A generic and language-independent Algorithm for Typo-Recognition

This Algorithm calculates Typo candidates using text corpus statistics based in Zipf's Law. The assumption is that any word that appears only once in the the text corpus is a typo candidate if there exists a more frequent word with an edit-distance of 1. # A generic algorithm for language independent typo recognition threshold = 100min_word_length = 5 single_use_words = SELECT word WHERE global_word_frequency = 1 and LEN(word) > min_word_lengthall_words = SELECT word WHERE global_word_frequency > threshold and LEN(word) > min_word_length found_typos = {} foreach hapax_legomenon in single_use_words: foreach word in all_words: if edit_distance ( word , hapax_legomenon ) > 1 : found_typos.append ( hapax_legomenon ) # found_typos contains generically calculated language independent typo canididates.