SULLIVAN 2.0

Teaching it your way

Team SULLIVAN 2.0

. Neda Ashraf
. Rishika Raveendran
. Rosemary Benny
. Tessa Ann Josy

Inspiration

The name Sullivan2.0 was chosen as a token of respect to the greatest teacher of all times - Anne Sullivan who taught and nurtured Helen Keller into the beautiful lady she was. We are a group of students who have at many instances found ourselves surfing through the net for reference and on most days this has proven to be time-consuming. A video that embodies the right amount of information, sound clarity and visual representation was a hard to get factor. That’s when we thought of employing our skillset into creating the AI teacher ‘SULLIVAN 2.0’.

What it does

The project is basically a machine learning model for lecture video generation. The model is trained using the datasets which contain tens of thousands of publicly available videos on Youtube. The best and average videos are segregated by the model based on the number of views and likes for the videos. Now, from the best videos, the model analyses and learns the speed of speech, the modulation in the voice, usage of examples and the amount of theory dictated by the teachers among other factors. After the training is done, the model produces a video. This is how - The model is fed the videos on C programming from Youtube for training. Produces precise textual material on C programming (with appropriate examples and just the amount of required theory as learnt from the datasets) Translates text into speech considering the speed, modulation of audio (again from the analysis after training) Puts the textual material and speech together to generate a new video. This video inevitably has the best teaching practices as the model had learnt from the best and most popular videos on Youtube.

How we built it

First, through various brainstorming and online meetings we gave our idea a basic structure. Then we divided the work among each other, some dealt with the coding part, some writing the contents. We used the live share feature in vscode to work together. We made a presentation using google slides so that all could work on it and make necessary changes. We used Google Meet for brainstorming and meetings. Constant Communication made this project possible.

Challenges we ran into

As we were beginners, it did consume a whole lot of time to learn the basics of AI and ML technology. Even if the idea and its workflow are ready, it would still take up enough time, work and knowledge for the model to be functional.

Accomplishments that we're proud

We gained considerable knowledge of AI and ML technology in such a short span of time. For a few of our teammates, this was their first hackathon and the experience was great. We started off knowing nothing and now we sure have made ourselves proud by gaining some knowledge on machine learning and employing it in our project.

What we learned

The things we learnt were innumerable. We never thought that we could learn in such a short time. Firstly, we learnt the basics of artificial intelligence and machine learning and got to work on this project using ML technology. The experience of working remotely as a team was fascinating and gave a lot of insights.

What's next for SULLIVAN 2.0

We plan to take up it a notch by integrating interactive doubt clearance sessions and also by bringing AI-generated human models for visual appeal. We desire to bring our AI model’s teaching to be as close to real human as possible. Advancements in technology will surely help us to make this leap.