AggieBot

Inspiration

We've always had trouble going through the TAMU website, maze of tabs and links, to find the information we needed as students.

What it does

The bot goes through all the information on the TAMU website and makes it easily accessible for students by using the chatbot to pinpoint and relay the required information.

How I built it

We used Google's cloud computing services and Dialogue Flow to create and train a working chatbot. It was trained on the information on the Texas A&M website.

Challenges We ran into

On the software side originally the chatbot was going to be made from scratch using a custom deep learning neural network with TensorFlow. Unfortunately, we soon realized that the test data was insufficient for training from scratch. Google DialogueFlow helped to redeem this as prebuilt language networks helped the chatbot understand much more easily. Although learning to use an entirely new API in less than a day proved challenging. On top of this we had to dust up on a language we hadn't used in a while, Javascript, more specifically, Node.js.

Accomplishments that I'm proud of

Learning the use and finding Dialogue Flow within 24 hours was probably the thing we are proudest of. It showed that we could be flexible and were capable of finding solutions when the plan went awry.

What's next for AggieBot

AggieBot can be deployed in various other websites with relative ease, this would help customers find the information they need much quicker than they would with a website. It can even be modified to find data from large databases with ease and even explain live sensor data.