RapidRead

• 

  Initial screen

• 

  After submitting a file

• 

  Final output (clipped)

Inspiration

I was frustrated by having to read long beginner unfriendly articles whenever I wanted to learn anything new. I specifically chose to use AI for this project because no other software is able to do something like this. This project requires the ability to understand and write text so naturally, AI is the only option in this scenario. Additionally, I am a first time builder participating in my first real hackathon who learned about implementing AI only a few weeks ago so I wanted to try and create something new with AI models.

What it does

Input a pdf file into the program and it will extract the text and summarize it for the user so it is easier to read.

How we built it

I began my implementing some frontend infrastructure to allow users to upload a file. Then I used the python library fitz to collect all of the text on the document. I researched about different AI models and chose one suitable for this task. I implemented the model into the program and accommodated for its limitations Once the core functions worked, I spent some time cleaning up the frontend to enhance user experience.

Challenges we ran into

Choosing an AI model was extremely difficult as the vast majority of quality models are paid. Furthermore, many of the free ones often give extremely poor responses and cannot handle being given the context for this particular task so picking a model that worked toa satisfactory level was one of the biggest challenges. In addition, many of the python libraries are rather hard to debug, especially concerning the AI models.

Accomplishments that we're proud of

I was able to finish this prototype from start to finish within an extremely short time span due to conflicting events during the hackathon I'm also proud that the model was implemented without much issue since my previous attempts with AI did not work out nearly a smoothly

What we learned

I learned how to properly incorporate and call an AI model in python and learned about how they work with tokens as the input/output

What's next for RapidRead

If possible, finding a better model as even the current one had its own faults Improving the frontend is another next step as it is not styled much at all

Built With

• ai
• fitz
• python
• streamlit
• transformers