This project was built in collaboration with Harini Narayanan and Michael Murillo-Martinez.
Inspiration for this project:
Expression is one of the first steps of mental health and well being. A common fear among people posting on social media is that someone will have a strong negative reaction to something you share.These comments have a negative impact on the mood of the user. What if there are negative comments on a user posting about mental health problems. It may push them over the edge.
What it does:
The tool finds tweets about Mental Health and collects replies for each tweet. If the tweet has an overall of negative replies, a chatbox replies to the user tweet with a positive message to help them feel better.
How we built it:
We used Python to scrape tweets and replies and used NLP neural network model to predict the negativity of the replies with their confidence values. Then, we used selenium to create a bot that can navigate the tweets with negativity and can reply with a positive message to that tweet.
What we learned:
We learned that there are some tweets where people express negativity and that needs to be addressed. We learnt a ton of new libraries, packages and tools such as Deepnote and Selenium.
What's next for Tweety:
- Generalizing the tool to study the user tweet replies on any topics.
- Expanding it to other social media sites.
- Develop a virtual chatbot that can reply positively to the tweets of users with more negativity.
My contributions:
- I worked on scraping the text of the replies of each tweet, aggregating the data, data cleaning and preprocessing.
- I also implemented a sentiment analysis using NLP neural network to predict the negativity scores of the replies of each post.
Below is the video explaining the whole project and also a demo of our tool: