Skip to content

Latest commit

 

History

History
7 lines (6 loc) · 598 Bytes

README.md

File metadata and controls

7 lines (6 loc) · 598 Bytes

neural_ime with tf.keras

This is a deep learnig based model which convert Japanese hiragana into kanji and hiragana mixed. The original source code is written by @yoh_okuno and I modified the codes to work with tensorflow 1.12.0 and tf.keras. Credit: @yoh_okuno Original source code is here.

You can test this kanakanji_convert_4_train.ipynb to train a model and test the model with kanakanji_convert_4_check.ipynb. The datasets is wiki_dataset_mecab_80000.txt, which is a Japanese hiragana and kanji hiragana mixed pair originally from Wikipedia. uno