word_forms

Why we included this project

Most NLP tooling reduces a word to a single base form, which is fine until you need to match 'run' with 'ran' or treat 'politician' and 'politics' as the same thing. word_forms takes the opposite approach: one function returns every related surface form, so 'president' comes back with 'presidential', 'presidency', and the verb 'preside' all at once. The result is a plain dictionary keyed by part of speech, and a similarity threshold lets you tune how far the library reaches. That makes it handy for search normalization, text analytics, or any preprocessing step where you want related words to count as one concept without writing your own morphology rules.