Integrating the Elements: The Current Integration Process among Translation Tools and Translators

In the orchestra of modern translation workflow, translators are the musicians and translation technologies, represented by MT (machine translation), TM (translation memory) tools, and corpora are the various instruments. In order to produce a symphony that resembles the sound in the composer's mind, all the instruments must be properly utilized and the good performance of musicians is also vital. Likewise, in the translation process, to ensure a high quality product, making good use of different translation tools and incorporating the human factor is the key. Different translation tools and trained translators have largely contributed to the advancement of the language industry on the whole and now these elements start to cooperate more and more seamlessly. It happens very much like the author's experience with his PC: The author bought a PC a few years ago and it had been a good assistant and companion to him but as new software became more and more demanding it became slower and slower. Then he decided to upgrade the computer. The plan was to replace the meagre 64 MB memory with a larger one and the old CPU with a Pentium 4. However, some computer expert informed him that if he set aside a larger portion of the hard disk as virtual memory, a new memory would not be needed and if he could increase the frequency of the CPU by switching some buttons on the motherboard a new CPU would not be necessary. The author thus tried these methods. Surprisingly, the performance of his old system had increased tremendously without spending a penny. The translator's role in modern translation process can be compared to the CPU; TM, memory and MT, hard disk. They appear to be different parts in a system but they can actually help and more importantly cooperate with each other to improve the overall performance of the whole system. During recent years, there is an observable trend that different translation tools and the human factor start to integrate in both simple and sophisticated ways. There are many projects having already adopted this concept and try to combine and configure these elements to work together and many others are underway, ushering in a new age of translation.

The major elements come into play in modern translation process include MT, TM tools and sometimes corpora and last but not the least, the translator.

MT has opened a possibility for high speed and low cost translation. Warren Weaver wrote a memorandum in July 1949: If one examines the words in a book, one at a time through an opaque mask with a hole in it one word wide, then it is obviously impossible to determine, one at a time, the meaning of words... But, if one lengthens the slit in the opaque mask, until one can see not only the central word in question but also say N words on either side, then, if N is large enough one can unambiguously decide the meaning. . . Derived from the simple concept of Warren Weaver's, MT has gone through a long and winding road. Burgeoning in 1950s, "bubble burst" in 1960s in the U.S with the publication of ALPAC, resurgence in 1980s in western Europe and Japan and flourishing again during 1990s. At first it was used for gathering intelligences; later its application had widened to many other areas including commercial translation. There is a great debate in the academia about MT, which is manifested by the stark contrast between the optimistic prediction of Raymond Kurzweil in his The Age of Intelligent Machines that "that language is no less complex or subtle a phenomenon than the knowledge it seeks to transmit" [1] and "by 2012, machine translation will be powerful enough to dominate the translation field." [2] and Bar-Hillel's theory, in A Demonstration of the Nonfeasibility of Fully Automatic High Quality [3] . Some scholars argue that it seems all the MT theories are dead-end approaches because no significant improvement has occurred. [4] However, in fact, MT is improving fast with the addition of a larger collection of terms and improved algorithms. Judging by the current development, on the contrary, MT will, if not dominating, play an important role in the translation process. The top five reasons for using MT given by John Hutchins are, sometimes the sheer amount of work is too much for human translators; some materials are too boring for human translators; MT can better keep consistency; it provides increased speed; it can lower the cost. The chief advantage of MT is it saves the analysis and input time and it also provides the translator a quick solution to terminologies although it is far from perfect as for now and there are many problems to be solved.

Contrary to the computer engineers' initial predictions that if the dictionaries were large enough and the lexicography good enough, then the programs would be able to do quality translation'(Schank&Kass,1988:182), the natural languages we speak and write, with all their exceptions and ambiguities, their subtleties and idiosyncrasies, are far more complex than this statement suggests and have proved to be beyond the capabilities of computer technology. [5] Thus, MT has largely given up its ambition to replace human translators completely but focuses on how to aid human translations.

TM tools help improve translation efficiency by "remembering" translated phrases or sentences (segments) for future use. They do not only save the translators from repetitive inputting but also provide terminology management and file analysis functions. According to Melby, the concept of TM originated in the 1970s with its subsequent implementation in the 1980s. [6] Now it has seen its wide applications in various translation tools. A TM is a type of linguistic database that is used to store source texts and their translations. The texts are broken down into short segments that often correspond to sentences [7] . These aligned segments can be reused if the same or similar sentences appear again.

According to the LISA report of 2004 the adoption of TM tools is increasing rapidly:

1. The majority of companies are planning to extend their use of TM technology

2. The majority of those using TM tools use it for at least half of their localization needs

The new trend of TM is the concept of 'text memory' rather than just translation memory [9] incorporating XML technology. This enables a wider application of TM tools.

