Taiwan International Conference on Translation & Interpretation: Technology and T&I Development, Taipei

Transcription

1 Filip Majcen Director Directorate-General for Translation European Commission Taiwan International Conference on Translation & Interpretation: Technology and T&I Development, Taipei Friday 8 November 2013

2 Translating for the European Union from Past to Future Ladies and gentlemen, It is a pleasure to be here with you today. Let me start by thanking very sincerely the National Academy for Educational Research and the National Taiwan Normal University for inviting me. I would particularly like to thank the Acting Director General of the Department of International and Cross-strait Education in the Ministry of Education, Dr Yang, the President of the National Academy for Educational Research, Dr Ko, and the Professors at the Graduate Institute of Translation & Interpretation at National Taiwan Normal University, Director Dr Liao and Dr Chen, for the hospitality that you have shown. There are almost kilometres between Brussels and Taipei, and it took me more than 15 hours to get here. But when I got the invitation to speak to you today, I was delighted to accept. Why? Because, in today s globalised world, we can shrink such distances when we discuss topics of common interest. The meetings and seminars I have had over the last two days in the Ministry of Education, Fu-Jen Catholic University, National Academy for Educational Research, National Taiwan Normal University and the European Union Centre in Taiwan, have already shown that. Many of the issues we have to deal with in Europe are quite similar to those you encounter in your part of the world. Now this conference is another opportunity to explore those issues together. Taiwan is one of the world s technological powerhouses and a leading player in the information technology industry. It s the perfect place to discuss today s topic: Technology and T & I development. 2

3 Technology has been transforming the translation industry for some time now. And it s made translation far more widely available to the public. Many people who browse the web today can have a page translated from an unfamiliar foreign language into something that they can understand in their own language with a few clicks. Many of us have access to translation in our pockets, embedded in apps, devices and screens. Technology has inevitably had a huge impact on translation in the institutions of the European Union. Before I dive into that topic, let me set the scene. Let me tell you something about the European Union and its languages. Then I ll look at how translation in the European Commission, where I work, has evolved over time. Finally, I want to look at the challenges that lie ahead for us. The European Union and languages My work has taken me to Brussels. But originally I come from Slovenia, a small country in southern Central Europe, next to Italy and Austria, with a population of about 2 million. Despite the small number of Slovene speakers, my mother tongue is one of the European Union s official languages. That puts it alongside far more widely spoken languages, such as English, French, German, Polish, Spanish or Portuguese. Now it s important to keep in mind that Europe, with its 500 million citizens, is a mosaic, not a melting pot, with religious, linguistic and ethnic diversity. Without that diversity, the European Union would cease to exist. The principle of equal value of all languages lies at the heart of the European project. In a democratic and transparent European Union, all citizens and players on the economic and social scene consumers and companies, stakeholders and associations must be able to understand their rights and obligations. They must be able to interact with the institutions without language barriers. 3

4 That is why the first regulation the founding members ever approved back in 1958 defined the official languages to be used for legislative purposes, for its Official Journal and for correspondence with national authorities or citizens. That regulation is still in force. What has changed is the number of official languages. It all started with four languages, French, German, Italian and Dutch. That grew to six, seven, nine, eleven and then there was the big bang of 2004, when nine languages were added in one go. Then came another three, and on 1 July this year, Croatian joined the family. You may have lost count, so let me tell you that the EU now has 24 EU official languages. How on earth do the European institutions cope with functioning effectively on a daily basis? The answer is that we have to be pragmatic. Before I explain how, a quick word about the main European Union institutions. The three main ones are: the Council of the European Union, in which Member States are represented by their ministers; the European Parliament, whose members are directly elected by Europe s citizens; the European Commission, which proposes laws that the Council and Parliament can adopt, and which ensures that Member States apply those laws correctly. Now, it would be unrealistic to expect all internal proceedings in all the European institutions to be conducted in all official languages. 4

