I am travelling on business almost every month as part of my job in market development. Juggling between client meetings requires me to adjust my flight schedules quite frequently. Looking up the airline website and searching for the right links is tedious. The airline app on my phone gets lost between 100 other applications installed and sometimes it does not have all the details I need
What it does
Flightzilla is a personal chatbot travel assistant for business travellers. The assistant is available 24x7 to help busy corporate travellers manage their travel schedules, manage seat upgrades , change flight dates, lookup hotels. It uses natural language processing(NLP) to provide an easy to use interface over Facebook Messenger.
TRY IT by clicking on "Message Us" button at the site [link][http://www.flightzilla.co]
How I built it
I used Node.JS and integrated it with a NLP engine. The NLP engine was trained by providing it various data classification e.g. airport, day of week, date and other keyword with grammar structure. This was integrated with Facebook messenger. The Node.JS server is the glue that holds this together by having an input-output with FB messenger and the NLP engine.
Challenges I ran into
Designing conversation flows is a new field and there is no tutorial available. Normal chatbot frameworks do not understand context. They are simple single line NLP engine. I had to deal with the concept of context in the conversation. example: The conversation should remember what the user mentioned earlier and link it within the same conversation.
Accomplishments that I'm proud of
There are no text books for designing conversation flow. I am self taught NLP engineer after tinkering with open source NLP engines. Just like UX/UI , I believe conversation flow design will be very important as chat based commerce is gaining traction.
What I learned
To do is to learn. To learn is to focus. Focus to enjoy.Enjoy the journey.
What's next for Flightzilla Assistant for Business Travellers
I will be focusing on training the NLP engine. I am now confident that one can build great domain specific NLP engine for the travel industry. I am looking forward to partner with travel industry on this journey. Some obvious enhancements are required
- integration with Facebook External authentication. This is now available on FB e.g. Facebook Messenger can link the FB id with a travel website's user id.
- payment integration. Facebook messenger now supports this and provides a seamless experience to customers.
- examples of travel itinerary using rich templates in messenger.