In recent years, I’ve read that the quality of machine translation has improved drastically. Nevertheless, I was quite unprepared when machine translation became an issue in my own course on academic reading and writing for international students.
When a recent group of students turned in the first drafts of their first essays, two of the essays included long passages that set off alarm bells. Both students’ paragraphs began with sentences that reflected what I knew to be their usual level of grammatical accuracy, but after two or three sentences, the accuracy increased sharply. I believed that the suspect passages had been plagiarized from an English-language text, but when I talked with the students, both said that they had used Google Translate to create the passages I was asking about. They said that they had written their ideas in their native languages, then used Google Translate to create the sentences that ended up in their essays.
Later in the week, I noticed that one student was looking at two texts on the screen of her laptop: on the left, in English, was an article I had assigned the class to read, and on the right was a text in the student’s native language. When I asked her about it, she explained that the text on the right was a machine translation of the assigned text. I asked if she was using it to check her understanding of the English version, and she said no, she was reading it only in her native language.
Concerned by those developments, I decided to experiment with Google Translate to see if the results my students were claiming could be replicated. I contacted several friends in academe who are non-native English speakers. I asked them to write sentences in their native languages, use Google to translate those sentences into English, and then send me both texts.
I was very surprised by the results. They sent me examples of translations into English from Albanian, Chinese, Kazakh, Russian, and Spanish. The translations from Albanian, Kazakh, Russian, and Spanish were good. They were not perfect, but the mistakes were such that, had I encountered the sentences in a student paper, I would have attributed them to grammatical errors on the part of the non-native speaker of English. Here is one of the Spanish examples:
Original Spanish: Si yo fuera tú, le cobrara hasta por respirar.
Google Translate’s English translation: If I were you, I’ll even charge you for breathing.
Accurate translation: If I were you, I would even charge him for breathing.
Not knowing that the English sentence was a translation, I would see a couple of typical errors and not be suspicious of the source of the text.
In contrast, Google Translate’s English version of a Chinese passage was unintelligible:
But the second language acquisition is a steady process, if you take a cure to do, that is, chronic diseases and even health care, unlike acute disease, once the recipe to treatment, the consequences will soon be reflected.
The accuracy of Google Translate’s Spanish-to-English translation is food for thought. Specifically, it raises the following questions about non-native speakers of English earning academic degrees in the United States:
- To what extent are they required to read assigned texts in English, rather than in other languages?
- To what extent are they required to write their own original texts in English as opposed to writing them in another language and then having them translated, whether by a person or a machine?
- To what extent are they required to speak about assigned texts or their own work in English?
- To what extent is it acceptable for students to cite works written in a language which their professors or teaching assistants do not understand?
For my own course, the answers are simple: The syllabus of my course clearly states that students who pass the course show that they can write an essay in English. Students must also be able to read English academic prose at a certain rate, and show that they have understood the text by immediately answering questions about it with no opportunity to read it in translation first.
I decided to experiment with Google Translate to see if the results my students were claiming could be replicated.
What about the students whose work originally sent me down this path? The quality of the text in their essays that initially set off alarm bells was much higher than the results I got from Google Translate in my little experiment. As a result, I seriously doubt that these students used Google Translate; it was probably a case of garden-variety plagiarism. Fortunately for the students, I require everyone to submit three drafts of each essay. Plagiarism on the first or second draft can be corrected, although plagiarism on the final draft has severe consequences.
I confronted the students about their drafts and required them to rewrite their essays with help from tutors in the writing center. As for reading assigned texts in translation rather than in English, I have explained the drawbacks to my students, but I know I cannot prevent them from making this mistake. I can only make sure that every reading assessment in my class is done under conditions that make it impossible for students to translate the text rather than read it in English.
For academe as a whole, the answer is not as simple. Having taught at several institutions, I’ve encountered a range of opinions about the importance of English proficiency in academic-degree programs in the United States. In some fields, the department conferring the degree is concerned only about students’ demonstrated mastery of content; it is satisfactory if students can demonstrate that they have acquired the expertise associated with the degree through projects, exams, and papers substantially edited by someone else. In other fields, mastery of academic English is generally implicit in the expectations for conferral of a degree.
Like every new technology, machine translation has benefits and drawbacks. It might be tempting simply to ban Google Translate across the board, but such a solution overlooks the complexity of the world. Judicious use of machine translation can certainly support, rather than replace, student effort. Nevertheless, it is high time for academe to establish formal policies about student use of machine translation for coursework. Different disciplines may have different policies, but these policies need to be founded on consensus, articulated clearly to students, and applied consistently.
Dianne Loyet is an adjunct instructor at the University of Illinois at Springfield.