A rule‑based parser for Chinese regional languages (Sinitic languages).
NLP tools for Chinese regional languages remain scarce, which limits the development of downstream language applications.
While a purely rule‑based parser cannot achieve production‑grade accuracy due to inherent ambiguities that rules alone cannot resolve, it can serve as a bootstrap foundation for machine learning and deep learning parsers.
The current stage of development is focused on the Taigi (Taiwanese Hokkien) lexer/tokenizer projects. The main objective is to design a rule‑based lexer that can consistently handle unambiguous words. In this initial phase, particular attention is placed on grammatical markers and high‑frequency vocabulary, ensuring that the foundation of the system is both stable and practical for real usage.
Work on the Hakka lexer/tokenizer continues, but regional distinctions are not implemented at this stage due to insufficient data. Lexical entries will be restructured once more comprehensive sources become available.
Mandarin lexer/tokenizer has been added as a scaffold for a Sinitic language NLP codebase. Further development is not planned in the near future.
- OCaml
4.14+(recommended) - ocaml-clean-compile — compile OCaml code without Dune
MIT License © 2026 ByteBard