CSE 591 Special Topics on Natural
Language Understanding (NLU) and Question Answering (QA)
Instructor: Chitta Baral
Description of the course: In this
course we will learn the science and engineering behind building a question
answering system using natural language understanding and reasoning.
The theoretical content that
will be covered in the course includes:
á Knowledge Representation
á Translating natural language to knowledge
representation languages
á Probabilistic Combinatorial Categorial
grammar
á Lambda Calculus and Inverse Lambda
á Machine learning approaches to natural language
processing
á Ontology development
á Building a natural language corpus
There will not be any textbook for the course. The
course material will consist of several chapters from books and have several
research papers.
As part of the course groups of 3-4 students will
build a system that can take natural language text as input and can answer
questions about them. The system will be a learning based system that will take
as input natural language sentences and their translations to appropriate
knowledge representation languages.
It will then learn how to do such translations for new unseen sentences.
The system will then use such translations to answer questions. We expect our
systems to have deeper reasoning ability than the WATSON system developed by
IBM that could beat the Jeopardy champions.
Some of the domains that will be offered for student
projects are:
á Combinatorial puzzles
á Wordnet glosses
á Framenet annotations
á Planning domains expressed in natural language
á Robocup commands
á Geoquery
á Chapters of high school Biology, Chemistry and Physics
books
á Human robot interaction languages
á Policy descriptions in natural language
á Natural language descriptions of databases
á Natural language text in Archaeology
á Natural language text from Pubmed
The grading will be based on a test (40%), class
participation and presentation (10%) and on the project (50%).