When I used to be in middle school decades ago, I discovered what wonderful tool an English dictionary is. My first dictionary was an Oxford English Dictionary that I inherited from my father, originally published in the 1960s. It had the words, some funny looking alphabet which I later understood was the Phonetic Alphabet, various meanings, synonyms, antonyms and examples of use in sentences.
For someone who is intending to build a pronunciation checker, I would start with the words and their International Phonetic Alphabet (IPA) equivalent. With today's text-to-speech technology and machine learning algorithms, going from voice to IPA and vice versa should be quick.
Once you have that for words, you could build on it to add evaluation of vocabulary, grammar, sentences etc., A lot of it is around assembling the data and collating training data sets, imo.
Cool project!