Sign In

Communications of the ACM

ACM TechNews

Teaching Computers the Nuances of Human Conversation

View as: Print Mobile App Share:
Artist's impression of a computer using natural-language processing.

Natural-language processing continues to struggle to understand the crucial nuances of human speech, according to Marilyn Walker, a professor at the University of California, Santa Cruz.


University of California, Santa Cruz professor Marilyn Walker says natural-language processing has become very good at recognizing human speech, but it still struggles to understand its subtle, yet crucial, nuances. In particular, it cannot yet pick up on the ways that word choice and sentence construction reveal a speaker's personality and the social relationships between people.

Walker says developing software that can do that would be a major step along the road to creating companion robots and other naturalistic technology.

One of the projects being worked on by Walker and her colleagues analyzes thousands of conversations from online debate forums. Using crowdsourced summaries of these conversations--which touch on a variety of topics from politics to pets to technology--and analyzing them linguistically, the researchers hope to tease out information about how sarcasm is conveyed in text, how people present and support arguments, and other ways people communicate. The goal is to create software that will be able to summarize online conversations and readily identify what position any given participant is taking. Walker says such a tool could make debates less polarized by enabling people to identify and understand opposing arguments.

From University of California Santa Cruz Newscenter
View Full Article


Abstracts Copyright © 2014 Information Inc., Bethesda, Maryland, USA


No entries found