Transsum
The purpose of the application is to summarize and translate documents without a word limit. The most complicated part was managing the token size and dividing the string into chunks for GPT-3.5-turbo to handle, allowing for unlimited document size. After some research and help from ChatGPT, I solved this challenge.
The application does not upload documents to the server but converts them into a string that’s stored in memory. This design made the project easier to build and improved security by avoiding the storage of files on the server.
Currently, the app cannot adjust the summary length, a feature I plan to address by using a different model to make the summary length proportional to the document size. I also plan to add speech-to-text functionality and a simple frontend to allow users to adjust settings with buttons.
In the future, I’d like to make the app voice-activated, summarizing automatically when no voice is detected, though this will depend on the user’s microphone quality and background noise.
