YouTube Sentiment Analyzer

Summarize uneccessarily long video reviews using the co:here NLP sentiment analysis.

Comment

Inspiration

As YouTube consumers, we find it extremely frustrating to deal with unnecessarily long product reviews that deliver numerous midroll ads rather than providing useful and convenient information. Inspired, we wanted to create an application that stops all that by allowing users to glean the important information from a video without wasting time.

What It Does

  • Determines whether a YouTube product review is positive or negative.
  • Summarizes the contents of the review.

How We Built It

  • The auto-generated closed captions for a YouTube video are obtained
  • Punctuation is added to YouTube auto-generated closed captions.
  • The punctuated script retrieved is then fed into Cohere, an NLP toolkit.
  • Cohere performs an analysis of the script and returns sentiment and summary.
  • Chrome extension is then utilized to display the information to the user.

Challenges We Ran Into

  • Settling on a use of Cohere.
  • Punctuating the auto-generated YouTube closed captions.

Accomplishments That We're Proud Of

  • Deftly transitioning between ideas.
  • Combining many new technologies.
  • Figuring out the puntuator2 library.

What We Learned

  • How to develop chrome extensions.
  • How to utilize Cohere to perform analysis and process summaries.

What's Next For The Project

  • Implement a mobile app component that allows the user's to make use of the service at the convenience of their phone.
  • Keep track of users' past history and engagement.
  • Incorporate a points system that encourages and drives individuals to continue to make use of the application.
Share this project:

Updates