No video of the event yet, sorry!

Currently, ibus-mozc is used as the default kana-kanji conversion method in openSUSE. However, although the mozc dictionary has been updated to the latest version, the conversion algorithm has not been updated for over 10 years. Currently, LLM is the mainstream in NLP, so shouldn't OSS kana-kanji conversion methods follow suit? Shouldn't we explore new kana-kanji conversion methods that utilize deep learning and word embedding?

I am currently implementing a completely new kana-kanji conversion method that I have named ANYA. I will explain the following points that I am working on in the process.

• Collecting and preserving copus using open data
• Creating an OSS Kana-Kanji conversion dictionary using word embedding
• Implementation of conversion algorithm using LLM
• Development of input methods that run on ibus on openSUSE
• Future operation plans using federated learning etc.