Kitabi Keeda

As an NLP Engineer, having corpora is very vital. While protecting endangered and indigenous languages, having datasets of recorded language is also vital. This project combines these missions in one!

Comment

Inspiration

Working in a bilingualism research lab, one of the tasks our lab has taken on is protecting Quechua and other indigenous and endangered Latin-American languages. One of the ways we have done this was by creating data-sets to preserve documented language. Studying Natural Language Processing and Computational linguistics, I've seen additional value in these data sets. These corpora allow us to build tools such as machine translation softwares that allow for us to communicate better with minoritized communities. We wanted to streamline the process, and allow for users to document their own language as well as export the datasets for their own use.

What it does

Users can create an account, and then enter text for any of the languages supported. They can also filter text by language or difficulty, which helps language learners see example sentences of varying difficulty. They also have access to an Export tab, where they can export the data for any specified language as a CSV.

How we built it

This project was built with a frontend in pure HTML and CSS, with a backend developed using the Flask API in Python. All of our server side content is processed in python, making additions and queries to a SQL server where relevant.

Challenges we ran into

Initially, we planned to work in ASP.NET Core as one of our members has extensive experience using it. However, the current longtime release is not the version that they were trained in, and a lot had changed. It was also not the most intuitive for new members. While we were able to salvage CSS and HTML, we had to transition to a different API for today. Only one member had used flask, however all members knew Python so it was an OK transition.

Accomplishments that we're proud of

We are very proud of our quick querying for the dashboard, as well as being able to communicate with a database server side. We are also proud that we were able to host the project for submission so judges can view it.

What we learned

Each member learned a bit from it, Carmen had to transition to frontend development and work on CSS and HTML primarily, Bill had to develop skills with querying relational databases, and V had to work with compiling language data.

What's next for Kitabi Keeda

Ideally, we would like to add translations as a column for machine translation, and add stronger validation for user logins.

Share this project:

Updates