JamSpell

Why we included this project

Most spell checkers look at each word on its own, so they miss errors that only make sense in context. JamSpell instead uses n-gram language models trained on large sentence corpora, letting it pick the correction that fits the surrounding words rather than the most common spelling. That approach fixes noticeably more errors than dictionary-based tools like Hunspell, and at roughly five thousand words per second it can run inline in a request path without slowing things down. The C++ core ships with Python bindings and a small HTTP server, so you can call it from a script or plug it into an existing service. You can also train your own model on domain-specific text, which helps when your vocabulary differs from general web content.