Inspired by a lack of clarity over text messages - which can often lead to emotional trauma - we decided to leverage an AI to help analyze the intentions behind text messages. This could prove especially useful to individuals with emotional learning disabilities such as with Autism and Aspergers'.
What it does
Emotional Assistant processes a message and returns the confidence of six common emotions. We return the highest confidence and display it to the recipients of the message.
How we built it
We built a chatroom using Socket.IO and Express/NodeJs. The messages are intercepted and fed through paralleldots' emotional analyzer API. The output is brought back to the users.
Challenges we ran into
We started with wit.ai - manually training the model. It was insufficient, and close to the deadline we had to swap to parallelDots emotional analysis. We also started with a login-implemented boilerplate and decided to drop that for a simpler, sleeker UI.
Accomplishments that we're proud of
Wiring so many libraries was incredibly convoluted and the team is proud that the final output is (more or less) polished and functional as originally specified.
What we learned
Building a chat room is a lot harder than it seems. Even with a wonderful library like Socket.IO.
What's next for Emotional Assistant
Implementing a custom LTSM AI NN with brain.js to analyze the conversational tone with respect to the emotional analysis of all chat messages in a conversation. In short, analyzing the tone of a developing conversation based on the tone of the messages.