Put AI to Work
April 15-18, 2019
New York, NY
Please log in

AutoML in the Chatbot Builder Framework

Jaewon Lee (Naver/LINE), Sihyeung Han (Naver/LINE)
4:05pm4:45pm Thursday, April 18, 2019
Implementing AI
Location: Trianon Ballroom
Secondary topics:  Automation in machine learning and AI, Media, Marketing, Advertising, Models and Methods, Text, Language, and Speech

Who is this presentation for?

  • Deep learning model engineers who want to implement better chatbot models, data coordinators who want to better manage large data corpora, and business planners who want to apply chatbots to their own business

Level

Intermediate

Prerequisite knowledge

  • Familiarity with NLP and chatbot concepts (useful but not required)

What you'll learn

  • Explore a detailed pipeline with the Chatbot Builder Framework
  • Understand NLP engines, especially dialogue models
  • Learn how to adopt AutoML in the Chatbot Builder Framework

Description

There are lots of steps to building a chatbot, and each requires tremendous work. Now, with the Chatbot Builder Framework, you no longer need to worry about building a chatbot. The Chatbot Builder Framework only requires a raw data corpus to create a high-performance chatbot on your own business domain. Just click the Build button and the framework will manage that data corpus by clustering into scenario, preprocess data, optimize hyperparameters for deep learning models, propose the best ensemble model for your domain, and finally serve to multiple messenger platforms such as LINE and Facebook Messenger. This entire pipeline will suggest the most optimized chatbot model and serve it to users so that they can apply to their own business.

On the Chatbot Builder Framework, clustering all queries into similar clusters helps to easily manage large text and log data corpora. Also, auto-hyperparameter tuning allows deep learning engineers to focus more on current model architecture or developing other cutting-edge models. In the end, each model validates each status by computing self-evaluation scores.

Jaewon Lee and Sihyeung Han walk you through implementing a self-trained dialogue model using AutoML and the Chatbot Builder Framework. You’ll discover the value of AutoML, which allows you to provide better model, and learn how AutoML can be applied in different areas of NLP, not just for chatbots.

Topics include:

  • The necessity (and the difficulty) of using AutoML in NLP and chatbots
  • How to apply AutoML in the Chatbot Builder Framework
  • Data clustering and hyperparameter tuning
  • Autoevaluation by models
  • Current business cases using AutoML for chatbots at LINE
Photo of Jaewon Lee

Jaewon Lee

Naver/LINE

Jaewon Lee is a data scientist working on NLP at Naver and LINE in South Korea. His team focuses on developing the Clova Chatbot Builder Framework, enabling customers to easily build and serve chatbots to their own business, and undertakes NLP research to improve performance of their dialogue model. He joined Naver/LINE after his company, Company.AI, was acquired in 2017. Previously, Jaewon was a quantitative data analyst at Hana Financial Investment, where he used machine learning algorithms to predict financial markets. He holds a BS in applied math and statistics with computer science from Johns Hopkins University.

Sihyeung Han

Naver/LINE

Kevin Han is a business consultant and service planner at Naver/LINE, a Korean company known for the biggest domestic web portal (Naver), mobile messenger (LINE), and AI-related solutions (Clova). His team provides an end-to-end AI service for clients, from improving dialogue models to consulting clients to create maximum value out of the company’s chatbot service. Previously, Kevin was a cognitive consultant at IBM. He holds a bachelor’s degree in psychology and business from New York University.