fugashi

Why we included this project

Japanese text doesn't use spaces, so any serious NLP pipeline has to split it into words before doing much else. fugashi wraps MeCab, a widely used Japanese tokenizer, behind a Python API that returns lemma, part-of-speech, and other UniDic fields as named tuples, so you don't have to parse feature strings yourself. It ships prebuilt wheels for Linux, macOS, and Windows, and dictionary installs come down to a pip extra, which removes most of the usual setup pain. You can also point it at any MeCab dictionary, not just UniDic, and it's fast enough for production use. That makes it a practical base for search indexing, sentiment analysis, or anything that needs Japanese split into meaningful units.