Inspiration

With the steady increase of mental health issues such as depression and anxiety, we wanted to create something that will offer assistance to those who need it. Curi was made with the aim to support people who don't have an immediate outlet to express their struggle(s). If someone is in need of help with mental health and doesn't know how to seek help, they can talk to Curi about their struggles and receive a response.

What it does

Provides a safe anonymous space for people to get help with mental health issues such as depression and anxiety by talking to our chatbot, Curi. Users can ask Curi questions about depression, seek information on how to get help from a professional, or simply tell Curi how they are feeling.

How we built it

We first had to find public information that holds questions related to mental health and the responses to those questions that were ethical. We then scraped that data into a format that we could modify and parse in order to train the behavior of Curi. Once we had the data, we had to lemmatize and stem the data to get clean and meaningful data. With this data we were able to finally create a pretty JSON to use for our bot. Using Rasa from python gave us the tools due to giving a powerful natural language understanding engine, dialogue management, and customizability. Rasa required us to create yml files from our data to be able to make nlu, domains, and rules files for our bot. This led to having to create more python scripts to transform our json data to yml. On doing so, we were now able to use Rasa and give it a large data set to train with. We created test scenarios and got results from Rasa to adjust its language flow. Overall we were able to create an innovative chat bot, that can help a person in need.

Challenges we ran into

It was a challenge to find enough data sets that were appropriate for our program. All the data sets that we used had different structures so we needed to modify each data set individually to merge them together. We also ran into a lot of issues with being able to choose the right kind of chat not to create. We tried using IBM Watson, Microsoft’s Bot Framework, and a few other third-party chat bots but they gave a lot of issues. Being able to transfer data from one form to another also created a lot of issues.

When converting finished, we had difficulty turning our basic HTML into a react frontend because we could not get some of the functionality of our chat to carry over. We also had the issue of initially creating our bot as we tried out different nlps and bot frameworks.

Although we created a fully functioning bot that could engage in conversation, we did not finish connecting our frontend to our backend in time for submission.

Accomplishments that we're proud of

We are proud to have a functional and working chatbot that can ask questions, reply to, and overall hold a conversation with the user about mental health. Furthermore, we are also proud to have trained the bot and further push the abilities of ai to a place that aids good mental health for humanity.

What we learned

  1. Learned how to scrape public data
  2. Merge different data sets
  3. Create scripts to transfer data to different dataset
  4. nlu
  5. TensorFlow
  6. Node.JS
  7. Rasa

What's next for Curi

We hope to further expand and train Curi to become a more fluid AI as well as be able to support any mental health issue and support/save multiple chats/users at one time. We would love to expand the design of [our web name] to a mobile app.

Share this project:

Updates

Submission history