Election Guard

Inp

Inspiration

We were inspired by the upcoming US election, and also, more locally, by the recent AMS election. We realize that there is a lot of disinformation online, and that may negatively affect one’s voting intentions. Keeping this in mind, we were inspired to create a platform to help people discern the underlying motive behind any published news article, including its political leaning and truthfulness. We also realize that people who speak other languages often receive news from articles written in their native languages, which also face the same threats of disinformation. Thus, we have decided to implement Spanish and French support into our application, making our platform more accessible and educative for all.

What it does

Our project uses machine learning technology to analyze news articles in order to determine its truthfulness, tone, how polarizing it is, bias, supportiveness of the subject, and political-leaning. It provides a short summary of the article, and highlights which parts of the article leans in what political direction. We store submitted news articles for analysis and compare the articles parsed in order to obtain data on the average bias/falsity of current day news. In addition, our application supports articles in English, Spanish, and French, and our interface is multilingual as well.

How we built it

front end react + Ant Design component library for a modern & easily navigational interface.
back end we integrated conhere api via prompt-engineering with cohere's classify & summary to analyze news bias, truthfulness, and political standpoint.
We used DeepL api for multilanguage support, including interfacing the entire website in supported languages.
MongoDB for data persistence, designed with consideration of inter-collection relationships for easy full-picture data retrieval

Challenges we ran into

Some challenges we ran into were deciding what exactly to implement, we made sure to prioritize features that would make up our MVP, and added extra functionality if time allowed it. During implementation, we mainly had issues with conflicting version control.

Accomplishments that we're proud of

We were able to integrate external APIs successfully, such as Cohere classify, and DeepL translator. We were able to fully link the frontend to the backend and create a functional app.

What we learned

Some of us learned more about the Git command during the project production, including resolving errors due to the same branch commits and solving merge conflicts. Our front-end developers also learned to use various components of React libraries to build and complete the project's front-end interface, such as using state change functions and props to change the displayed language for switching between languages from the dropdown menu on the project. We learned about the Cohere API and used it to train a classification model to determine fake vs real news.

What's next for Election Guard

If we had more time to work on this project, we would implement more options for languages to provide accessible information for more language backgrounds. We would also implement a "past inputs" dashboard to display previous entries that the user has made, or make it a universal platform where all entries can be seen by all users. We believe everyone should make informed decisions regarding the political scene, and should know whether they are subject to misinformation from the media. We would also implement a web scraper so that users could input any URL and receive the same statistics back, as well as a summary of the given piece of text.