Inspiration
Our inspiration for VeriTrue came from our own struggles with reading long articles and research papers. We often found ourselves overwhelmed by the sheer volume of information and wished there was an easier way to extract the key points. That’s when we decided to explore AI models like BERT and BART on Hugging Face. We wanted to create a tool that makes summarization effortless and helps users save time.
What it does
VeriTrue is a Chrome extension designed to help users simplify lengthy content. It takes text inputs, Word documents, or .txt files and generates concise summaries. Currently, we support these file types, but we plan to expand to PDFs and other formats in the future. With VeriTrue, users can focus on the most important information without wading through unnecessary details.
How we built it
To build VeriTrue, we used HTML and CSS to create a clean, user-friendly interface. The backend, written in JavaScript, interacts with Hugging Face’s Inference API, where we integrated the BART-Large-CNN model fine-tuned for summarization. This allowed us to leverage pre-trained AI capabilities while ensuring seamless functionality in our extension. The development process involved multiple rounds of testing and debugging to fine-tune the user experience.
Challenges we ran into
While building VeriTrue, we faced several challenges. First, the Hugging Face API had rate limits, and GPT APIs were too costly for our project. We also encountered issues with summarizing very short texts using the BART-Large-CNN model, which sometimes led to incoherent outputs. Another major hurdle was integrating PDFs. Libraries like pdf.js had limitations with client-side rendering, and attempts to use Python or NPM packages didn’t align with our browser-only environment. Despite these constraints, we persevered and worked within free resources to deliver a functional extension.
Accomplishments that we're proud of
As students who have just started out on exploring the topic of AI and Analytics, we felt a sense of accomplishment just by having the chance to be able to use and experience with AI models, especially those that are free of charge. This helps us with our learning as we continue to expand our knowledge on the topic of AI.
What we learned
We learned about the different AI models and each of their functions as well as how to create a Google Chrome extension as this is the first extension that we have created. Having chosen the BERT AI model as the desired model for our extension, we also learnt how the model works as well as how to use it to summarize long text content into simpler sentences.
What's next for VeriTrue?
Making use of valuable feedback from every user of our extension, we will continue to implement more changes and improvements to VeriTrue, ensuring that it stays competitive with the other extensions in the market. Our public repository for the extension will continue to be open for people to make changes should they have any suggestions and improvements to the codes in mind.
Log in or sign up for Devpost to join the conversation.