Sign In

Communications of the ACM

ACM TechNews

How Computers Parse the Ambiguity of Everyday Language


Ohio State University researchers' Madly Ambiguous game highlights the issue of ambiguity in Natural Language Processing.

Ohio State University researchers developed an online game as part of their investigation of the challenge of ambiguous language for computers.

Credit: Mike Lawrie/Getty

Ohio State University researchers investigating the challenge of ambiguous language for computers used an online game to clarify work in the field of natural language processing (NLP).

The researchers' Madly Ambiguous game challenges players to stump a computer by filling a blank space in a sentence with a word whose semantic role the system attempts to infer.

The computer either follows a rule-based approach traditionally used in NLP by focusing on the main noun and looking it up in a semantic database, or uses word embedding to map words and phrases onto a vector space in which similar words appear closer to each other.

Since the game's launch, the computer has corresponded with user judgments 64% of the time using the first method, and 70% of the time with the second technique.

From The Atlantic
View Full Article

 

Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA


 

No entries found