-
This is the home page of the main project. There is the option to view the mood once and see the overall graph of a single account.
-
This is one of the moody tweets landing page.
-
Here is what it looks like to open the page.
Inspiration
After looking through several projects, our team landed on the idea of detecting mood from a Twitter account. When expanding on the project as we worked on it, we found that the many uses of mood detection on Twitter accounts is useful in identifying mental health issues. We understood that to tackle this issue, we needed a heavy stepping stone into analyzing tweets as data. We chose to use GCP for its machine learning and vision APIs so that our development could progress smoothly. Later in the project, we wanted to offer a graphing service of a user's tracked mood. To do this, we used FireStore to graph out the mood shifts of a user account as they were searched. The hopes of this is to supply concerned users of our application the ability to see the current status of a twitter account's mood while mapping it to a graph they can use later.
What it does
It takes a twitter handle name and uses Machine learning to give you a current mood reading of the twitter account.
How we built it
We created an app engine nodejs application to which we created cloud function to make requests to the Machine Learning Api's to get mood readings. All user mood data is being used and stored in firestore. We also bought a domain on domain.com to use in GCP's web-hosting service.
Challenges we ran into
In UI/UX, we had problems regarding the implementing a loading page where it either stayed in the screen for the entire time or it didn't appear at all. We also had problems in creating an input field for the user to enter a twitter handle to use our website. There's also the issue of resizing our website and it doesn't implement well on certain screens. We also struggled with linking our CSS files our corresponding HTML files. In setting up our domain, we ran into issues bringing our website up and running because of restricted administrator privileges. Firestore API, app engine environmental variables, Google Cloud functions creation, Domain.com SSL verification were all obstacles that we overcame. During the Machine learning stage of the project, data classification became an issue since fitting that data did not always equal similar associations.
Accomplishments that we're proud of
Having a fully functional finished project is something the MoodyTweets team is proud of.
What we learned
Taking smaller steps in Machine learning is useful before you start bigger data analysis. In adding, keeping an eye on your requests per second is important to do since you can exceed you limit on GCP if you are not careful. The front end connects to back end more often than anticipated. The biggest lesson we all learned however, was that adapting to change takes a team effort by delegating tasks smartly and by giving a team member the right problem to solve.
What's next for MoodyTweets
Spreading the idea of analyzing mental health in more than just social media by preventing accidents outside of the virtual space.
Log in or sign up for Devpost to join the conversation.