Table 124, seniors from TJ.
Inspiration
We were inspired by this project because we have a lot of friends who are worried about bothering their friends when they are having issues. By talking to a bot instead, they do not have to worry about being obnoxious or being judged, when they really just need some suggestions on how to improve their mood. Also, sometimes friends and family aren't always available to talk to, so a bot would be a more accessible option if needed.
What it does
CareBot is a Facebook Messenger Bot that has a conversation with the user. It asks the user how they are feeling, suggests types of advice to give, gives specific advice, and finally elaborates on one selected piece of information with links and suggested routines.
How we built it
We built this bot by creating a Facebook Messenger application and connecting it to DialogFlow, a conversational interface that uses natural-language processing and artificial intelligence to interpret user input. We coded a fulfillment in Node.js in order to create responses to the users programmatically and connect each piece together.
Challenges we ran into
Getting the application to the point where we could interact with CareBot was extremely difficult. Connecting Dialogflow, Facebook Messenger, and ngrok together was extremely difficult. This was further complicated about halfway through the hackathon when the bot stopped working, and it took a long time to realize that ngrok had timed out. We also ran into many problems with DialogFlow, as it struggled to properly interpret much of the user input, especially when DialogFlow Contexts were involved, so workarounds had to be placed into the code.
Accomplishments that we're proud of
We are extremely proud of being able to put together a functional bot on Facebook Messenger that actually interacts with users. We are also proud of all of the different suggestions we researched and came up with in order to help users, and the different ways they were classified. This took up a significant portion of our hacking time.
What we learned
We learned various different ways to help out those who are feeling down in some way or another. We also learned how to set up Node.js servers using ngrok, how to make a Facebook application, how to use DialogFlow, and how to start building a conversational user interface.
What's next for CareBot
Next, we need to work on expanding the User Interface, by letting users click on different options instead of having to type them all out. Furthermore, we hope to get better at using DialogFlow, which will allow us to accept user input that is more different than what we expect. Ideally, CareBot could figure out what a person is feeling and how to help them through more idiomatic language, such as "I'm a mess."
Log in or sign up for Devpost to join the conversation.