Using Computers in the Translation of Literary Style

Challenges and Opportunities

Roy Youdale

• Publisher: Routledge
• Year: 2019
• ISBN:
  9780429638497
• Edition: 1
• Topic: History
• Subtopic: Ancient History
• Index: History

This volume argues for an innovative interdisciplinary approach to the analysis and translation of literary style, based on a mutually supportive combination of traditional close reading and 'distant' reading, involving corpus-linguistic analysis and text-visualisation. The book contextualizes this approach within the broader story of the development of computer-assisted translation -- including machine translation and the use of CAT tools -- and elucidates the ways in which the approach can lead to better informed translations than those based on close reading alone. This study represents the first systematic attempt to use corpus linguistics and text-visualisation in the process of translating individual literary texts, as opposed to comparing and analysing already published originals and their translations. Using the case study of his translation into English of Uruguayan author Mario Benedetti's 1965 novel Gracías por el Fuego, Youdale showcases how a close and distant reading approach (CDR) enhances the translator's ability to detect and measure a variety of stylistic features, ranging from sentence length and structure to lexical richness and repetition, both in the source text and in their own draft translation, thus assisting them with the task of revision. The book reflects on the benefits and limitations of a CDR approach, its scalability and broader applicability in translation studies and related disciplines, making this key reading for translators, postgraduate students and scholars in the fields of literary translation, corpus linguistics, corpus stylistics and narratology.

1 Using Computers in Literary Translation

The focus of this book is on combining close reading with corpus linguistics (CL) and text-visualisation tools and techniques in the process of literary translation. The role of this chapter is to actively encourage the opening up of a wider debate about the various ways in which computer technology could be productively incorporated into this process, without involving either deterioration in quality of translation or translator deskilling. The close and distant reading (CDR) approach is contextualised by looking at the development and current usage of the two most widely adopted translation technologies: computer-aided translation (CAT) tools and machine translation (MT). The impact of these technologies on translators is summarised, and evidence is provided of varying attitudes towards their use on the part of both commercial and literary translators.¹ Some examples of the employment of both MT and CAT tools in literary translation are then given to show that these technologies need to be considered as part of the wider debate referred to above, particularly in the light of some current developments.

The third section of the chapter sets out the practical detail of the CDR approach as it relates to computer-aided literary analysis and literary translation. The functionalities of a number of software programmes which can be used in literary translation – CATMA (Computer-Aided Textual Markup and Analysis),² Sketch Engine³ and Voyant Tools⁴ – are listed and illustrated by examples of possible uses. These functionalities are then brought together, listed in a software comparison chart and linked to a provisional four-stage model of the process of literary translation. This shows how the software tools and approaches can expand the range and type of stylistic and narratological information available to the translator by interrogating both the source text (ST) and the draft target text (TT), in addition to their uses in the search for immediate translation solutions for given words and phrases. The chapter concludes by making the case for the relevance of the CDR approach to a range of neighbouring disciplines, and it is suggested that both they and Translation Studies (TS) would benefit from closer dialogue.

1.1 Computers and Translation: An Overview

The Development of Translation Technologies

In a chapter from the Routledge Encyclopedia of Translation Technology (2015) the editor, Sin-Wai Chan, makes the blunt observation that ‘In terms of the means of production, all translation nowadays is computer-aided as virtually no one could translate without a computer’ (2015b:44). He also notes that ‘It is estimated that with the use of translation technology, the work that was originally borne by six translators can be taken up by just one’ (2015b:45). How we have arrived at this situation and an overview of what the impact of technology on translators has been, form the focus of this section and provide a context within which to examine the actual and potential uses of computers in literary translation.

Warren Weaver of the Rockefeller Foundation and Andrew Booth of London University are held to have been the first two scholars to suggest using the newly invented computer to automate translation between natural languages, in 1947 (Chan 2015a:3). Over the next 20 years substantial resources were devoted to developing this idea, but with disappointing results. It was believed by the research funders that the achievement of the stated goal of MT – the ability to produce fully automated, high-quality translation – was unattainable. In 1966 the Automatic Language Processing Advisory Committee (ALPAC) report of the US National Academy of Sciences concluded that there was ‘no immediate or predictable prospect of useful machine translation’.⁵ Although this delayed the progress of MT, it is held to have led directly to the development of CAT for the human translator in the form of the translation memory (TM) (Chan 2015a:5).

MT research was given a boost in the 1970s by the needs of the Canadian and European Community governments to find a way of producing a large volume of acceptable multilingual translations of administrative documents (Qun and Xiaojun 2015:107). In the 1980s the original MT systems based on linguistic rules were supplemented by the use of example-based MT, which took advantage of enlarged databases to allow better automated string-matching searches. This was followed by the development of statistical machine translation (SMT), sometimes called data-driven MT, which is ‘a machine translation system that uses algorithms to establish probabilities between segments in a source and target language document to propose translation candidates’.⁶ Before they can be used SMT systems, like all MT systems, need to be ‘trained’: they are ‘fed’ bilingual parallel corpora of texts related to the field and text-type being translated to analyse. They then search for probable matches amongst these corpora for ST segments (usually sentences), and these searches can be word-based, phrase-based or syntax-based (Qun and Xiaojun 2015:113). Despite this apparent technical sophistication, most MT still requires human intervention in the form of post-editing of translations, which is fast becoming a standard part of commercial translation. Two of the most recent developments in the field of MT are neural machine translation (NMT) – so-called due to an analogy with the human brain’s neural network – which performs analysis and output based on weights and biases; and a growing convergence between MT and CAT’s main tool, the TM (Choudhury and McConnell 2013:53).⁷

Born in a sense from the apparent failure of MT in the 1960s, the development of TMs and other CAT tools was boosted in the 1990s by the desire of corporations and institutions to widen their markets in goods and services (Garcia 2015:68). Whilst varying in the range of tools they offer and how they interact with other systems,

every CAT system divides a text into “segments” (normally sentences, as defined by punctuation marks) and searches a bilingual memory for identical (exact match) or similar (fuzzy match) source and translation segments. Search and recognition of terminology in analogous bilingual glossaries are also standard. The corresponding search results are then offered to the human translator as prompts for adaptation and reuse.

(Ibidem)

While the fundamental principles of TMs have remained the same, there have been significant changes in the context within which they are used and the features they can now offer. These include the following:

• a predominantly web-based rather than hard drive-based working environment, involving translation teams rather than lone translators
• large online databases can now be accessed, such as TDA (TAUS Data Association)⁸ and the Google Translate Toolkit⁹
• term extraction tools can now bring language-specific knowledge to a CAT environment, which was previously like an empty shell waiting to be filled by TMs
• speech recognition text input is now being integrated into some CAT tools
• some sub-segmental matching from internal databases is now possible, as well as the use of predictive typing
• the convergence of MT with TM is growing, with advantages accruing from using a suggested MT version of a segment where there is ‘no match’ in the CAT tool or a fuzzy match below 70% similarity.¹⁰

(Garcia 2015:82–85)

The Impact of Translation Technologies on Translators: Some Key Issues

The discussion so far has dealt with the development of translation technologies in purely technical terms, without regard to the effects these technologies have had on human translators, but a consideration of this is vital in any discussion of technology and translation. Even a brief review of the development of translation technologies will have made it clear that they have transformed the working lives of translators, particularly since the widespread introduction of CAT tools from the 1990s onwards. There now exists a body of studies on the impact of these technologies which reveals a mixed picture (Olohan 2011; Pym 2011; Screen 2016; Kenny 2017; LeBlanc 2017; Vieira and Alonso 2018). In broad terms, as Anthony Pym notes, the impact of translation technology on commercial translation is plain to see:

whereas much of the translator’s skillset and effort was previously invested in identifying possible solutions to translation problems (i.e., the generative side of the cognitive process), the vast majority of these skills and efforts are now invested in selecting between available solutions (i.e., the selective side of the cognitive processes).

(Pym 2013:493, emphasis in the original)

While TMs have undoubtedly increased productivity, improved consistency and eliminated some boring and repetitive tasks (LeBlanc 2017:48), they have also led to some adverse consequences: ‘TMs render the translator’s work more mechanical and, when misused, may lead to deskilling and may have an effect on the translator’s professional satisfaction’ (Ibidem) as well as lowering rates of pay.

According to the UK Translators Survey 2017, involving 588 respondents, 65% of translators used TMs and 22% used MT (2017:26). In relation to TMs, while two thirds rated them as important to their work, there were mixed views on their impact:

I use translation memories when required to do so by clients, but find them of little benefit.

(2017:28)

Many of my (agency) clients would not work with me if I did not use a CAT tool, it is therefore an essential skill to learn.

(2017:41)

With MT, there was more overt hostility: ‘Post-editing, which some agencies use a lot, is not improving quality, just reducing what the translator is paid’ (2017:32). There were also examples of suggested strategies for ameliorating the impact of MT: ‘Machine translation will be the natural evolution of the translation industry, to a certain extent. By fearing MT, we can make it worse. It should be embraced, but sold to clients as a different service to human translators’ (Ibidem). Given the rapid development of both the translation technologies themselves and changes in the contexts in which they are being used and integrated, perhaps the survey’s most surprising finding is that only a fifth of 585 respondents felt that technology would reduce the importance of human translators (2017:39). What is clear is that the future of translation technology is not entirely straightforward to predict, and that it is probably unwise to adopt either what Dorothy Kenny terms ‘cyber-utopian visions of a world without language barriers’ (2016:online) or the sweeping assumption that most translators will soon become just post-editors of MT (Pym 2013:488).

So far we have only looked at translation technology in the context of commercial translation, which is estimated to accoun...

Table of contents

• Cover
• Half Title
• Series Page
• Title Page
• Copyright Page
• Dedication Page
• Table of Contents
• List of Figures
• List of Tables
• Acknowledgements
• List of Abbreviations
• Introduction
• 1 Using Computers in Literary Translation
• 2 Analysing the Source Text: Structure and Style
• 3 CDR, Translation Theory and the Attempt to Create an ‘English Benedetti’
• 4 Applying the Methodology (Part 1): The Translation of Culture
• 5 Applying the Methodology (Part 2): The Translation of Punctuation
• 6 Applying the Methodology (Part 3): Comparing Source Text and Draft Translation
• 7 Applying the Methodology (Part 4): The Auto-analysis of Translator Style
• 8 Conclusions: Assessing the Potential of the Methodology
• Appendix A: Research Data
• Appendix B: Translations used for Chapter 7
• References
• Index