Computer-Assisted Translation (CAT) Tools in the Translator Training Process

ŁÓDŹ Studies in Language

Edited by

BarbaraLewandowska-TomaszczykŁukaszBogucki

Editorial Board

Piotr Cap (University of Łódź, Poland)

Jorge Díaz-Cintas (University College, London, England)

Katarzyna Dziubalska-Kołaczyk (Adam Mickiewicz University, Poznań, Poland)

Wolfgang Lörscher (Universität Leipzig, Germany)

Anthony McEnery (Lancaster University, England)

John Newman (University of Alberta, Canada)

Hans Sauer (Ludwig-Maximilians-Universität München, Germany)

Piotr Stalmaszczyk (University of Łódź, Poland)

Elżbieta Tabakowska (Jagiellonian University, Kraków, Poland)

Marcel Thelen (Zuyd University of Applied Sciences, Maastricht, The Netherlands)

Gideon Toury † (Tel Aviv University, Israel)

VOLUME 58

1Introduction

Translation has been employed in various forms since the dawn of civilisation. The very first organised human cultures needed translation, initially oral and then written following the invention of writing systems. Eugene Nida (1988: 23) associates the beginning of translation with Septuagint, which was probably the first translation of Hebrew Old Testament into Greek (ca. 3rd century BCE). Douglas Robinson (1997, 2002: 7), on the other hand, finds beginnings of translation in techniques employed by Marcus Tullius Cicero (106–43 BCE) in De optimo genere oratorum. Cicero translated two Greek letters into Latin, thus explaining the process:

The demand for translators resulted in a more conscious approach to translator education. Anthony Pym (2009: 1) suggests that origins of more extensive training programmes could be found in the “elaborate Chinese institutions for the translation of Buddhist texts (4th to 9th centuries), in the “House of Wisdom” in the 9th-century Baghdad, or in cathedral chapters as in the 12th-century Toledo.” Conflicts fostered translation since it was necessary to understand one’s enemy. Even European colonists in America employed basic translator training in order to train captured natives to serve them as interpreters and help them in contacts with the local population. One common thing was that translator training was carried out locally, at the place it was needed, which meant that state borders, places where civilisations met, witnessed more rapid development in the field. Again, Pym (2009) names such examples as schools training French interpreters in Constantinople (1669), the Diplomatic Academy of Vienna (1754), Al-Alsun school of translation in Egypt (1835) or an increasingly fast development of translation schools in China (19th century), with Yan Fu in charge of several translation schools in China, starting from 1896.

It is worth noting that initially translator training was state-controlled to ensure both translation quality and allegiance of translators, especially in Europe (see Caminade and Pym 1998; Pym 2009); however, such attitude was not uniform ←11 | 12→around the world. On the other hand, in Spanish America “translator training was more closely related to sworn translation which was a tool used to maintain juridical regime, employed in turn to control colonies” (e.g. Translation curriculum at the Law Faculty, University of Uruguay, 1885) (Pym 2009). It shows that translator training programmes were very diversified, depending on local political or institutional requirements.

The 20th century and events like the World War II (the Nuremberg trials in particular, which hinted at the future role of interpreters in international institutions) emphasised the importance of translation in national and cross-country communication. The first programmes aimed at “training professional translators and interpreters were introduced at the University of Geneva, Switzerland, in 1941, Vienna, Austria, in 1943, Mainz-Germersheim, Germany, in 1946, and Georgetown, USA, in 1949, for example” (Schäffner and Adab, 2000: vii). The start of the Cold War was a real turning point for Translation Studies. Suddenly everyone realised the import of knowledge that could be gained by translating foreign data, like military strategies, plans, and other classified information. The need to be able to translate vast quantities of data in short periods of time led, at least partially, to Warren Weaver memorandum in July 1949. The memorandum was crucial for the development of machine translation (MT) as it outlined aims and methods of linguistic research in the field of MT way before computer capabilities became known to researchers, first in the USA and then around the world (see MT News International, no. 22, July 1999).

The rise in general awareness and need both for regular translators and MT led to the creation of specialised academic translation programmes. Those were meant to produce highly skilled individuals who could provide translation services in the civilian world on the one hand, and research and develop suitable MT solutions on the other. In 1954, Georgetown MT research team presented their working MT system (see Hutchins, 1994). In 1964, ALPAC (Automatic Language Processing Advisory Committee) was formed to study MT and evaluate overall progress in computational linguistics. The committee had a very negative impact on MT in general. Over the next 20 years the research in the field was severely hampered but ultimately led to the development of, among others, translation memory technology, first commercially employed by SDL in the 1990s.

The European Union and its need for legislation and other documents to be rendered in the official languages of all member states significantly increased the demand for both highly skilled human translators/interpreters and automated translation systems. However, even more important drive factors behind this growth of importance of translation included a total overhaul of economies of developed countries in the 1980s (privatisation and deregulation) (see Castells, ←12 | 13→1980), the “advent of computer technology (silicone semiconductors and computer on a chip), and emergence of computer software” (Cronin, 2003), which made it possible to develop computer-assisted translation (CAT) tools.