5 The European Commission, where I work, employs about people from 28 Member States. They may speak any of the EU s 24 official languages (as well as others). To work together and understand each other, we use three internal working languages: English, French and German. But the legislative and policy documents that we produce have to be published in all 24 official languages. Ladies and Gentlemen, Last year s conference on Quality Enhancement and Professionalisation, heard Marco Benedetti and Brian Fox make a presentation on the Commission s interpretation service, where they are Director-General and Director respectively. Today, it s my pleasure to speak about translation at the European Commission. I work for the Directorate General for Translation. The Directorate General for Translation The Commission s translation service has grown exponentially over the last 55 years. It started with four official languages and 25 staff, translating pages a year. Now we have 24 languages and staff, translating around2 million pages a year. We are the biggest service in the Commission, also the biggest of a galaxy of language services in the EU. There are about translators and support staff, and many more external contractors, working on translating and localising material from or about the EU. The work of the European Commission s translators has changed considerably over recent years, though our core business is still translating legislation. 5

6 Let me put under the spotlight the Internet first. The Commission has set up portals to inform the public, and to help and encourage people to take an active part in discussion and debate on Europe. Our portals are part of the inter-institutional Europa website, the biggest multilingual public service website in the world. This gets about 40 million visitors a month, with many pages available in 24 languages. Managing such a big site and so many language versions is a huge challenge. DG Translation edits and translates web content. But we cannot provide all information in all languages, as we have to strike a balance between costs and benefits. So we make our decisions according to factors such as: the kind of information and the target audience. Is it general or specialised?; its 'shelf-life'. Is it short-term or long-term?; the urgency and relevance of the material. How important is it, how will it be used? We have given special attention to niche products such as press releases, to make them user-friendly for readers who pick them up on line or on the go via mobile phones. Our colleagues are currently working on a Commission-wide project to rationalise what has become a gigantic website. The aim is to make it more relevant, coherent and affordable. We advise other departments in the Commission what language to use on which part of the website. Turning to products and services other than full-length translations, we have a host of these. We have produced glossaries and terminological aids. We can provide oral or written summaries on request. We have a hotline for urgent translations of short texts. We can provide subtitling. We can offer machine translation, about which I will say more later. We also translate classified texts. 6

7 Most originals are written in either English or French by people who are neither English nor French speakers. That is why we offer editing and linguistic advisory services to other Commission departments. Formal editing is also a good way to save 23 translators from wasting time trying to decipher an unclear text. Informally, translators contact the author to clarify ambiguities and improve quality. The Commission, with my Directorate-General in the lead, is running a campaign for clear writing throughout the organisation. We publicise the principles of clear writing and help Commission staff to put them into practice. We also have translators stationed in each of the Member States, one per country. They work on localising texts, adapting the message, especially press material, to the local context and a specific target audience. In recent years, the Commission has become more active internationally. Demand for translation to and from non-eu languages such as Russian, Arabic and Turkish has grown. We also provide about 900 pages a year from and into Mandarin, most of it translated by freelancers. Though our output is huge, we cannot satisfy all the requests we get for translation. So we have to set size limits and priorities. Our translators cover around 75 % of our output, and the rest, currently almost pages per year, is contracted out to freelancers. By involving them, we also help to develop the market for translation in the 28 Member States. Now let me turn to the changing role of our translators. This conference focuses on technology, which is becoming increasingly important in our work. Without it, managing our output would be 7

8 impossible. But let me stress that the human touch is vital. Only humans can deliver real quality. Only a native-speaker s use of language or choice of register can guarantee the end product is fit for purpose. So, our translators are our most important resource. And as the world around us is changing, the role of our translators is changing too. They need to acquire skills beyond that of translating. Most importantly, they need to be flexible. They have to be willing to use technologies which are having a strong impact on their profession. They need to be willing to perform new tasks. Translators are now communications experts in a multilingual organisation. Our translators are revisers, evaluators, linguistic counsellors. They provide targeted advice for authors, specialists working for other Commission services. They help develop, test and roll out new translation tools. The EU s areas of responsibility have increased considerably over time. Translators have had to keep up with these changes to be able to handle material in a multitude of fields. They may be asked to cover topics ranging from detailed technological descriptions of nuclear power plants or economic and fiscal policies, to brochures for children about the importance of a healthy diet. As an organisation, we constantly aim to achieve quality in every step of the translation process and the end product. Our translators have the important task of ensuring the quality of translations through terminological consistency and multilingual concordance. 8

