The Human Factor in Machine Translation

Sin-wai Chan

• Publisher: Routledge
• Year: 2018
• ISBN:
  9781351376235
• Edition: 1
• Topic: Languages & Linguistics
• Subtopic: Translating & Interpreting
• Index: Languages & Linguistics

Machine translation has become increasingly popular, especially with the introduction of neural machine translation in major online translation systems. However, despite the rapid advances in machine translation, the role of a human translator remains crucial. As illustrated by the chapters in this book, man-machine interaction is essential in machine translation, localisation, terminology management, and crowdsourcing translation. In fact, the importance of a human translator before, during, and after machine processing, cannot be overemphasised as human intervention is the best way to ensure the translation quality of machine translation. This volume explores the role of a human translator in machine translation from various perspectives, affording a comprehensive look at this topical research area. This book is essential reading for anyone involved in translation studies, machine translation or interested in translation technology.

1 Localizing websites using machine translation

Exploring connections between user eXperience and translatability

Lynne Bowker and Jairo Buitrago Ciro

1 Introduction

Over the past two decades, the World Wide Web and its associated navigation tools have developed into a resource that is currently unmatched as a means of providing users across the globe with access to information. Companies and organizations of various types create websites to promote and sell a wide range of products and services, and a significant factor in determining whether or not these efforts are successful is something that has come to be known as the User eXperience (UX).

Hassenzahl (2013) describes UX as an “evasive beast”, noting that experience is an almost overwhelmingly rich concept, with a long history of debate and many attempts to define it. UX is a slippery concept because it focuses on the total subjective experience of the user – such as whether using a product or service produces a positive, a negative, or a middle-of-the-road reaction – and on whether the product or service meets the user’s needs. According to Garrett (2011: 6), if someone asks you what it is like to use a product or service, they are asking about your user experience. This could include questions such as whether it is hard to do simple things, whether the product is easy to figure out, or how it feels to interact with the product (e.g. awkward, or satisfying). Every product that is used by someone – from a toothbrush to a digital camera, from a microwave oven to a pair of shoes, from a book to a website – creates a user experience. Designing for a positive UX has become an important aspect of website development. As more and more companies and organizations compete for business online, users have many possible websites to choose from. Providing a positive UX on a website is therefore becoming increasingly important as a means of attracting and retaining customers.

To date, much of the emphasis of website design has been placed on the visual aspects of a site, such as font size, icons, or colours, as well as on the navigational aspects, including screen layout, scrollability, and hyperlinks, among others (e.g. Lindgaard et al. 2011). Commonly asked questions that relate to UX include whether the site is easy to use, attractive, and appropriate. However, in the context of UX, relatively little attention has been paid to the textual content, even though much web content is text-based.

Some organizations have developed style guides for content creators that suggest the particular tone and voice that should be adopted to appeal to site users (e.g. MailChimp Style Guide, Microsoft Style Guide, WebAIM). However, to the best of our knowledge, a question that has not yet been deeply explored is how the UX of a website’s text influences the translatability of that text, and vice versa. In other words, if a source text generates a positive UX for a reader, how easy is it to translate that text? And if the text is written in such a way as to make it more translation-friendly, how does that affect the UX for both the source- and target-language readers? Bowker (2015) reports on an initial exploration into the relationship between UX and translatability. The results of that study, which involved one text and one language pair (English and Spanish), suggest that UX and translatability have the following relationship: as translatability increases, the UX of source-language readers begins to decrease, while the UX of target-language readers begins to increase (Bowker 2015: 25).

This question of the relation between UX and translatability becomes important for organizations that wish to localize their websites into multiple languages in order to better market their products and services to customers in other countries or regions, where linguistic and cultural differences may come into play. Presumably these organizations would like their websites to generate a positive UX for both the source-language audience and the target-language audience. However, in business, there is also a need to be mindful of the bottom line. Translation and localization have the potential to be time-consuming and expensive undertakings, which means that organizations are also interested in seeking ways to minimize these costs (Lefeuvre 2012; Nantel and Glaser 2008; Spates 2014).

A strategy that can be used to reduce translation costs is to have the goal of translation in mind when creating the textual content of a website in the first place. So-called “writing for translation” or “controlled authoring” involves reducing linguistic ambiguities and simplifying structures in the source text so that it can be more easily translated (Ó Broin 2009; Sichel 2009). This approach is particularly attractive if the source text is to be translated into multiple target languages because eliminating a potential problem (e.g. an ambiguous construction) in the source text is much faster and cheaper than having to address the resulting problem in each of the different target texts (Brown 2003). Meanwhile, another strategy to reduce translation costs for website localization is to integrate the use of technology, such as a machine translation system, into the translation process (Garcia 2010; Jiménez-Crespo 2013). Machine translation systems attempt to automatically translate a text from one language to another; however, professional translators may also play a role, such as post-editing the machine translation output (i.e. revising the draft translation that was produced by the machine translation system). Indeed, these two cost-saving strategies can even be combined. A text that has been authored using a controlled language can often be processed more easily by a machine translation system and will typically result in higher quality machine translation output than would a text that has not been written in a controlled fashion (Clark 2009; Ferreira 2017).

