Comments on Youtube videos hold a lot of insight and feedback valuable to the creator but spam spoils the insights. Text response and feedback holds a lot of value but it is hard to comprehend due to its size, but if we can condense it down to numbers and summarize it, it can be made useful
What it does
It either takes a youtube video URL and gives stats on the comments or takes a twitter profile and gives stats based on the tweets. It also helps to find out all the questions asked in the comment section
How we built it
We built the backed with flask and TensorFlow serve. We used a sentiment analysis model whose data pipeline also uses a spam filter. The model was built using tf 2.0 and was trained on the IMDB dataset.
Challenges we ran into
Setting TensorFlow serve for the first time. Learning about tf 2.0.
Accomplishments that we're proud of
The TensorFlow model and the intuitive UI.
What we learned
What's next for Textstat
- Comparison of stats with that of previous videos, tweets, and posts.
- Stats on facebook posts.
- Measuring amount of spam in the comments, tweets, and posts.
- Measuring toxicity in the comments, tweets, and posts.