9 Translators roles have multiplied, and new technology now helps them to fulfil them all, while adding flexibility to their working lives. Many now choose to work part of the time from home, as teleworkers. Now for another aspect of the human touch. When we have a clear idea of what our customers need, we are able to give them a better service. So, last year we launched a new customer relations approach. We get and stay in close touch with the authoring services. We inform them about the services we provide, and our constraints. We get to know their needs in terms of translation and other language products and services. This enables us to anticipate and improve forward planning. Technology, in the form of electronic content management systems, enables us to plan and process our workflows, both for translations carried out within DG Translation, and for those sent to external contractors. Ladies and Gentlemen, Given the number of official languages we now work with, technology and translation tools are vital. Let s take a look back at how things were at the European Commission, and how far they have developed over the years: Tools When the Commission s translation service was first set up, translators used a Dictaphone to record their translations. The secretaries struggled with typewriters to get the translations onto multicopy paper with the layout of the source text. Changes were made with the help of a pair of scissors, sticky-tape and Tipp-Ex. Each language department had a well-equipped library with books, magazines and paper cards with terminology. It was heavily consulted. 9

10 Our document workflow was managed on a board with little coloured cards indicating work allocation and progress. We have moved from these boards to an electronic document workflow management; and from typewriters to word processors, computers, speech recognition, computer assisted translation, and machine translation. Now we are at the cutting edge of translation technologies: both a consumer of these tools and a driver for their development. ICT tools help us deliver consistent translations of good quality and to increase productivity. Let me run through the main milestones in the introduction of IT in our work: In 1991, there was only embryonic IT, consisting of word processing, terminology and databases with legislation. In 1994 to 1995, all translators were equipped with a PC. In 1995, we launched a huge central translation memory, called Euramis. This made genuine data sharing possible between translators in all EU institutions. It made it possible to reuse past translations. The memory is not used directly during the translation process. It is merely a database layer which is accessed to retrieve or store data processed locally by translators. In 1996, we introduced a document workflow management tool which became indispensable. Because of the sheer number of translation requests and languages, we needed to organise the work and manage the workflow more efficiently. 10

11 In 1997, it became possible for translators to share and consult all translations through an electronic archive, which became essential for documentary and terminological search. The same year, the first computer-assisted translation tools, now the backbone of all translation activity, were deployed in pilot projects. CATs had arrived. In 1998, translators got general access to Internet. And, because much of the information translators needed in their work was available on the Internet, we created a tool to facilitate the organisation of links to these new external sources of information. Today, this tool is fully user-provided. Translators add links and maintain the resource collectively. In 2000, CAT solutions became available to all of DG Translation s translators, who were relieved of the more repetitive part of their work. The same year, we also got full electronic management of translation requests. All originals arrived in electronic form, which made electronic processing of translations possible. And we developed tools for communication between translators working with the same document at different stages or into different languages within our service and with translation services in other EU institutions. Then in 2002, we created a one-stop-shop for terminology, which draws on the contents of about 30 terminology sources in the time it would normally take to search a single source. A button on the toolbar in Word makes it possible to launch searches for selected terms directly from Word. 11

12 In 2005, the Euramis central memory was used globally and we had an automated, central electronic pre-processing. We can say that computer-assisted translation became a reality. Euramis is today the bread and butter of daily life for our translators and currently contains more than 205 million phrases in all official EU languages. The same year, 2005, the joint terminology database of all the EU institutions, called IATE (Inter-Active Terminology for Europe) became fully operational. And in 2010, the Commission decided to adopt and adapt to its needs the open-source Statistical Machine Translation System Moses which is now state-of the-art and the most widely used MT system. Let me come back to those two milestones in just a little while. In 2012, we completed the development of a module in the Commission s Corporate Web Content Management System that automated the transmission of translation requests and the import of finalised translations on the web page. The most recent IT milestone is the migration of all our staff to a new CAT tool. We completed this operation, unprecedented in its scale, just one week ago. Here are some figures: staff and work stations were migrated over a period of six months. You can imagine that preparations had to start well in advance (almost two years ago) and that many translators had to take on additional responsibilities as testers, master trainers, trainers, trouble shooters etc. 12

