The inspiration was a way to have an overhead view on what the sentiment of comments are on any YouTube video.
What it does
This application and api takes a YouTube video url, it then uses the YouTube api to pull all the comments from the video. It then uses pytorch through the library flair to run a sentiment analysis on all the comments. It then shows the results back to the user in a table.
How we built it
This application was built with Flask, the flair library that built on top of pytorch and using the YouTube api.
Challenges we ran into
A challenge that I faced was deploying the application to Heroku for testing. The application requires downloads of particular data models that when unzipped exceed the size limit for a free application on Heroku. I plan on trying to deploy this on a digital ocean droplet or other deployment site.
Accomplishments that we're proud of
I'm proud of getting the application up and running correctly.
What we learned
- I learned how to clean text data before running a sentiment analysis on them.
- I also learned how to use the flair library.
What's next for Youtube comment analyser
The next steps are:
- To get the csv download of results feature complete
- Ability to add a playlist or multiple video urls
- Increase the number of comments that the user is able to pull from a YouTube video.
- Deploy the application