Overview
Making a choice can be hard, let alone when you’re hungry. We’ve solved some of the toughest decisions through voice and chat technologies.
Business Case
Information Provided Instantly
As a decision agent, voice and chat technology allow users to quickly work through layers of questions in a natural format (human conversation) in order to get to an answer. While we solved a simple problem (ordering lunch), these technologies could be applied to solve real-world problems. For example, it could be leveraged in hospitals where users could leverage chat or voice UI to find their way around the space, or by staff to locate a nurse or doctor. In the end, this application goes far beyond getting lunch, and is effortlessly presenting information that the user wants and needs in a timely fashion.
The Experiment
Eliminating Indecisions Through Various Technologies
In order to eliminate the indecision about deciding what’s for lunch, we developed an Alexa skill that searches Yelp for nearby restaurants. When prompted, Lunchbot extracts the user’s request and sends it to Yelp’s search API. Designed to simulate conversation with human users, Lunchbot sorts through menu items and then proposes a few options on where the user can grab a bite to eat. Using the user’s order history to personalize their experience, it then delivers that experience to various forms of communication platforms including Slack, Alexa, Twilio (SMS), a custom web-based chatbot and more.
Technology Overview
-
Conversation Management Tools We used a combination of Dialogflow, a cloud-based service for developing voice interaction for various platforms including Slack and SMS and Amazon Alexa to build a customized voice experience. -
Cloud Data Processing To help us hit the remote endpoints such as Yelp Fusion API, connect to the user profile database and process data we enabled Google Cloud functions and AWS Lambda to hook everything together. -
Datasources - APIs and Databases In order to find the best match for the search query, we used Yelp’s API to locate open lunch spots in the local area. Simultaneously, we used Firestore, a NoSQL document cloud database, to manage Lunchbot’s user profiles and order histories.