Best Thesis - 2007

Federico Gaspari

-The Role of Online MT in Webpage Translation

Abstract: 

This thesis is concerned with the use of online machine translation (MT) services to translate web-based text and enable the production and management of multilingual content on the Internet. The perspectives of interlingual dissemination and assimilation of digital information are both considered, to address the key challenges of using MT in the online environment. Adopting a language-neutral and system- independent approach, this thesis examines in depth a range of under-researched questions regarding the successful deployment of web-based MT software in a number of scenarios, with a view to raising the profile of MT on the Internet.

After the introduction, the literature review in chapter 2 includes contributions published mostly over the last ten years that deal with the topic of online MT from a variety of angles. Chapter 3 presents a questionnaire-based survey which investigates the actual usage of a range of free online MT services among 104 UK-based university students, and discusses the most widespread and unanticipated usage patterns that emerge from the data. Having identified some areas where further research is needed, we put forward some suggestions regarding how the user interface of web-based MT services could be enhanced.

Chapter 4 reports a case study considering the role played by free online MT services for the purpose of information assimilation, in which subjects have to find some information from five foreign-language websites. Two interesting findings are revealed: firstly, Internet users who take advantage of MT tools to extract basic information from a monolingual website available only in a foreign language actually tend to perform worse than the control group, who carry out the same task without the help of translation tools, despite not knowing the language in question; and, secondly, in spite of this worse performance, individuals using online MT consistently have a higher confidence in the accuracy of their answers, compared to the control group who do not use web-based MT. This suggests that users perceive Internet-based MT as a reliable tool to assist them in the assimilation of multilingual information for everyday real-life translation tasks on the Internet when they encounter language barriers.

Chapter 5 focuses on a redesign project during which an online MT facility was embedded into an existing monolingual English-language website to automatically disseminate its content in at least ten other languages, avoiding the need for human translation and localisation. The main technical and practical issues involved in this MT-based strategy to produce and manage multilingual web content are illustrated by four prototypes that were designed for this implementation project. These four prototypes were tested for usability and their effectiveness was evaluated by a group of 72 multilingual international users, who provided information about their design-related preferences. The results of these experiments are analysed in detail to suggest which implementation strategies are most successful and which approaches to the design of an online MT facility are most appealing for international users visiting a website powered by Internet-based MT.

Building on this work, a follow-up project covered in chapter 6 sought to address a number of issues flagged up by the previous user testing and usability evaluation. We present a set of ten user-centred and usability-oriented heuristics that we have developed according to well-established principles in human computer interaction and web usability, which can help web designers to successfully integrate online MT facilities into the interface of monolingual websites. The chapter explains how these heuristics can guide the whole life cycle comprising the design, development and implementation of websites powered by free online MT. In addition, we show that the heuristics can also be used to conduct thorough diagnostic evaluations of the success with which monolingual websites offer online MT for the benefit of their multilingual visitors. We mention possible improvements to extend the heuristics, so as to address areas that are relevant to online MT but are not covered at the moment.

Finally, chapter 7 draws some general conclusions and summarises the main contributions of this thesis. A number of open questions and issues that have been identified during the course of this research are discussed and some areas worthy of future work are outlined.