Our inspiration comes from the opening statements from HomeAway President John Kim. He mentioned how one of the major problems that HomeAway faces is providing customers the most relevant recommendations. Customers have to click through so many home listings that do not match their preferences before finding the right home or just leaving the site. Using artificial intelligence, we created a photo organization tool to segment photos based on the features that customers care about most. We then used this tool in conjunction with a chatbot to streamline the entire booking process. Our goal is to provide each customer with the most exceptional and personalized experience.
What it does
The image organization tool categorizes photos based on these five features: bedroom, living area, kitchen, bathroom, and amenities. With the ability to segment photos, we then created tabs within each listing to sort photos based on what type of home feature they were considered. This allows customers to quickly browse photos based on their personal preferences. The tool also allows us to optimize which photos show up first in the thumbnail of listings. This decreases the number of clicks needed to find what they want and increases the probability of a customer booking a home. Out chatbot streamlines the entire booking process. Through real time communication, our chatbot allows customers to browse listings, view home features, compare listings, book, and receive customer service.
How I built it
We trained a CNN with keras on a dataset we gathered using the University GPU’s. We then designed a tabbing interface in HTML. Lastly, we used the facebook messenger API and wit.AI to create an intelligence facebook bot. Challenges I ran into The biggest challenge was cramming and finishing the project on time. Dealing with 40 gigs of data was challenging because we had to move the data over to multiple machines, which caused a lot of downtime. Other challenges included deploying the Facebook messenger bot service to the cloud.
Accomplishments that I’m proud of
We are proud of creating the tab feature under each listing, developing a photo organization tool, using Facebook’s bot API to communicate and learn from customers, and working together as a team. We are also proud of gathering a huge dataset that spans multiple gigabytes. Successfully building a FB messenger bot end to end that can generalize multiple responses and having our network learn successfully.
What I learned
I learned that it is very important to map out the entire process of the project, starting from the very beginning. This way, everyone can be on the same page regarding what we want to accomplish. We also learned that we need a data handling mechanism beyond the basic local file system and file transfer over servers to handle big data.
Whats next for Homie AI
Homie AI’s next steps is to become a product build at HomeAway and to eventually be on every user’s Facebook messenger to provide everyone an exceptional and personalized vacation booking experience.