What if you had a tool so you can peek into a youtube video and analyze what it is said without watching it? YTTFJS does that!

What it does

It gets the transcription of the Youtube video, displays the text, the Toxicity Classification results and a word cloud of the text.

How I built it

A Flask backend with javascript functions to handle the predictions and displya of the results. NLTK and Wordcloud libraries to plot the wordcloud. All deployed on Heroku.

Challenges I ran into

I wanted to handle the predictions on the backed and wanted to convert the tfjs model into a graph model I also asked a question and still would love an answer, so..please Tensorflow? a little help :D ?

Accomplishments that I'm proud of

Finishing on time! I would like to keep working with Transcriptions, NLP and Tensorflow models.

My concerns for responsible AI are here: PDF

What I learned

Lots of new tricks on Javascript and Python!

What's next for YTTFJS

So this is a prototype, a lots of things can be done from here, I will work on further visualizations and improvements to the backend in the near future.

Share this project: