Compendium

Summarizes given text or hyperlink and provides relevant links and resources.

Inspiration

In an era where AI is becoming increasingly prevalent, it’s essential to avoid over-reliance on technology. As such, we were inspired to create a tool that allows for both efficient data summarization and critical thinking. By combining AI's capability to summarize and highlight key information with the exploration of recent, relevant news, we encourage users to engage with their material more thoughtfully.

Functionality

Our tool delivers concise key points and summaries while also providing related resources by using Claude AI and our web-scraping program. We suggest recent news articles and supply an extensive database of research papers, allowing them to dive deeper into topics and develop further understanding.

Users can make use of our tool by summarizing things like notes or long readings. This will allow for them to quickly digest what would otherwise be a long read and dive into more fleshed-out articles and papers.

Building Process

We started by playing around with generative AI to understand its capabilities. Then, we focused on using it to generate keywords from a text block and connected it to our idea of scraping the web for relevant information. Next, we created a front-end page where users could input text and quickly get relevant information. In the end, we incorporated newsAPI and arxivAPI, Claude AI, and a front-end UI to create a nice page to get quick information and other relevant resources.

Challenges

We had trouble with Flask while attempting to connect our frontend and backend programs. We attempted to incorporate React, but similar to Flask, had trouble and ended up using basic CSS, HTML, and JavaScript. We had trouble parsing through generated messages and figuring out which portions should be left as either plain text or links. We spent a considerable amount of time figuring out the best inference parameters and prompt messages to generate the most accurate results.

Accomplishments

We are proud of finishing the project! We were proud of being able to use the generative AI to not only provide us with information but also to successfully parse the provided messages. We are also proud of the work we did with connecting the front and back ends to successfully work together and output nice results.

Learning Points

We learned how to use Flask to connect our front and back end programs. We also learned how to incorporate APIs into our programs and provide outputs from user input. We expanded our knowledge of CSS and HTML to create a nice-looking page.

What's Next for Compendium

We will look to expand the sources we have access to, instead of only having more recent ones. We can do this by improving the related source accuracy by using vector comparison and taking into account the frequency of word appearances. Relevance is not based only on timeliness but also on accuracy, which is important for us to consider when suggesting outside resources.