The advantages and disadvantages of TM are discussed in terms of time and quality. TM saves translator's time by reducing the time reproducing the same or similar contents but it takes time to build up a TM that is large enough to be particularly useful; TM helps keep consistency of translations thus improve the translation quality but it is only useful when the previous translation was correct in the first place.

Corpus as a tool primarily for linguistic studies has been adopted more and more in translation studies as well as in practices.

In the language sciences a corpus is a body of written text or transcribed speech which can serve as a basis for linguistic analysis and description. [10]

The rationale behind it is well explained by Isabelle: Given the staggering volume of translations produced year after year, it is quite obvious that existing translations contain more solutions to more translation problems than any other existing resource." [11]

There are many kinds of corpora such as monolingual corpora, multilingual comparable corpora, aligned parallel corpora etc, among which aligned parallel corpora are of special interest to translation process. Parallel corpus is a bilingual corpus that contains source texts and their translations. [12] From the first Hansard Corpus compiled by Bell Communications Research and the IBM T.J. Watson Research Center in 1980s to the European Corpus Initiative (1992), to MULTEXT and MLCC (1994) to many other projects being carried out in recent years, parallel corpora are growing fast and steady and they have wide applications.

The applications of aligned parallel texts are extremely diverse, and include compiling translation memories, deriving dictionaries and bilingual terminology lists, extracting knowledge for cross-language information retrieval, retrieving examples for computer-assisted teaching or contrastive linguistics, etc. [13]

In translation studies, apart from those mentioned above, the current applications also include creating terminology bases, user dictionaries, evaluation of MT qualities etc. The main advantage of using corpora is they provide protean examples of natural usages of languages and with the advent of electronic corpora they can be easily adopted to do various studies. The pitfall of corpora is it takes a long time to develop and for parallel corpora it is costly to align.

Translators are a factor much neglected but crucially important in the discussion of modern translation process. [14] There are enough misunderstandings of translators:

As we know to our cost, non-translators have little understanding of the translation process, even laboring under the delusion that translation is a mater of looking up words s in a dictionary and writing them down in the right order. [15]

The learned agents of cultural transmission who circulate information, knowledge and passions around the globe as they work around the globe as they work discreetly in the service of two masters. [16]

The role of translators has been defined by Edmond Cary as instruments of humanism, peace and progress. (Edmond Cary 1956: 181) However, as the translator is becoming more and more dependent on information technology [17] he/she seems to be less and less taken notice of. There are even scholars claiming that:

Since all translators (or nearly all) now use word processors and PCs they are considered on a par with the office typist. [18]

Thus, it is important and urgent to point out when discussing translation technologies that the basic skills of being a translator are still relevant even though these tools have largely levitated the work load on the translator's part.

It's only natural that after so many years' development, these elements like small trenches developed after the rain, now start to converge into a big river that will possess more currents and power than any of the streams alone.

MT and TM have a homogeneous origin, as when MT development encountered its ebb many people started shift their resources to develop TM tools instead. During recent years people begin to realize the synergy these two tools can achieve by combining them together:

Integration is already evident: many commercial MT systems now incorporate translation memories, and many TM systems are being augmented by MT methods. [19]

An example is SDLX translation tools. There are integrated MT functions as instructed on its user's manual: If you use Machine Translation (MT) software, and if the MT software provides the appropriate SDL interface DLL, you can now use that software from within SDL Edit and SDL Apply. Another example is @promt. It has demonstrated a Built-in technology of Associated Memory (a kind of translation memory database integrated with machine translation system) [20]

The concept of translation tools integration has well manifested itself in EURAMIS project set up in the European Union as well:

The system is still under development, but already operative in large part, at the European Union's Translation Service (Theologitis 1997, Blatt 1998). The core of the system is EURAMIS, which combines: (a) access to full-text databases, particularly the CELEX multilingual database of EU official publications; (b) access to centralized translation memories, topical TMs, and personal TMs for each individual translator; (c) facilities for creating TMs; (d) access to the Commission's MT systems; (e) automatic extraction of terminology from the EURODICAUTOM database, conversion of terms into MultiTerm format, and access to the MT dictionaries. [21] ...Those who prefer working with a translation memory can still get suggestions from machine translation where translation memory could not provide satisfactory results; and those who prefer machine translation can use translation memories as an added value, e.g. in order to reuse binding translations or to be consistent with previous ones. [22]

Machine translation has also been used to create fuzzy matches in TM tools. [23]

Corpora study has yet to be defined as either theories or methodology in language studies but corpora and especially parallel corpora can provide translators with an invaluable tool: The most helpful is probably the chance to test one's own tentative translation against the background of a large selection of original text written in the target language. [24] Considering the fact that there are already a considerable amount of aligned files available in parallel corpora, they can be used in ways such as merging into the TM files in TM tools to enlarge the translation memory in certain domain.

