What inspired us to build it?

Watching Logic speak about different societal issues during his VMA performance made us realize that mental health is an issue that needs to be talked about more. There are countless people suffering from depression and suicidal thoughts. This where we realized that the best way to help people in these situations is to provide an outlet to talk about these issues anonymously. We realized how difficult it is to actually determine if someone is in need of assistance, which is why there is a need for another method to help people in this situation. This is where we decided to take action and make that outlet for concerned friends and elders.

What it does?

When We Listen is an anonymous social media outreach application that helps your concerned friend or acquaintance avoid depression or suicidal thoughts. When We Listen integrates Twitter's API with Indico's Emotional Analysis and Microsoft's Text Analytics API's with the intent to determine a user's mental state. The combined data sets and statistics from Microsoft and Indico are analyzed to provide us with enough information to successfully output how strong one's mental state is as a percentage.

How we built it

We started by developing the back end components and getting the twitter API to retrieve all of the tweets from a specific user. After retrieving all these tweets, we filtered out all of the tweets related to depression and suicidal through keyword searches. The keywords were chosen based on extensive research about past deaths and depressive states on both famous and regular people. After all the tweets that matched with the keywords were received we used a combination of Microsoft text analytics and Indico's emotional analysis API's to generate a numerical value associated with how sad or negative tweets are. From these values, a percentage of how happy you are is obtained. This algorithm was then implemented on a webpage developed using HTML CSS and JavaScript. This webpage has a simple input in the center for the twitter handle. After typing a twitter handle, the user is redirected to a page with a status bar that displays the percentage happiness along with a message that tells the user how to proceed based on the output.

Challenges we ran into

In the beginning, we each branched off into design,back end, front end implementation and research. The research component resulted in several controversial topics and presented us with several situations where the text analytics did not provide the expected emotional output. To combat this, we decided to use two separate text analyzers, this allowed for mixed results that ultimately improved our overall accuracy. Both analyzers returned multiple outputs based on levels of emotion for each input text. To successfully determine the exact emotion related we compared these values from both text analyzers and chose the highest emotion and its value.

Accomplishments that we are proud of

We are proud of being able to use Twitter's API efficiently and extract only the information that we need for obtaining data before analysis. This allowed us to have a large open data source through a fast and efficient code. We are extremely proud of the fact that we combined the two analytic API's from Microsoft and Indico to create a precise algorithm to determine mental state as a percentage. This was accomplished by combining research data on suicide events related to Twitter and a data on depressed and emotional weakness throughout existing tweets. Moreover, we are extremely satisfied with the front end of our entire project. Our Web App is a simple and extremely effective method of determining user's mental state. We are proud of the fact that we combined bootstrap's CSS and HTML applications along with our very own self designed webpage design. This illustrates how we were able to use all the tools available to us in the most meaningful way possible.

What we learned

The use of various API's was the main focal point of learning throughout this development process. The first API we used was Twitter's open source API, this was used to obtain tweets from selected public users. We then used Microsoft's Text Analytics API to determine the emotional state of various input texts. A similar API was used from Indico, however we mainly focused on Indico's emotional analysis and disregarded the other parameters they offered. We learned that the best way to analyze data is through expansion and filtration. Throughout the front end web application development process, we learned new techniques that helped improve the quality and professional look of our final application.

What's next for When We Listen

To add a direct messaging option, for the user to anonymously message their friend. We believe that the user should have a way to speak their concern however they feel will help the most. This will be done through the use of a private When We Listen Twitter account and will be implemented into the Web Application.

Built With

Share this project: