September 26-27, 2016
New York, NY

A peek at x.ai’s data science architecture

Angela Zhou (x.ai)
4:35pm–5:15pm Monday, 09/26/2016
Implementing AI
Location: 3D10 Level: Beginner

Prerequisite knowledge

  • A very basic understanding of or interest in NLP and NLG systems
  • What you'll learn

  • Understand an approach to the understanding (NLP) and response (NLG) dimensions of a human-machine interaction
  • Description

    In any human-machine interaction, you need a dialogue model: the machine must understand and be able to respond appropriately. Angela Zhou discusses x.ai’s AI personal assistant, Amy Ingram, who schedules meetings for you, focusing on the way x.ai has approached both understanding and responding.

    Topics include:

    • How x.ai uses various machine-learning techniques (deep learning, neural nets, etc.) to teach Amy to understand scheduling conversations (NLP)
    • How x.ai has developed its natural language generation (NLG) capability so that Amy responds appropriately at any given moment in a scheduling conversation, with a particular focus on its resolvers
    • x.ai’s offline data pipeline, which allows it to derive features from millions of email texts and train algorithms to automatically interpret them semantically and extract information relevant to meetings, and how x.ai uses this dataset to train a deep learning model that detects customer intents as expressed in text
    • x.ai’s online data pipeline, which has to respond within seconds to an arbitrary number of emails entering its system and automatically generate responses (To achieve this, x.ai frequently serializes updated models and seamlessly plugs them into production through a staging environment. New models get integrated into a “mock production system,” which allows x.ai to test them before final deployment.)
    Photo of Angela Zhou

    Angela Zhou

    x.ai

    Angela Zhou is a data scientist at x.ai focused on understanding scheduling-related email through data analysis, machine-learning, and NLP methods. In the past, she has used statistical machine-learning methods for iris detection and recognition. Angela holds a BS degree in mathematics from Northeastern University in China and a master’s degree in statistics from Columbia University.