Lumos: A light for all voices

Lumos is a web application that identifies unique voices to give learners and educators information to formulate personal opinions without the influence or bias of modern search engines.

Comment

Story

Inspiration

We were inspired to make this product by the proliferation of AI-generated content that is now flooding the internet. We felt like as AI gets better and better, this problem will only get worse as companies seek to push as much generic content out as possible in order to gain their attention. This also drowns out individual authors pushing their own unique perspectives on complex, nuanced issues.

What it does

Our application supplements Google by adding unto the excellent search engine they've built with our grouping algorithms that automatically detect when a group of articles is conveying the same point and meaning. Our app will group these articles together, helping the user find more unique voices and viewpoints rather than the same mainstream opinions that every media company puts out.

How we built it

We used react for the front end and node.js for the backend. In addition, we utilized python libraries such as sci-kit learn in order to create our grouping algorithm. We also relied on several react css libraries to make the user experiences as smooth as possible.

Challenges we ran into

In general, we had to do a lot of research to optimize our grouping algorithm so that it ran as efficiently and smoothly as possible. However, our biggest challenge was ensuring that the user experience was the best ever. We knew that user experience contributes a lot to how an user perceives an app and its information, and since our app is designed to serve such an important and vital role in people’s lives, we knew that our user experience needed to be the best as well. Through many trials and errors, however, we managed to develop a smooth user experience.

Accomplishments that we're proud of

We developed a very valuable grouping algorithm that tokenized articles and grouped them. We also developed a world-class front-end that provides smooth user experience.

What we learned

Through this project, we greatly improved our frontend and data science skills. We learned how to create a clean, user-focused interface in react. Additionally, we learned how to use a modified K-Means Clustering algorithm, with K selected through the Elbow Method, to group data based on viewpoint. On top of this, we learned how to use the Google, Bing, and Yahoo APIs to power our customizable search engine. In addition, we learned a lot about working with other programmers in the same project. Most software cannot be built by a single person, and having a system that programmers can easily collaborate in is extremely important. We are all happy to have been given the opportunity to learn about one of those systems in popular use(git, github) at this hackathon.

What's next for Lumos: A light for all voices

In the future, we will refine the credibility score algorithm by training our model with more online databases of credible sources. Additionally, we plan to handle more results. Realistically, it would be possible to search and group around 1000 web pages.

Share this project:

Updates