-
Positive input
-
Response to positive input
-
Neutral input
-
Response to neutral input
-
Negative input
-
Response to negative input
Inspiration
All members on the team are current freshmen studying CS or EECS at UC Berkeley. Being part of one of the most competitive college campuses in the country, we have noticed that many students deal with stress and anxiety. Despite the many medical resources and support groups formed by students and staff on our campus, most students tend to become isolated and, in severe cases, depressed. Coupled with an unwillingness to seek help, these characteristics are key risk factors in depression and thoughts of suicide. A study at the University of Virginia showed that 1000+ college campus suicides occur within the United States each year, a figure which is currently on the rise. In addition, Emory University reports that fewer than 25% of students with depression or suicidal thoughts receive adequate care and medical attention. We decided to create a project that would help ameliorate this situation by identifying risk factors and warning signs and implementing protective factors to improve students' mental health.
What it does
StressLess takes input text in the form of sentences describing a student's mood and daily activities, similar to a journal entry. The application then uses the IBM Watson AlchemyLanguage API to analyze the text and extract keywords, then determine if the student is in low, medium, or high risk of developing suicidal thoughts. This information is then used to generate a response for the student, indicating what his or her next steps should be (reflecting on their day, giving suggestions on how to improve their mood based on activities that they mentioned in the past, or seeking professional help).
How we built it
Our frontend is a web portal built using HTML and CSS, which consists of a login screen and a text box in which a student enters statements about his or her day, a couple of sentences at a time. We used the Watson Bluemix AlchemyLanguage API to analyze the input text. We used Emotion Analysis to analyze the overall tone of the entry; if the overall tone was "joy," we extracted keywords from the text using AlchemyLanguage's Keyword Extraction and logged them as entities that made the user happy. If the overall tone was not "joy," the app returned suggestions to the user on how to improve his or her mood using the entities that were logged from joyful entries. If, however, the text contained extreme words linked to depression, anxiety, or hopelessness, the app gave a list of ways to reach professional help lines.
Challenges we ran into
The web framework was difficult to work with, as none of us had experience using Django. In addition, we had difficulty getting the login screen set up, as well as exceeding our limit of Watson API calls during testing and debugging.
Accomplishments that we're proud of
We learned to use the Watson and Django APIs, which none of us had experience with. We also created a more time-efficient solution for simulating user accounts, which we could not build in the allotted time.
What we learned
We learned how to use two new APIs, as well as how to host a website using Google App Engine.
What's next for StressLess
Implement user accounts so that different people may log their mood on a daily business and see continuous results. We would also like to build a mobile feature, which pings users when they need to do tasks that improve their mood (example: if eating lunch makes a user happy, the app would alert them at noon to remind them to eat).
Log in or sign up for Devpost to join the conversation.