We wanted to make it easier for those trying to learn languages to immerse themselves in real world content while also providing them with tools to improve their knowledge and promote ongoing learning.
What it does
The website aims to use machine learning in conjunction with a web crawler to continuously collect news content in a variety of languages. The user can then select a language to learn and begin engaging with real world content. While reading the articles they can call up real-time translations and store challenging phrases to work on later. Should they choose to return to their saved phrases they would be able to quiz themselves on the correct translations.
How we built it
We attempted to use a trainable neural network fed by a web-crawler to collect content. We then parsed this content in to a database and built a user front-end to present the articles to our users.
Challenges we ran into
Having succeeded at projects within our comfort area at past hackathons, our teams goal at this hackathon was to maximize the challenge we presented to ourself in the hope of learning new skills. Unsatisfied with the multiligualism of news aggregators that we were familiar with, we set to make a spider to discover articles on a list of international news websites. News websites are similar-looking — frustratingly so because their structure to a crawler can differ significantly, requiring some extra sophistication to correctly classify article pages from non-articles, and grabbing the article text to be hosted on the Multilingo servers. Interestingly, we learned that this problem is largely solved by DiffBot and like services with cutting-edge machine learning techniques. The crawler then became more of a curiosity for the group as we tried to see how long it could continue to traverse the internet. While we were able to complete a functional project at this hackathon, unlike at past events, we were able to focus on trying new things, which at the end of the day maximized our enjoyment and learning at this event.
Accomplishments that we're proud of
We were happy to get some basic implementations of Tensor Flow functioning, create a well designed home page with some new special effects, and enjoy the hackathon experience.
What we learned
We certainly learned a lot about new server methodologies with respect to collaborating and deploying with git, and connecting a variety of web frameworks like express.js and apache on one server.
What's next for Multilingo (Multilingo.me)
We would like to spend some more time on the upfront challenge of collecting and storing the news articles on an ongoing basis which will allow us to provide up to date content to our users.