Abstract:

The nationality of university students is increasingly diverse. A student might use different languages to communicate with each of their parents and to communicate outside their home, giving them greater scope for understanding information in a range of languages. Nonetheless, in many cases they need to read or write in a language in which they are not proficient or with which they are entirely unfamiliar. In such situations they may use machine translation (MT), which we define as translation from one language into another by a computer program without human input.

Against that backdrop, and as part of its ProjecTA-U project [FFI2016-78612-R], Tradumatica, a research group of the Autonomous University of Barcelona (UAB), is investigating how university students use MT for academic purposes. In the academic years 2015-16 and 2016-17 we collected information from such students through focus groups and semi-structured interviews. In 2017-18 we conducted a case study with the participation of 71 students (future teachers) from the UAB's Faculty of Education. We asked the participants to carry out a standard reading task as part of a subject and recorded the results. The text they were given to read was an MT system's output, although they were unaware of this. The task included answering seven questions related to comprehension and the difficulties encountered.

Our focus groups and interviews showed that students use MT to carry out university work if they have only a minimal understanding of or are entirely unfamiliar with the language involved. In our case study, however, we saw that how successful they are in doing so depends on various factors. We have observed that reading an MT system's output does not always involve an in-depth understanding of the information it contains; in many cases, readers' understanding is merely superficial. In that light, MT, in conjunction with the reader's knowledge of the subject matter, would usually be sufficient to obtain a rough idea of the gist of a text. Understanding more specific details is much more complicated, however, as is interpreting highly imprecise or significantly distorted texts.

Keywords:

Technologies, learning, higher education, machine translation, information access.