In keeping with this volume’s theme – the human factor in machine translation – this chapter seeks to build on the initial findings of Bowker (2015) with regard to the relationship between UX and translatability when machine translation is used to translate the textual content of a website. UX is an important human factor to consider when contemplating the use of machine translation, and so this chapter begins with a brief overview of UX, examining some definitions from the literature, and focusing in particular on the role of text as part of the UX of a website. Next, the notion of “writing for translation” is presented, including a very brief overview of controlled language and machine translation. Once the key concepts have been explained, we then present the results of a multilingual recipient evaluation. In this study, a group of English speakers act as source language recipients who weigh in on the UX of two different versions of a text that has been written to encourage students to apply to a graduate programme in information studies. Next, machine translations of these two versions of the source text are evaluated by target-language groups with different user profiles, including French-speaking undergraduate students, Italian-speaking professors and graduate students, and Spanish-speaking university library employees. The final part of the investigation introduces post-editing, which is another type of human factor in machine translation. In fact, O’Brien et al. (2014: vii) have described post-editing as “possibly the oldest form of human-machine cooperation for translation, having been a common practice for just about as long as operational machine translation systems have existed.” In the present study, the French versions of the machine translated text are post-edited to see how this might affect the UX of a website.

2 User eXperience

The International Organization for Standardization (ISO DIS 9241–210:2010) provides the following definition for UX: “A person’s perceptions and responses that result from the use or anticipated use of a product, system or service.” This seems to be a reasonable high-level explanation. However, as emphasized by Law et al. (2009: 719), there is no universally accepted definition of UX because this concept is associated with a broad range of fuzzy and dynamic notions, including emotional, affective, experiential, hedonic, and aesthetic variables. Inclusion and exclusion of particular variables in the evaluation of UX is inconsistent and appears to depend largely on the researcher’s background and interest. Law et al. (2009: 719) also observe that the unit of analysis for UX is very malleable, ranging from a single aspect of an individual end-user’s interaction with a standalone application to all aspects of multiple end-users’ interaction with the company and its merging of services from multiple disciplines.

Two other factors that make it challenging to provide a firm and consensus- based definition of UX are that it is subjective and dynamic. UX is subjective in nature because it is about an individual’s perception and reaction with respect to a product or service. Moreover, it is dynamic because it is modified over time as circumstances change (e.g. the user becomes more accustomed to a product, or new features are added).

According to Garrett (2011: 7–8), a distinguishing feature of UX design is that it goes beyond aesthetics or function to deal with questions of context. Often when people think about product design, they first think of it in terms of aesthetic appeal: a well-designed product is one that looks or feels good. Another common way that people think about product design is in functional terms: a well-designed product is one that does what it promises to do. Garrett (2011: 7–8) clarifies that products might look good and work well functionally, but UX design seeks to ensure that the aesthetic and functional aspects work in the context of the rest of the product and in the context of what the user is trying to accomplish. He gives the example of a coffeemaker, noting that aesthetic design makes sure that the button on the machine is an appealing shape and texture, while functional design ensures that this button triggers the appropriate action on the device. Meanwhile, UX design makes sure that the aesthetic and functional aspects work in the context of the rest of the product and the intended task by addressing questions such as whether the button is the right size relative to the importance of its function, or whether it is in the right place relative to the other controls that the user would be using at the same time.

2.1 UX and websites

User experience is a vital part of any product, and this includes websites (Hartmann, Di Angeli and Sutcliffe 2008). Every organization that has a website creates an impression through UX. Increasingly organizations are recognizing that providing a quality UX on a website is an essential, sustainable competitive advantage. As explained by Garrett (2011: 12), UX forms the customer’s impression of an organization’s offerings, differentiates an organization from its competitors, and determines whether the customer will return.

Hassenzahl and Tractinsky (2006: 95) observe that conventionally, UX design has been largely focused on preventing usability problems. However, they go on to make a convincing case for considering a high quality UX of a website as being not merely one where there is an absence of problems, but rather as one where users have an overtly positive experience when interacting with the site.

2.2 Text as an element of UX

A main goal of almost every website is to communicate the information content of the site as effectively as possible. It is not sufficient to simply put the information online; it must be presented in a way that helps people to absorb and understand it. If the information is not conveyed effectively, users might not be able to establish that an organization offers the service or product that they are looking for, or if they only manage to find this information after a challenging exploration of the site, they might determine that if the site is difficult to work with, then the organization might be too. Effective communication is therefore a key factor in the success of the product or service. Creating an information-rich UX is about enabling people to find, absorb, and make sense of the information that is provided on a website.

According to Garrett (2...

Table of contents

• Cover
• Title
• Copyright
• Contents
• List of figures
• List of tables
• List of contributors
• Introduction
• 1 Localizing websites using machine translation: exploring connections between user eXperience and translatability
• 2 Norms and controlled language for machine translation
• 3 Caught in the web of translation: reflections on the compilation of three translation encyclopedias
• 4 A comparative study of term extraction methods in translation
• 5 Introducing corpus rhetoric into translation quality assessment: a case study of the white papers on China’s national defense
• 6 Evaluating term extraction tools: system performance vs user perception
• 7 Terminology resources in support of global communication
• 8 Corpora and CAT-based CN-EN translation of Chinese culture
• 9 Issues of language structure in machine translation between English and Kiswahili
• 10 Quality estimation of machine translation for literature
• 11 Optimising the use of computer translation systems by examining disciplinary differences and characteristics of genres as well as various approaches applied in computer translation
• 12 Crowdsourcing translation in contemporary China: theories and practices
• Index