Montclair, NJ - Linguistics Professor Anna Feldman’s work in natural language processing provides a link between humans and computers. As a researcher, her work will have direct benefits to machine translation, specifically when it comes to translating figurative speech. “If you ever use Google Translate, you know machine translation is still not a fully solved problem,” she says. “While getting the structure and form of the translation is difficult, translating figurative language is back-breaking. First of all, the machine needs to detect figurative language, such as an idiom.”
Idiomatic phrases such as “hit the sack,” “eat my hat,” “blow my top” or “go cold turkey” are confusing for computers – and for language learners – to translate because they can often be taken literally, as well as figuratively. “Therefore we need to find a way to tell literal phrases from idiomatic phrases in different contexts,” Feldman says.
With the help of a $176,000 grant from the National Science Foundation, Feldman is working with Computer Science Professor Jing Peng and several linguistics and computer science students to develop a language-independent method to automatically decide whether expressions like “stabbed in the back” or “blow the whistle” have either a literal or figurative interpretation in the text.
The team has been able to quantify many linguistic properties of idioms and incorporate them into their idiom detection algorithm.
“This work will contribute to the advancement of research in sentiment analysis, opinion mining, machine translation and natural language understanding,” Feldman says.
With the help of a $176,000 grant from the National Science Foundation, Feldman is working with Computer Science Professor Jing Peng and several linguistics and computer science students to develop a language-independent method to automatically decide whether expressions like “stabbed in the back” or “blow the whistle” have either a literal or figurative interpretation in the text.
The team has been able to quantify many linguistic properties of idioms and incorporate them into their idiom detection algorithm.
“This work will contribute to the advancement of research in sentiment analysis, opinion mining, machine translation and natural language understanding,” Feldman says.
