The problem UTube Channel Analyser solves

UTube Channel Analyser helps content creators get insights into their Youtube Channel by analysing the Sentiment of the comments on their Youtube Videos using Natural Language Processing powered by Bi-Directional LSTM with more than 500000 trainable parameters. The Portal is designed in such a way that it offers a smooth user experience.

The results are easy to comprehend and can be used in machines with access to the internet, thus making the solution accessible for small content creators. We also talked to a few content creators, and they found the platform helpful, and they sought to improve their content based on the sentiment analysis of the comments.

Challenges we ran into

The biggest computational challenge was training the Deep Learning Bi-Directional LSTM Model with limited resources as we do not have access to GPUs. Collecting a proper dataset for training was also an issue. We overcame the computational challenges by using Google Colab which though provided us with GPU but had a very short runtime so we had to run the Notebook multiple times. We have ideas to handle emojis and emoticons but we couldn't implement the same due to the aforementioned issues which stopped us from increasing our model's accuracy.

The final challenge and the most frustrating one was deploying the backend solution as we used Tensorflow the overall size of the solution was quite large(around 600MB) and hence we were not able to get it down to the size which can be hosted on free on Heroku. The challenge remains unsolved due to the lack of resources but we have deployed the backend partially with the Tensorflow Model which comes under the free hosting limit. To use the complete version of our solution, install the backend and the frontend on your local machine from the repos provided.

To check the Backend:-

1. To see the Youtube Videos fetched after entering the URL of a channel
   https://utube-backend-ml-leopard.herokuapp.com/videos?url={url}
   for example, https://utube-backend-ml-leopard.herokuapp.com/videos?url=
2. To see the comments(without sentiments) for a particular video
   https://utube-backend-ml-leopard.herokuapp.com/comments?id={id}
   for example, https://utube-backend-ml-leopard.herokuapp.com/comments?id=ba-HMvDn_vU