What is more, translation as such started to be seen in terms of a vocation, which facilitated integration of translator training programmes into university structures. According to Pym (2009), unemployment amongst students resulted in increased demand for translator training programmes since young people saw this as an opportunity to find a job, even though there was no rise in demand for full-time translators and interpreters. There are countries, though, where demand for quality translators exceeds the capacity of an education system, e.g. China (see Schäffner and Adab, 2000; Pym, 2009). In the end, and thanks to the access to all kinds of computer tools facilitating translation, a significant portion of all translation traffic is translated by underqualified people, prone to errors and major setbacks. Data from a number of European Commission surveys show that freelance (self-employed) translators constitute about 74 % of the total number of translators in Europe. Apart from Slovakia, no European country regulates the profession of translator. Therefore, even an unqualified person can work as a freelance translator. Some states do impose specific requirements on sworn or authorised translators, but those requirements vary from state to state, if they are in force at all (see Pym, Grin, Sfreddo, and Chan, 24 July 2012).

The book will discuss student–computer interaction in the context of CAT course. An attempt will be made to show that the currently employed CAT teaching methods may result in students relying too much on the CAT software, and that the incorporation of problematic aspects of external data verification leads to increased awareness and better quality of student translations.

The primary goal of the book is to diagnose how CAT tools work and to identify students’ needs in regard to CAT skills acquisition process. What is more, the book will attempt to define teaching goals for the translator training process in the context of the use of CAT tools.

CAT tools are state-of-the-art computer programs that allow human translators to translate faster and with better quality (Bundgaard, Christensen and Schjoldager, 2016). However, their advantages are to be clearly seen in the case of certain types of (mostly scientific and technical – see Fernández-Parra, 2010) texts due to the fact that such texts follow similar style and structure patterns and, as a result, are repetitive to a certain degree. In order to stay competitive on the market (the need to work with client-provided translation memories; optimise work time to translated pages ratio; and increase output volume and rates per page), a translator is left with no choice but to utilise the latest translation technology (Pym, 2009; Christensen and Schjoldager, 2016). ←13 | 14→A variety of different tools is available on the market at the moment of writing this book (2017). They include advanced formatting-capable word processors, TM-based solutions, automated glossaries, machine translation, corpora, and more. A successful translator is required to know and accommodate these resources in order to achieve best possible results.

However, one has to be aware of the dangers the technology brings about and know how to negate them. The quality of contemporary computer resources may lead to misplaced trust in the feedback obtained from them, occasionally resulting in lower quality translation. Such over-reliance brings about questions whether computers could replace humans in translation. In fact, some researchers agree that while combination of various computer resources “can be expected to replace fully human translation in many spheres of activity” (Pym, 2012, p. 487). At the same time, translators are expected to become post-editors, refining computer translation (idem.). Even today some types of specialised texts allow CAT users to revise and approve automated translation output, providing their own translation only when necessary.

The book assumes that over-reliance on computer resources, especially in the case of CAT tools, may lead to [serious] errors in translation (see Doherty, 2016), especially in the case of students, graduates and less experienced translators in general. In order to prove this theory, a research has been conducted, aiming to show that the users of CAT tools are prone to accept proposed translation with no proper revision, even though these suggestions may contain errors¹. Such errors may result from a number of reasons. These may include, e.g. incorrect context, or previous translation errors in external resources (Barbu, 2015), i.e. translation memories (e.g. MyMemory², which can be accessed directly from many CAT tools via proper plugins, e.g. in memoQ³; or resources obtained from ←14 | 15→translation service ordering parties [TSOP]). The research focuses on the errors found in the TSOP-delivered translation memories. It analyses how they were processed by test subjects. Finally, the book summarises conclusions to be drawn from the study in relation to the translator training process.

Secondly, the main research is conducted in order to determine how trainee translators deal with the problem of faulty external resources in a live project. The research was conducted in two stages. The first stage took place in 2015/2016 academic year and involved a group of 22 test subjects. The second stage took place in 2016/2017 academic year with 18 participants. Each stage was divided into two phases in order to check initial and final CAT skills after a course on CAT tools. Both test groups included 2nd-year MA students undergoing CAT programme during their final semester at the University of Łódź⁴. The first group learned how to operate the software. In addition to the content covered by the first group, the second group discussed dangers posed by CAT tools and the methods of avoiding them in the process of CAT.

While translators have no influence on the quality of received translation memories, they do have full control over what will be done with the problem during the translation itself. The study assumes that a significant portion of the previously committed errors is retained in translations performed by CAT mechanics-oriented (see Chapter 5.3) trainee translators (I stage) to a much greater extent than in the case of CAT specifics-oriented (see Chapter 5.3) trainee translators (II stage). Such result would imply that it is not enough to understand how to operate a CAT tool since it will not yield the desired outcome. Therefore, it is critical to learn how to use advantages of CAT tools and consciously avoid any disadvantages they may have. Developing a set of best practices regarding the practical use of CAT tools may significantly improve the overall translation quality through the reduction of the risk of committing an error.←15 | 16→