CS&E Team Triumphs at Hackathon
A Department of Computer Science & Engineering (CS&E) hackathon team won the recent AI x Journalism Hackathon hosted by Hacks/Hackers, the Hubbard School of Journalism and Mass Communication, Poynter Institute, Trusting News, and the MacArthur Foundation. The hackathon invited journalists, technologists, civic leaders, and community members to come together to build AI-powered solutions that transform how Minnesota residents access and engage with local information, while centering important questions of ethics and equity in how these tools are used and deployed. CS&E master’s students Arunachalam Manikandan, Dipan Bag, and Erina Karati collaborated on the winning project, MinneDigest. Their project earned a $10,000 prize to support further development of their platform.
“The best part was meeting different people who work at different companies and startups throughout Minnesota, and getting to know the kind of work they do,” said Bag.
MinneDigest is an AI-powered news digest and podcast platform designed to make local Minnesota news more accessible, personalized, and engaging. It aggregates stories from local outlets, generates AI-powered podcasts, and supports multiple languages so users can catch up on the latest news in their preferred language. Manikandan, Bag, and Karati designed MinneDigest to retrieve the latest news from various local Minnesota websites, including MinnPost and MPR. Each time their program runs, it accesses these websites and gathers the most up to date news with properly summarized information.
“We want MinneDigest to be as personalized as possible,” said Karati. “Users should have control over their news preferences, audio formats, and even the style of news delivery. Ensuring news accuracy is one of our top priorities. We are working on ways to have our news feed verified by journalists, so that users can trust they are receiving credible, well-sourced information.”
“We’re digesting all the news we have collected,” said Manikandan. “We also use Perplexity to collect news. With all the sources we get, we digest it and convert it into a simple format, which has the title and content, and then we send it to summarization using ChatGPT. Once the news is summarized properly, we send it to the voice model, where we get the 30 second audio clips. Finally, we send all of today’s news articles to our custom-built module, which generates AI-powered podcasts.”
Their main source of inspiration was NotebookLM, which was developed by Google. A lesser-known tool, NotebookLM allows users to drop any PDF (such as a resume) or YouTube link, and Google will generate a podcast talking about whatever material you posted. Their approach uses ChatGPT and other large-language models to make podcasts featuring current news. Their second inspiration was to have a quick, on-the-go news podcast for people.
“We wanted 30 second audio clips where people can listen to them on the go, kind of like TikTok for the news,” said Manikandan. “These inspirations together helped us form MinneDigest.”
Their primary resource involved utilizing various platforms to conduct research about the project, such as Claude, ChatGPT, and Perplexity.
“Problem statements took some time, since journalism is an entirely new field for us that we didn’t have any previous exposure to,” said Manikandan. "Fortunately, the hackathon had experienced journalists share their problems, so we were able to consult with them while searching for ideas through ChatGPT and Perplexity. These elements helped us narrow down the ideas and propose which technologies would be best.”
After speaking with journalists, the team also added a multilingual audio support feature. Manikandan, Bag, and Karati are currently planning and brainstorming new features, such as customizable audio clips and adding more personas into the podcast. They are planning to add more user-facing customizations to make the news more personalized.
“We felt that MinneDigest could be something that would help marginalized or minority communities of Minnesota to stay up to date with the news,” said Bag.
“We believe in transparency and want to make MinneDigest open-source, so that anyone can see how it works,” said Karati. “While development will remain in-house for now, this approach ensures accountability and fosters trust in our platform.”
“This [grant] will help us add new features, boost user engagement, and collaborate with Minnesota news outlets to bring civic information to the community more effectively,” said Bag.
Moving forward, Manikandan, Bag, and Karati hope to develop MinneDigest into a much bigger app that would support more people. They are planning to partner with more local news outlets to integrate their news into their application.