Machine Translation, NCTA


At the NCTA September meeting, Dr. Anthony Pym discussed his research findings and explained “what happens” when translators work under pressure. BY RAFFAELLA BUSCHIAZZO

The September General Meeting took place on Saturday the 13th in downtown San Francisco and was presented by NCTA President Tuomas Kostiainen. Vice President and Translorial Publisher Yves Avérous offered potential volunteers free training on layout and Translorial blog site management. Then he praised the excellent work Translorial’s new editor, Nina Bogdan, did on the September issue. He also showed everyone how to join the new NCTA Group in LinkedIn, the professional network website. All active NCTA members are welcome to join the group. At the December General Meeting we will present the most popular network websites where you can promote your professional skills online.

Translating faster

NCTA was honored to host Anthony Pym, who presented some findings from academic studies of translation that he and his colleagues conducted at the Monterey Institute of International Studies (MIIS) and Tarragona and Copenhagen Universities. Anthony Pym is head of postgraduate translation programs at the Rovira i Virgili University in Tarragona, Spain, and Visiting Professor at MIIS. He holds a doctorate in sociology from the École des Hautes Études en Sciences Sociales in Paris and has worked as a technical translator for some 15 years. His main research work has been on sociological approaches to the study of translation, especially with respect to the use of translation technologies, and he has written or edited 15 books and over 130 articles in the field.

Industrial work conditions and new translation technologies have made time-on-task a key element in much translation work. We have to work faster, and speed has thus become a factor in many quality equations. Pressure on translators comes from localization work cycles, translation memories, machine translations, teamwork, importance of revisions, etc. Anthony Pym reviewed the main findings of the research in several areas: from the way translation strategies change under time pressure to the proliferation of errors in translation memories. The research projects are mostly European, on a select number of language pairs, usually involving small numbers of subjects and with questionable generalizability. His survey should nevertheless lead to one fundamental question: When time is the key, can we still talk about translation as communication?

Dr. Pym trains his students on how to translate under pressure. There are several elements that need to be taken into consideration when you translate fast: more errors, less communicative quality, more economic rewards, less job satisfaction, and different translation styles.

The translation process

The studies to understand how translating under pressure affects the translation processes use mostly think-aloud protocols (translators reading and speaking aloud, verbalizing what they do), screen-recording (translators recording their voice while translating then listening to their translation), keystroke recording, and eye-tracking (observing the translator’s eye patterns during the translation process with software). Dr. Pym showed us an example of three translators translating the same text in a study at Monterey. He pointed out the varying allocation of resources: some translators spend more time making the first draft nearly complete, whereas others quickly sketch out a basic outline to be filled in later.

“Styles” of translation

Many different variables operate in the translation process. Several studies showed how a long initial orientation phase doesn’t necessarily produce a high quality translation, and that long revision times may in fact lead to lower quality. The translation process is composed of several phases such as interpreting the source text, composing the translation, and then various phases of revisions. Translators have different and personal ways to approach these phases that can be summarized as follows: The “architect” style with minimal revising after the initial drafting of the translation; the “water colorist” with minimal revision; and the “oil painter” style with major revising during and after the initial drafting. You could be any one of them and do an excellent translation.

Another important factor is the differences in how people translate into their first or second languages. Studies show more revision when translators translate into their second language. Also, professional translators finish the first draft of a translation into their first language faster than novices do. Professionals then spend more time on the revision phase, while introducing fewer changes. Under time pressure professionals spend more time looking at target text rather than source text and decrease planning activity, which leads to a more rapid translation process.

Humans are the slowest

Dr. Pym ended his presentation with a discussion about translation memories (TMs) and how they affect these processes. Some studies show that TMs allow faster work and enhance terminological consistency but also propagate any errors. Using Google for an initial translation and then editing the result gives higher productivity than using translation memories for European languages, but not for Asian languages.For technical texts, human translation is the slowest. Translation memory with fuzzy matching is initially fast, but then requires time consuming error-checking to find subtle mistakes. The fastest, oddly enough, is machine translation such as on Google followed by human editing. Why? The initial draft is faster to do than with TM, and the mistakes the computer makes are usually major and immediately noticeable, thus easy to correct.

On behalf of NCTA I would like to thank Anthony Pym for his presentation. He both captivated and informed the audience with the results of his research. RB