Even today, mental illness takes one in five Americans. We wanted something for people with mental illness, something which can assist them in crisis situations or just be there to converse.
What it does
Our website has a built-in intelligent chatbot which talks to people in distress and helps them overcome anxiety, stress, depression, anger, etc using various suggestions and techniques and also learning from the user's personality traits.
How we built it
We used IBM Watson's conversation service to build and train our chatbot. We configured specific intents, entities and dialogues to train the chatbot. The conversation service is deployed on a specific workspace on the IBM cloud instance. We save the conversations using CouchDB on IBM Cloudant. The website provides various self-help options for people facing different kinds of emotions. We present the user an array of blogs, videos, gifs, games, and even some stress relieving exercises published on the webpage based on the mood of the user. We aim to always improve the performance of the bot with each conversation and so as an additional feature,we developed an analytics engine that processes feedback from the user. The feedback forms ask the user which technique helped them overcome a specific phase they were going through in the ongoing session and how useful it was talking to the bot. We developed a Python Web App that is seperately deployed for this.The chatbot performs ajax queries to post to this engine. A url can be accessed by domain experts which then displays pie-charts/graphs using GoogleAnalytics to improve suggestions in future conversations.
Challenges we ran into
The learning phase for the bot is still on going as the chatbot still has a lot to learn and then respond accordingly. We tried to develop a ID3 classifier for better analytics and explored IBM-experience manager/corpus for analyzing the Couchdb but ran short of time for implementing this.
Accomplishments that we're proud of
Completing the chatbot, webapp and analytics given the time constraint. Integrating IBM's watson conversation service, cloudant and OpenWhisk in the web app.
What we learned
IBM-Watson, distributing team-work, integrating chatbot, analytics and web application.
What's next for Solace
An advanced analytic engine using IBM Watson's Knowledge Studio/Discovery to enable the bot to learn on the go. Offer the service as a mobile application. Integrate with hospitals nearby using push notifications to assist in emergency situations by checking-in a patient history. A peer network to interact with other people experiencing similar issues.