Ai Summarization Service

A handy summarizer to speed up search and comprehension of new information.MVP: Twitter bot that can get tagged under a tweet, to reply with a summary of the text, video or audio.

Comment

SLIDE DECK https://docs.google.com/presentation/d/1z80xeymzJUvLw5Kum9iW95eusYFxmKYERrh2kkJXzQM/edit?usp=sharing

CODE: https://github.com/kon-rad/ai-summarization-service

Inspiration

A friend of mine recently commented about the need for a service to summarize long videos, for example there was a tweet that contained a 30 minute long keynote presentation at a conference. This seemed like a really good problem to tack for this hackathon.

What it does

This is a twitter account that when it get's mentioned with an '@' under a tweet or thread, replies with a summary of the content in the tweet.

How we built it

In this hackathon I came in with a fresh perspective and researched the best ways to summarize content using Ai. I found that there are many ways to do this, identified the best architecture.

  1. If content is greater than context window, use Map Reduce Chain Langchain functions to create a summary. This splits the text into chunks that fit in the context window, summarize each chunk, then create a summary of all the summaries. 2 prompts. requests done in parallel.
  2. Else, use Stuffing Langchain function.

Challenges we ran into

Time was very tight in order to research and implement some code to prove this could work. In the limited time I was able to find the state of the art techniques for doing this and to implement a proof of concept.

Accomplishments that we're proud of

Completed research Completed presentation slide deck Created GitHub repo with template [Almost] Generating a summary of the text bible using my code.

What we learned

Many ways to do summaries, there are many apps that provide this service, yet it was discovered that there is still a long way to go before we have tools that we use everyday to provide us summaries on al the content that we come across everyday. The potential is massive for improving our productivity if we have a handy tool that is with us to quickly summarize large content.

What's next for Ai Summarization Service

Build out the MVP prototype and unleash it on the world.

Built With

Share this project:

Updates