13 On the technical side, integration of an off-the-shelf tool into our existing IT environment (of workflow management, translation memories and terminology databases),was and still is a major challenge. Several developments and a lot of fine-tuning are still under way. On the organisational side, we had to deal with the usual unease about the change, which varied across departments, age groups, and so on. There are significant differences between the new tool and the old one. For instance, content and format are dissociated now, meaning that translators do not work with the original documents in Office format, but rather with raw texts without formatting. We had to plan adequate training and helpdesk/troubleshooting support to ensure business continuity without putting the quality and efficiency of our work at risk. So, now a couple of words about our interinstitutional terminology database IATE: It builds on Eurodicatom, one of the first on-line language-oriented databases in the world, first set up in 1973 with data entered on punch cards. The advantages of IATE compared to other sources of terminology is that it contains data on all of the European Union s fields of activity and that it is highly reliable. Mother-tongue terminologists validate all entries to guarantee quality. IATE has been publicly available on the Internet since 2007, covering all 24 official languages, plus Latin. Internal users also have access to non- EU languages, among them Russian, Turkish, Arabic, and some 300 terms in Mandarin. 13

14 With about 1.5 million phrases, 9 million terms and half a million abbreviations, IATE has many users, among them translators in the EU institutions and our freelancers. The daily number of queries normally ranges between and , some also from Taiwan. The record for a single day is almost searches. When it comes to machine translation, the European Commission was in fact among the pioneers, having funded research in this area since the 1970's within its Framework Programmes on Research and Technological Development. The system then used proprietary technology based on grammatical rules and dictionaries. But after almost 25 years and considerable investment, the service only offered 28 language pairs, covering just nine of the then 23 official EU languages. When we took on nine new languages in one go in 2004 and three more less than three years after that, it became clear we needed a more advanced system. Now we have it. The service is called MT@EC and is provided by DG Translation. Developing MT@EC is a European Commission policy objective, and financed by public funds. Our aim is to make it a standard tool for our translators, who use machine translation output as one of the language tools at their disposal; all staff in the EU institutions, and soon Member State administrations, which will be able to use it as a self-service system, offering raw translations. It offers a total of 552 language combinations between all 24 official EU languages. 14

15 We officially started providing the service to all Commission departments in June We are about to start offering it to public administrations in all 28 EU Member States. uses Moses the world s most heavily-used open-source system to create machine translation pairs with statistical methods by crunching millions of sentences. Why have a machine translation system of our own, instead of simply using freely available tools on the Internet? The added value for us is: First, a matter of quality: we exploit our own linguistic expertise and build on our own corpora; Second, it ensures we remain independent and retain ownership of intellectual property; Third, for confidentiality and security reasons. Many EC documents are not allowed outside the Commission s firewall, not even for translation purposes. What is needed for MT to become successful is loads and loads of data. And that is another of DGT's major assets: multilingual documents that have been stored over almost 2 decades in an ideal format to be used to build MT engines adapted to the translation of EU legislation and documentation. So, translation memory databases that had originally been developed in the 1990's, to allow the development of what later became statistical machine translation, are now used for that purpose. Our in-house input is supplemented by translation memories from other EU institutions, which makes Euramis one of the biggest translation memory systems worldwide. By the way, these memories are available to the general public. 15

16 Building on a constant output of high quality human translations, MT@EC is trained to understand and translate EC documents and to keep improving. We have the right people for this challenging task. Not just the right language experts, highly-skilled translators in all of the EU languages, but also the right technology experts. We have a team of computational linguists with specific expertise in statistical machine translation who work hand-in-hand with translators. Machines cannot replace people, and they rely on people s expertise to ensure a continuous cycle of improvement. That was just a brief survey of the main IT changes within our organisation. Now let me turn to our cooperation with the world outside. How are we responding to new challenges beyond the EU institutions, in our dealings with the outside world? We are trying to master the changing context, and to influence it, by means of our outreach activities, connecting with the world around us. This applies to terminology, for example. And we need to keep informed about technological developments. Our partners, our stakeholders are universities, research centres, language service providers, translator associations, language technology firms. We work with them to make the language industry as a whole stronger, and to position us as a centre of excellence. We want the language industry to develop its employment potential, which in the current economic context climate is becoming ever more important. There is a common interest both for public and private translation providers in developing the educational attainment and skills of language 16

