HKIME

HKIME is an on-going project that aims to create an Input Method Editor for Cantonese. Progress is documented in the Jupyter Notebooks located under the directory notebooks.

Our latest milestone is a simple bi-gram based statistical language model with no smoothing from a corpus made from scraping Cantonese Wikipedia. We are currently working on training a NN-based language model.

Relevant papers and sources that have tremendously aided our progress are in resources.md and sources.md.

Core Components of the Project

All of these are in progress at the moment

• Fuzzy Jyutping / Processing different romanizations
• Jyutping Segmentation
• Jyut2Char: Jyutping to Character Conversion
• Scraping / Corpus Generation
• Neural Net Language Model