Inspiration
Yup’ik is an Eskimo language for which few resources exist. It is an endangered language from western and south-central Alaska with less than 20,000 native speakers left.
Yuarcuun means "searching tool" in Yup'ik. We want to build an app that anyone can use as a translator and dictionary, to encourage the language use and development.
What it does
In order to start a machine translation toolkit English<->Yup'ik we parsed an open-access dictionary and grammar for Yup'ik language. Since this language is polysynthetic and very rule-based, we are writing a parser for the language based on its grammar rules with two aims:
- better results than unsupervised tokenization for the machine translation scheme using deep learning,
- displaying informations in the translation/dictionary tool about root/postbases/endings and their meanings. This feature could be very helpful for language speakers and learners.
We plan to release the toolkit as a phone app Yuarcuun. As a first step we created a dictionary app using our parsed dictionary.
How we built it
Using Python and Tensorflow to leverage cutting-edge deep learning for the machine translation task. We built the app using React Native.
Challenges we ran into
Yup'ik is a low resource language, thus our dataset is not as big as usual machine translation datasets are.
What's next for Yup'ik Machine Translation
As soon as the toolkit is complete, release it to encourage young native people in Alaska to use and keep learning Yup'ik Eskimo.
Team Members
Christopher Liu, Kevin Chavez, Laura Domine
Built With
- python
- react-native
- tensorflow
Log in or sign up for Devpost to join the conversation.