Good News Hunter

Project Title: Good News Hunter.

Problem Statement: News nowadays is fed to us like fast food, most of it focused on shock value, often sensational, negative, or misleading for attention, while meaningful and uplifting stories are buried and overlooked, impacting people mental life and society’s perspective of world tremendously. This project aims to use AI to automatically find good news buried by the social media algorithm and share them with everyone.

AI Solution Overview: We address the challenge of filtering and presenting positive news stories from a sea of information. By leveraging the News API, we fetch articles from reputable sources and apply a two-stage AI filtering process.

Initially, a sentiment analysis model is conducted to gauge the positivity of each article. Subsequently, we use LLM (powered by Groq API) is employed to classify articles into categories such as "Cute/Fun," "Improvement," or "Heartwarming", with curated prompt.

Additionally, users can submit their own news stories, which are evaluated by the LLM to ensure they are safe and genuinely positive before being added to the feed. We automated the pipeline for every 15 minutes and deliver through an interactive website where users can browse, interact, and save their favorites.

Impact & Ethics: This is a project that can have impact in many fields. We live in a world where the news we see every day often fuels anxiety, sadness, and anger. Countless studies have shown that this constant exposure to negativity is a major contributor to declining mental health. But the truth is there’s no shortage of good news. Inspiring stories, acts of kindness, progress, and hope are happening all around us. They're just buried, ignored by algorithms, and overlooked by mainstream media. With the powerful LLM technology nowadays, we can easily use leverage them to uncover the hidden gem. By uncovering and amplifying these hidden moments of positivity, we can not only improve emotional well-being but also help reshape how people see the world.

Technologies Used: Programming Languages- python
Libraries and Frameworks-
Streamlit: For building the web application interface.
aiohttp: For asynchronous HTTP requests to fetch news articles.
Pydantic: For data validation and settings management.
TQDM: For progress bars in asynchronous operations.
SQLite: For local database storage of articles.
Poetry: For dependency management and packaging.

Deployment: Streamlit Community Cloud

AI Models and Tools
Transformers (Hugging Face): For sentiment analysis using pre-trained models.
GroqAPI: For calling the LLM that classifies articles and validates user submissions.
NewsAPI: The external API used to fetch news articles from various sources.

What we learned/Challenges Faced: Can we actually find good news under constraint?-
Since the free-tier news API only allows a maximum of 100 items per request, and I'm still capturing from major news outlets for maximum volume, I was afraid there would be little to no good news at first. However, there are actually plenty - not many, but enough to make me feel that this project and AI can really do good for the world, and that there are truly hidden gems buried around us.

Design Obstacle-
I actually saw this competition two days ago and decided to participate. The time was short, but I somehow managed to pull through. The challenges I faced were countless, including the pipeline not refreshing automatically, getting stuck due to errors, database issues with inserting, updating, and fetching data, and syncing the backend with the UI to ensure the results were displayed correctly. All of these took time to get right.

When I saw the final working website, everything felt worth it. It's definitely in the MVP stage and not perfect. There are still many areas that can be improved. But I'm satisfied with what I built in a short time and how it actually made me feel happier and more content. I hope it brings you joy too.

Built With

• ai-applied-sentiment-analysis
• api
• grok
• llm
• poetry
• python
• sqlite
• streamlit
• transformers