17 professionals. One of our responses is to promote the European Master s in Translation, which has been set up to foster cooperation between universities and translator training institutes around a common curriculum. In recent years, we have moved from being an inward-looking service to one that is well connected and which attaches great importance to relations with groups of language professionals and translation services in EU Member States, but also in other parts of the world. We are involved in international language and translation forums, and we have well-developed bilateral cooperation with many countries. We very much appreciate the cooperation we have with Taiwan, which began in Since then, we have had continuous contacts with a number of players, including educational institutions and the language industry. We also have other contacts in your region. Taiwan, China, Japan and Korea have all expressed a common need to translate legal information into English. That is why one of my colleagues went to Japan some months ago to give a detailed overview of the practical translation process in the Commission. We hope such Asian-European cooperation will continue in future. Now let me try to sum up our story. Since the European Commission s Directorate-General for Translation was set up almost 55 years ago, it has grown exponentially. 17

18 In the process, technology and translation tools have become increasingly important. There are four main reasons for this: The sheer size of the operation has required increasingly sophisticated tools to help organise the work efficiently. The ever-increasing complexity of the subject matter and the large number of language combinations to be handled have created a need for new tools to assist translators, enhance quality and ensure language concordance. Cost considerations, as well as supply and demand constraints have pushed the Directorate-General for Translation to invest more in technology to improve productivity and cost-effectiveness. Last but not least, information and communication technologies have created demand for new products and services and have influenced customer expectations. We can expect these trends to continue in the years ahead. Demand for translation will continue to grow. New countries may join the European Union, adding new languages. Digitalisation in society at large and greater use of online services will also grow. Our ability to deliver will depend to a great extent on us making the right technological choices. On the one hand, for technology to give the best results, it will have to be properly integrated in the translation workflow. To take one example, so far, statistical MT has focused mainly on providing translations for gist purposes, rather than outputs that minimise the effort of translators. The European Union is funding research projects like Mate CAT and CasmaCat that aim to push what can be considered the new frontier of 18

19 CAT technology. These projects aim to facilitate translators' work by developing methods that: Allow the MT system to learn in real time from translators' corrections and feedback (User Adaptive MT); Supply translators with additional information to enhance productivity (Informative MT); Only display MT output if the system considers it is of sufficient quality, thereby promoting translator s uptake of MT. On the other hand, we will have to fine-tune the integration of technology into the services we offer. The MT we provide is a building block for a number of multilingual online public services. Today, for instance, we have the Internal Market Information System which enables national, regional and local authorities to communicate quickly and easily with their counterparts abroad in all official EU languages. Tomorrow, we could be looking forward to e-justice, e-procurement, e- health, and so on. The Commission is investing in further research in this field so that we can continue to develop this infrastructure and widen the range of services we can offer. One of the Commission s objectives for the future is to provide professionals and the public with new tools to create, exploit and preserve all forms of digital content in any language. As a supplier of EU translation aiming to offer good value for money, we will have to step up our efforts to approach translation and technology in a holistic way and not just focus on individual parts of it. 19

20 We have to bear in mind the people involved, our institution s needs, EU citizens needs. We have to bear in mind the media we are translating for, and the type of content. We have to bear in mind the mix of resources and tools we have at our disposal. We re committed to developing an MT service for the future that learns from our translators consistently. Human translation is and will stay our core task. There is no risk that in the future translators will turn into just post-editors of machine translation. They will remain at the crossroads between all technological changes, innovation and the documents to be translated and they will adapt themselves to take advantage of all the tools that are being developed. Human beings, human translation are the ultimate warranty of quality and legal certainty. I am sure today s conference will give us all new insights. I hope it will enable us to anticipate how best to develop the interaction between humans and technology. I look forward to gathering new ideas about what we can do and what we should do to achieve the results we want! Thank you for your attention. 20