YANA (You are not alone)

YANA Chatbot

You Are Not Alone (YANA)

You Are Not Alone (YANA) is an application that was built in response to the opioid overdose crisis in Canada. YANA is aimed to be a support system for those with substance use challenges and offers a way for their friends to be immediately notified in the event of an overdose. The prompts of our chat-bot, YANA, are phrased in a way to show that the main issue is NOT the usage of these substances, but reinforces that they are being safe and taking precautions so they do not MIS-use.

Users sign up via Facebook, provide a list of friends, and authorize the application to notify these friends in the case of an emergency. In the future, there will be an option to notify healthcare authorities in case of an emergency, but this will be completely up to the discretion of the user.

If a user indicates that they are about to use, YANA will offer a reminder to have naloxone nearby. YANA will then standby and prompt the user at regular intervals to ensure that nothing has gone wrong. If the user is unresponsive for multiple check-in intervals in succession, YANA will immediately send an alert message to all of the friends that were listed during sign-up along with some information about naloxone. Furthermore, this alert message will also be sent if the keyword ‘help’ is typed into chat. In future iterations, the application would be able to call friends instead of only sending a message.

Users can also have conversations with YANA, as it implements the Google Sentiment Analyzer Machine Learning API and responds to the user based on the “sentiment score” of the statement the user inputs or specific trigger keywords.

Collaborators: Eric Kim, Aidan Matthewson, Leo Sporn, Shaqil Rahemtulla

Built With

Submitted to

RU Hacks

Created by

I worked on YANA's internal logic as a chat-bot. Using Google-cloud's natural language processing API makes this a lot easier, especially as it wasn't my first time working with this tool. However, care has to be taken in processing and interpreting the host of data Google's API returns. For instance shorter responses decrease the reliability of the analysis, or result in an error. I addressed this using a data augmentation method commonly used to address missing data in computer vision applications, slightly modified to work with missing text. Additionally, in some cases, the presence of certain words and phrases in a user's chat text required a hard-coded response, rather than relying purely on the sentiment analysis results to make a decision.

Leo Sporn
I worked on the React front end, backend routing, integration and the face-book-chat api.

Eric Y. Kim
Aidan Mathewson
Shaqil Rahemtulla