Inspiration

We were reading about how more and more people face some kind of loneliness. And many studies including one published by Harvard says that it severely affects your mental wellbeing, your health, immunity and can cause acute depression, heart-diseases. And most of it is preventable, if these people get right help at the right time. Usually they don't have accessible means to get help. So we decided to create a chatbot that resides in your phone and is just a few taps away. We expect this will make people more comfortable to seek help.

What it does

When user comes to Yuumi, We engage him in a conversation in order to gauge how their sentiment is, how they are feeling. Once their sentiment determined, we have a set of offerings based on their mood, say if they are feeling happy we offer them something interesting to read or watch. But if they are feeling sad, our offering ranges from offering a professional help to our very important feature - User will be able to ask community anonymously and will be able to get an unbiased opinion from them.

How I built it

We have used messenger's API to deliver first few messages to induce an input from user, once we get that, we run a sentiment analysis (Built in PyTorch) on that text to measure user's sentiment. Based on that sentiment, we offer user various options using Messenger API template messages.

Challenges I ran into

Gettign the PyTorch sentiment analysis to work, training our models took a lot of time.

Accomplishments that I'm proud of

Getting a lot done on second day, and finally be able to stitch a solution that actually WORK !!

What I learned

Whole team was quite amazed how easy it was to set up a Messenger bot, and API's documentation was quite good, and so was few easy-to-setup guides such as 'test-drive'. We never worked on PyTorch, and we feel that it is a powerhouse, I would never come across it if not for this hackathon.

What's next for Yuumi

Very exciting about this, in its full form,

  • Yuumi will be able to have conversations with user by its own. It will have an intelligence of its' own by reading that user's profile and their preferences, thus offering very tailored conversation.
  • Yuumi will be able to see where do user stand in terms of sentiment scale, if his recent activity and current conversation gives us an alert of some severity, we can add further measures to provide appropriate help or alert authorities that will be able to help them.

Built With

Share this project:
×

Updates