NSF grant helps research into computers learning idiomatic phrases


Cassandra Smith

When you hear a phrase like “throw away,” you probably know of a few ways to use it in a sentence. While you may be able to use an idiomatic phrase like that accurately, a computer may not be able to do so. A project out of the University of Illinois Urbana Champaign is working on fixing that problem. 

Assistant Professor Suma Bhat (Electrical & Computer Engineering/Coordinated Science Laboratory) is working on a project that will allow computers to learn many idiomatic—expressions commonly used by a native speaker—phrases and understand them in both literal and figurative senses. Take “throw away” for example. Right now, a computer may not know if that phrase means to throw something away in the garbage can (literal) or to describe, say, wasting an opportunity (figurative). 

With the help of a $150,000 grant from the National Science Foundation, Bhat and her team will be able to pool together many phrases to help machines understand nuances and variability in language. 

Researchers said this will “create new natural language processing algorithms for accurate interpretation and generation of idiomatic expressions towards a more human-like processing ability in machines.”