The user is notified when their speech indicates that they may be feeling anxious or depressed
We were inspired by the recent surge of discussion of mental health across our nation. Statistics show one third of college students feel depressed. So we thought the health and wellness theme of Hackduke's Do Good Challenge, provided a great opportunity to pursue this idea.
What it does
TriggerNow uses Speech Recognition and Semantic Analysis to determine a person's state of mind from the words that they use.
How we built it
We used react native to make the app itself. To perform the semantic analysis we used an open source classification data set of over 20,000 words ranked with positive and negative values to give a sentiment score for each phrase the app picked up.
Challenges we ran into
Too many dependency configuration issues to count, especially with Python. Android development in general is very difficult to setup, even without having to use Android Studio.
We also originally planned to use a heartbeat sensor, but we realized that the pebbles didn't have a heartbeat monitor, so we have to adjust our plans.
We considered using a muse headband but that was simply too much temerity data and we decided we wouldn't' have enough training data to get reliable results.
Accomplishments that we're proud of
We never used react-native before, and this is our first ever NLP-based hack!
What we learned
Basics of NLP and neural net algorithms as well as react native and android dev. A very educational 24 hours!
What's next for TriggerNow
Future steps will include refining the NLP algorithm as well as adding more parameters to our consideration to give more nuanced diagnoses of a person's mental state. We think by including analysis of heartbeat and examining more precise speaking patterns such as inflection, frequency, and what sequences of words someone utters we can gain a more subtle understanding of his or her mood.
Moreover by tracking someone's location we can determine locales that might cause someone to feel anxious and inform them and try to put them at ease with a either recommendation or a distraction.
We plan to eventually build out a decision engine that will evaluate and weigh these factors across time frames to create a mental health profile which users can track and look back to see the progress they've made.
More importantly we want to provide more interactive feedback for users so as to resemble a kind of personal diary. Since many people who suffer from some form of mental illness or are just going through a tough time keep a diary as a means of expressing their struggles and finding relief. Saying something out loud can provide an even greater sense of relief because it gives the speaker a feeling of permanence. Thus we want to respond back to someone's statement with something that might cheer them up or distract them like an image of their cat or an event they can look forward to, or a quote with which they can find solace. Ideally we'd like to capture the intoxicating feeling of playing an old school text adventure where it seems like someone hard coded almost every possible combination.