_rules are letter-level things like
_d) am (i eIm
und) am (en @m
_) am (en a#m
which specify how to pronounce a particular combination of letters in a particular context. i suspect you could probably reduce the 7000 down to 100 with a more expressive system; the schwa-reduction of [a] in "fundamental", the second example above from en_rules, is the same schwa-reduction applied everywhere to english low vowels in unstressed syllables
by contrast, es_rules (for Spanish) is 285 lines long, and then there are 391 lines of exceptions (mostly things like "thunderbird" and "software") in es_list
in chinese, instead of several thousand rules, you have several tens of thousands of rules, one for each character (which are not in a one-for-one correspondence with words but are roughly as numerous)