Graffiti wall

Social media for your current emotional needs!

Comment

Inspiration

Social media for current emotional needs. graffiti wall - you see tens of hundreds of things scribbled by random anonymous people .. and you feel connected .. to some of those scrawls .. and maybe you wish you could talk to that person?

What it does

Technologies used :- The project uses NLP, word cloud and cockroachdb.

  • It's primarily a social media for your CURRENT emotional needs.

  • Posts are short sentences comprised of a reason why a person feels a particular emotion at that instant.

*Every post has a lifeline of 24 hours, in which random people can comment on random scrawls(anonymously). Eg: you had a really bad day because you lost a competition, you come to this wall, maybe someone else went through exactly the same thing? Right then, no one could understand your mood or state of mind better.

*Completely anonymous platform - no question of gender, race, age, ethnicity, attractiveness quotient, etc. It's about pure feelings.

*5 basic emotions - ________________, each emotion is color coded. Sentiment analysis based on relative appearance of scrawls. So the wall will appear as this jumble of colors. A glance will tell the general mood of "the world" at that instant.(if sufficiently random samples generated) Rosy overtones to the wall? - its a happy day for most, Green - most people are pissed off!

  • NLP used to club similar scrawls together. Eg: "I got divorced", "I got separated" etc.

*The scrawls in the graffiti are attributed a portion of the wall based on their popularity. Not in terms of popularity of comments, but rather because of the fact that more number of people went through the same thing. Say a particular day oversaw LOTS of divorces, but not so many promotions.

How we built it

Transformers for NLP. Flask for interfacing the UI and the backend.

Challenges we ran into

Idea finalization took more than we expected, UI-UX difficulties, Could not generate sufficient data for a proper graph.

Accomplishments that we're proud of

That we were able to work continously without tiring, and were able to achieve a basic but workable model of what we had in mind.

What we learned

Learnt how to use transformers for calculating the sentence embeddings and clustering similar sentences together.

What's next for Graffiti wall

  • Graffiti fading off the wall over a period of 24 hours.
  • UI can be improved further.
  • Scalability - lazy loading of the graffiti, load handling, etc.
  • Storing user IPs in database - in case of harrasment issues, an IP can be blocked.

Built With

Share this project:

Updates