Instructors:
Zhengzhong Liu, Zhiting Hu and Eric Xing
Data and Time:
10:45 am – 12:30 pm, Friday, Febuary 7, 2020
Introduction
The recent success and growth in the fields of Natural Language Processing (NLP) and Artificial Intelligence (AI) have presented the world with a large number of new applications, techniques, models, and architectures. In this tutorial, the audience will learn how appropriate abstraction and modularization can streamline both the development and deployment of NLP technologies. The tutorial will provide a systematic view of the NLP landscape, spanning text understanding, generation, and retrieval. We will present a principled breakdown of the broad tasks/techniques, and the actual systems that implement and operationalize modular NLP development. The tutorial also includes hands-on sessions, from which the audience will use the open-source systems to practice modular NLP and build complex applications. In sum, the tutorial delivers NLP in a modularized and systematic view, with a significant focus on practical development.
Tutorial Outline
- Motivation for Modularization (10 mins)
- Natural Language Prcoessing Overview (10 mins)
- Modularizing NLP Pipeline (35 mins)
- Q&A (10 mins)
- Modularizing NLP Model & Learning (30 mins)
- Q&A (10 mins)
Resources
- Tutorial Handouts
- AAAI Tutorial Forum
- Forte: A toolkit for modularizing NLP tasks and abstracting models from domain logics.
- Texar and Texar-Pytorch: The toolkit built for modularizing machine learning in NLP.
- ASYML project group: Our project group: Machine Learning as Machine Assembly, we aim at designing modularized components for NLP and Machine Learing in general.