DKSplit – Fast Word Segmentation for Python

See also:

Instant Segment: fast English word segmentation in Rust –https://github.com/djc/instant-segment

Author here. Thanks for the submission.

DKSplit tags character boundaries with a BiLSTM-CRF and returns top-k candidates instead of forcing a single split — built for domain names: coinages, transliterations, multilingual strings, not just dictionary words. 9MB int8 ONNX, Python and Go, benchmark data and script in the repo.