MentaliSAR

Inspiration

Since I have attended uni, I have seen many people suffer from depression and anxiety (sometimes ourselves). Therefore, we wanted to do something so we could help others in the best way we could. So we made MentaliSAR, a small gift to the community by Sashank, Arnav, and Ronan (SAR).

What it does

It is a Natural Language Processing chatbot that intelligently maps user input into categories of the questions and chooses a suitable answer as a response. It is specifically trained on a dataset with mental health-related questions.

How we built it

We built it using python and flask with the help of NLP library-nltk and deep learning library-TensorFlow (along with tflearn). We made our own dataset with sufficient data to train the chatbot well enough to answer basic queries.

Challenges we ran into

Getting the whole project onto a web server was difficult since none of us knew how to use flask. Setting up the environment was also very hard since we were getting a lot of import errors and were not able to continue.

Accomplishments that we're proud of

Successfully made our first working chatbot web app that answers accurately 95% of the time. We are proud of the solution we have devised with hard work and perseverance.

What we learned

We learned how to lemmatize/stem/tokenize words from sentences and use them to accurately respond to the user query. Furthermore, we also had fun learning how to launch a temporary web server to host the project with basic functional UI using flask.

What's next for MentaliSAR

We have only laid the foundations for an idea that can have a major positive impact on society. We hope to scale the project into a fully functioning conversational AI with a deeper knowledge of mental health problems, linguistic nuances, and a more personalized user experience.