native language

Method determines the native language of foreigners writing in English

Tecno idioma 2adaniel buenoSimply reading a text written in English by non-native speakers of that language might be sufficient to figure out which language is the mother tongue of the writer. Computer scientists from the Massachusetts Institute of Technology (MIT) and the Israel Institute of Technology (Technion) have developed an algorithm that predicted the native language of over a thousand foreign students who took an English test, with a 72% average success rate. The students were native speakers of 14 different languages, including Portuguese. The findings were presented last month at the 18th Conference on Computational Natural Language Learning, in Baltimore. When writing in English, the students used certain syntax patterns from their native language, such as the typical order of subject, verb and object or how to formulate negatives, which are unusual in the language of Shakespeare. The algorithm was not only able to identify these peculiarities, it was able to assemble a tree showing the linguistic proximity of the 14 languages analyzed. “The amazing thing is that our system inferred this tree without coming into contact with a word of those languages,” says Yevgeni Berzak, one of the authors of the study.