Inspiration
We often use online lectures when working on assignments, reviewing course concepts, and studying for exams. The problem is that a lot of the time we have whole playlists with over 20 videos but the titles just say "Lecture 2" or "Episode 10", etc. and you have to go through about 5 videos to find the one that actually contains the information you are looking for. There has to be a more efficient way to get to this information.
What it does
GoHere is an app that finds the exact video and time in the video that relates to your query about a course. It searches an entire YouTube playlist to find the best match for your information needs.
How we built it
GoHere takes the URL to a youtube playlist, then transcribes and summarizes each video in the playlist. The summaries and a user query are then passed to the Cohere API which searches the Youtube video summaries for the most relevant one. Once a top result is found, the relevant youtube video is opened in your browser at the approximate time that relates to your query.
Challenges we ran into
Summarizing the videos took a long time, and we had to decide on the summary length that was optimal both in terms of information/context provided and for generation.
Accomplishments that we're proud of
- understanding and applying Cohere's API
What we learned
- Cohere's Summary, ChatBot API
- RAG - Relative Augmented Generation
What's next for GoHere
- User inputs syllabus or practice problems, GoHere directs user to the list of videos to watch
- Creating a user interface
Built With
- api
- cohere
- python
Log in or sign up for Devpost to join the conversation.