Hello there! We are team Lancelot, a group of two students from Columbia University who pursued Master of CS degrees from The Fu School of Engineering. When looking at the competition we were inspired by the many new ChatBot / Virtual Assistants that have been released such as Apple Siri and Google's "Now" service. We realized that a highly customer centric business such as AirBerlin could greatly next generation enterprise application that both predicts customers needs and assist them with changes to their reservations in real-time.
What it does
We named our project "Amelia" the AirBerlin Virtual Assistant. The system is designed to track a business customers existing reservation, make new ones, and give them "intelligent" recommendations based on their needs. By designing a progressive web application we were able to build a platform agnostic backend that could potentially be integrated with AirBerlin's real production system. In addition to understanding several trained interactions the system actually processes dynamic language through Natural Language Processing - meaning a user can say a phrase in multiple ways and the system will understand it.
Ex: "Book me a flight from JFK to ATL" or "Get me a trip from JFK to ATL" are both understood.
Key "Intelligent" Features: (Proactive not Reactive Features) 1) Monitor travel/distance time to Airport from current user location using current GPS 2) Recommend to rebook a flight if the travel distance to the Airport reaches a critical threshold 3) Automatically schedule an Uber to take the user to the Airport 4) Provide real-time weather updates and suggestions based on travel destinations 5) Provide real-time flight updates 6) Offer easy interactions to get more information on their flight or destinations
How we built it
We built the system atop IBM Watson's Dialogue system, a generic infrastructure that can provide a framework for chat assistants. A great deal of time and development went into teach Amelia how to understand various phrases of natural language an giving his "intelligent" skills that allow her to proactively assist the client without having to be a solely reactionary system. The front end system is a fully cross-platform web application that can work on any mobile phone and integrated with existing AirBerlin web applications. We build the infrastructure on scalable AWS EC2 instances which can deployed globally.
Challenges we ran into
One of the biggest challenges we ran into was connectivity issues to the AirBerlin API, unfortunately due to the excessive number of participants the API would periodically go down - even during our demo. We worked around this.
Accomplishments that we're proud of
Aside from building a fully working demo one of the most proud of features we have is the automatic rebooking feature. This feature will proactively use the customer's GPS (after authorized by user) to check the travel time and distance between the user and the Airport. If it looks like it is probable they are going to miss the flight it will send them a push notification and help them rebook right in the app.
What we learned
We learned to think outside the box and combine existing technologies in a new and creative way.
What's next for Lancelot
We hope to actually partner with AirBerlin to build Amelia or some version of an enterprise chat assistant. If we win we would be very interested in exploring this exciting topic with you further.