An N-gram model specifically trained on USDA list of foods to de-abbreviate grocery items, specifically the way they are worded on receipts. Code was utilized in thesis project for class CP499. Heavily inspired by @weirdMath's abbreviation spellchecker for english.
After running word_abb.py, user must plug in abbreaviated word inside alongside language models to get predicted word with the accuracy percentage.
noisy_channel('cffe', big_lang_m, big_err_m)
{'coffee': 9.411900215724993}