Inspiration

Giving medical advice at the hands of a well thought out and truly conversational AI user experience. Many of the self checkers that were developed for the Coronavirus symptoms are simple web-form inside of a chatbot. We believe the user experience can be more natural in conversation and thus provide a more robust data collection of the user's conversation with the chatbot.

Many people in the United States, perhaps the world. do not readily have access to a doctor let alone a diagnostic test for COVID-19.

What it does

This Coronavirus questionnaire gives access to everyone who speaks with Mercy, our conversational AI and provides important recommendations based on the user responses.

How we built it

We built it with Luis AI cognitive services for the NLU and the Microsoft Bot Framework. This gives us the ability to design a very human-like interaction with a medical assistant like persona that asks questions and takes responses from the user in a natural back and forth dialog.

Challenges we ran into

The conversational AI UI/UX. Going beyond a button or web-form format for conversational chatbot systems has a layer of complexity that not many chatbot developers dare to endeavor.

Why is this so important? When you create your dialog flow normally people try to proceed from one dialog to another dialog using buttons or a web-form like interaction. This is very limiting because you confine a person to a particular choice. Rarely are there multiple selections of choices or the ability to repeat and ask for additional choices.

This interaction is more web-like and non-conversational but rather button selections. Instead, the challenge becomes how can you create a conversation that is natural and robust in questioning the user but also collecting pertinent data from the user's conversation to make pivots in dialog flow to follow up and or ask additional questions through conversation that pertain to something the user said.

For example, I could ask you what is your favorite color or I could show you a list of 100 different color choices... Furthermore, in our conversational AI instead of me giving you a choice of symptoms I simply ask you for the symptoms you are experiencing.

This is more natural, effective, and how a doctor would actually take a history from a patient.

Accomplishments that we're proud of

With our experience in conversational AI development and a very robust code based bot framework from Microsoft, we are able to develop a conversational AI experience that is vastly better than most, if not all, chatbots out there today.

What we learned

During the process, we learned that this technology and codebase can be used well beyond just checking for Coronavirus symptoms.

What's next for Coronavirus Health and Temperature Checker

Taking the user data from the conversational AI usage and mapping it to the 50 states so that data can be utilized through an API about potential COVID-19 patients. therefore, aiding in their testing, care, and hopefully successful outcomes.

Website and chatbot coming soon!

Built With

  • azure-cognitive-services
  • azure-maps
  • luis-ai
  • microsoft-bot-framework
  • power-bi
  • qnamaker
  • redis
  • typescript
Share this project:

Updates