Recent events such as the Black Lives Matter Movement and the COVID-19 pandemic have made it especially clear that people struggling with mental health are on the rise. Despite the importance of being informed, many people are now being overwhelmed with information. Social media and news platforms are constantly updating with the latest global and local reports -- factual or not -- and many people are up-to-date, but often at the expense of their mental health.
What it does
Truth-tweeting and sentiment analysis aim to address these issues of mental health, information overload, and reliability of tweets through our web application that lets users decide what topics they want to see, and providing them with true-or-false-indicated content and a sentiment analysis. This alleviates the stress of information overload by streamlining user-desired content, saving users the time and resources from fact-checking the information they see, and preserves their mental health by providing a sentiment analysis that lets them know the tone of tweets beforehand (e.g. in case they are not interested in reading a sad or fearful tweet before bed). Relating this to Twitter Principles, we deliver #trust-worthy, #fast, and #straightforward information to the users and improve overall platform-wide mental #health.
How we built it
Built using a Flask server backend with a React frontend. The react frontend serves to handle inputs and parse through the user queries, and also allows the users to get interesting trends near them. From there, the flask server handles all of the data manipulation, processes the text to determine the emotion, and attempts to determine the corresponding truth value associated with the tweet (especially if it had a url to an article)
Challenges we ran into
Building with API's, connecting flask backend to React frontend, understanding and doing data processing
What we learned
Nowadays, information is everywhere and it comes in a massive way. People read a lot of different tweets with many types of content, some are funny, some are rude, others are irrelevant and a lot of times they are not completely true. In times like these, the constant news coverage updates overwhelms people creating a serious mental health concern. With Truth Tweeting and Sentiment Analysis, the user will be able to decide which tweets to see based on specific hashtags or topics and also know whether they have fake content or not. Depending on the mood the user is having, they can also also decide if they want to see or not tweets categorized by mood, such as happy, sad, angry, graceful, etc.
What's next for our app?
During this time that we were able to work on the feature, there are small details that are yet to be completed. We see that there are areas of opportunity where we can improve for a better overall experience. For example, incorporating the dashboard connection which is currently not dynamic. Also the machine learning model in use is working fine but there are definitely ways to refine it. In terms of UI and user experience we are so open to hear users' feedback about all aspects in which we can improve the feature so we can give the best experience.
Backend Link: https://github.com/zhaovan/twitter-hackathon