HackStack

Project Info

ERNIE

Devpost

This project did not submit a demo video on Devpost.

Inspiration

Misinformation and bias in the media is extremely prevalent in todays world, where anyone can post and publish anything on the internet. There are many individuals who listen to and cite sources that are dishonest and manipulative without realizing, and we want to guide them to better information in a polite, friendly, and informative manner.

What it does

Our web app allows the user to enter a URL to a news article, processes it and selects multiple divisive quotes using Gemini API from the article. We run them through our trained model to develop a bias score, verify and adjust the score using Gemini API, and return these to the user in an easily consumable format. Additionally, we provide them with 10 alternate, less-biased articles on similar topics for them to read instead.

How we built it

We incorporated our current knowledge base to develop a full stack web app. Individually worked on the frontend, backend, and model training for the first half of the hackathon, before collaborating to integrate it fully into a working product.

Challenges we ran into

Figuring out how to train a model was difficult, as none of us had done that before. Working with Google Cloud Services was also new for us. Figuring out a work around for uploading our model to Github as it was too large

Accomplishments we're proud of

We managed to train our own model on a laptop using built-in GPUs. We have a product that surpasses our MVP goal. We set a viable scope for the project and completed it satisfactorily. This was the first hackathon for a few team members!

What we learned

How to train a model on pre-existing data using CUDA Learned to integrate Gemini API in unique ways other than just getting a prompt response.

What's next

Continue to train the model on a more extensive dataset that we weren't able to gain access to in this timeframe. Implement a google extension for this project, allowing for easier usability on the article page itself.

Analysis

Compare with all teams

View

Metric

Figures cover GitHub contributors during the hackathon window. A co-authored commit counts in full for each author, so per-member totals add up to more than the whole-team figures.

Technology

Found in codeClaimed only
  • CSSIn code
  • FlaskIn code
  • HTMLIn code
  • Hugging FaceIn code
  • JavaScriptIn code
  • Next.jsIn code
  • PythonIn code
  • PyTorchIn code
  • ReactIn code
  • Tailwind CSSIn code
  • TypeScriptIn code
  • Google GeminiClaimed

11 of 12 appear in the indexed code. 1 claimed on Devpost could not be matched to code, which may simply mean the tool leaves no trace in the repository.

AI coding agents

No AI coding agent signals were found in this repository.

Detected from committed agent config files and commit authorship. Absence of a signal is not proof an agent was unused.

Codebase size

Source size

81 KB

Source files

20

Counts recognized source files only; vendored directories, binaries and lockfiles are excluded, so this is smaller than the repository on disk.

0 stars