Computer scientists at MIT and Israel’s Technion have designed a system which can identify the native language (non-English) of individuals who write essays in English.

Initially, MIT professor Boris Katz asked one of his students Yevegeni Berzak to write an algorithm that could determine the native language of someone writing in English in order to develop a grammar-spelling software. Katz believed such software would be personalized and tailored to a user’s specific background.

With the help of Katz and another professor at Israel’s Technion, Roi Reichart, Berzak has created the system that has gone through more than 1,000 English essays written by native speakers of 14 different languages. After analyzing the syntax of every essay the system compares it with the syntax of the writers’ native languages. After this process, the software comes up with probabilities. For example, it can conclude that there is a 51-percent chance a particular essay has been written by a native Russian speaker, 33 percent –by a Polish speaker, and only 16 percent—by a Japanese speaker.

The researchers also distinguished family trees, so as it is easier to classify the languages.

“These features could be extremely valuable for creating better parsers, better speech-recognizers, better natural-language translators, and so forth,” Katz said.