DKSplit – Fast Word Segmentation for Python
SMRTR summary
DKSplit is a fast, CPU-only Python library for splitting concatenated strings like domain names, hashtags, and slugs into readable words. Using a 9 MB BiLSTM-CRF ONNX model trained on LLM-labeled data, it achieves 91.5% accuracy — far outperforming WordSegment and WordNinja. It's brand-aware, multilingual, and processes ~800 samples per second with no GPU or API required.
SMRTR provides this summary for quick context. The original article belongs to Hacker News.
Read the original article