As you can probably tell from my profile picture, I am quite fond of Japan. However, I cannot actually speak, read, or write the language, regardless of whether it is Hiragana, Katakana, or Kanji. As such, I have created a Seq2Seq model for translating basic English phrases to Japanese.
The utilized dataset is provided, with others able to be found here: http://www.manythings.org/anki/
GloVe word embeddings: https://nlp.stanford.edu/projects/glove/
For future improvements, a larger dataset would allow for custom phrases to be translated, but would also require more RAM than is available in Colaboratory. An attention layer could also be implemented to improve the accuracy of translations.