The following is a simple illustration of the current integration of these tools. More applications are yet to be investigated.

There is so much being said in the studies of human and translation technologies. The main school idea seems to be that human translators must adapt to the new technologies, and it seems that new computerized tools would not need to consider translators' practical use and experience. [25]

John Hutchins has outlined three categories of translation demand: that of publishable quality: translation for dissemination; that of short-lived documents: translation for assimilation; that of on-the-spot translation: translation for interpersonal communication. He argues that translation for dissemination should be done largely by human translators; that for assimilation can be done by MT and for face-to-face communication is dominated by human interpreters. Thus, it is reasonable to believe at this stage human translators are still needed in areas where the demands are beyond MT. After all, as translators and humans we have more moral, political and historical wisdom, than we know how to reduce into practice; we have more scientific and economical knowledge than can be accommodated to the just distribution of the produce which it multiplies.--- Defense of Poetry, Percy Shelley

It is clear that translators are not yet to be replaced by MT. However, translators also need to evolve in order to be more flexible. It is also worth recognizing the fact that the modern commercial translators have to make use of available tools that can help them improve the quality and productivity of translation.

In this sense, the translation industry is pointing to a new set of skills translators have to deploy if they are to act in a context of larger translation volumes, faster delivery times, stricter customization demands, and global production teams. The translator needs to develop the expertise of a project manager, a computer scientist, a documentalist, a DTP specialist, a terminologist, a language engineer... [26]

The TransRouter project in Europe has already started to examine this problem.

TransRouter project aims at developing a prototype decision support tool for users of translation technology, which would help them to decide whether a document or a number of files should be dealt with by human translation, human assisted translation using translation memories or phrasal lexicons or machine translation. [27]

TransRouter project is a good start to explore these issues of how to integrate humans in the process. However, there seems to be a lack in systematic studies of the latest development in this field as to how exactly translators should play his/her role according to different tasks in the translation process and with the emergence of technologies translators' role seem to be less romantic even less important than before as he must admit translation is no longer, under most circumstances, done by himself but a result with the cooperation of the software engineers, programmers and other linguists. The coexistence of translator and translation technology seems to be contradictory prima facie but in fact they are better to be compared to two sides of one coin, the foundation of each other. Translator's role may change from a manipulator of words to possibly a supervisor and maintainer of the machines and other tools but the human factor can not afford to be neglected. Above all, there are still translations that cannot be done by MT or solved by TM. The human factor must be integrated into the process although vary in degrees. Through proper integration automation and MT will not be a threat to the livelihood of the translator, but will be the source of even greater business and will be the means of achieving considerably improved working conditions. [28] The integration of different translation tools as well as the creativity of translators seems to be a natural and inevitable trend.

The integration process among translation tools themselves and translators and technologies has already begun. Apart from projects such as EURAMIS, there are also many other similar projects. The Translator's Workbench project developed in Department of Mathematics and Department of Linguistic and International Studies at University of Surrey, UK, proposed a machine-assisted term elicitation (MATE) methodology for LSP (language for special purposes). KonText software is adopted to "exclude frequently occurring 'noise' words from a text " and extract terms needed. In fact, these terms can be easily adopted into TM tools. Integration of parallel corpora has been put on the agenda as well: Today's main research trends are spread across a continuum, with machine-assisted human translation at one end and human-assisted machine translation at the other. All along this continuum, though, parallel corpora can be a valuable tool and resources... One of the main uses of parallel corpora is in automatic multilingual lexicon construction and term banking (Gaussier & Langé, 1994). [29] The integration of translators has also caused concern and many people have realized the difficulty and necessity of incorporating translators and technologies:

...most translators are traditionalists by nature, preservers of balances, accuracies and niceties... It takes a long time, therefore-probably a full generation - before translators will fully accept the new tools. [30] But it is a good start that we begin to address this problem and through means such as education the change of translators' mindset will eventually take place. And with the help of technologies, translation will itself receive a much higher profile than in the past. [31]

Although, there are already many examples for the integration phenomenon, the process on the whole has just begun. Theories on this subject matter seem to be sporadic. Systematic studies on this imminent trend seem to be lacking. Yet, the integration of different translation tools and humans seems to be imminent and natural. We need to take a humane point of view on translation technologies and a technical perspective on the process of translators' works. Like fine tuning the sound equalizer on the control panel of a sound system, done well, the bass will provide the firm foundation, treble fascinating special effects, mediant, the perfect vocal performance, together a good translation of music through the speakers, through proper interaction, translation tools and humans can achieve synergic effects; "A computer is a device that can be used to magnify human productivity. Properly used, it does not dehumanize by imposing its own Orwellian stamp on the products of the human spirit and the dignity of human labour but, by taking over what is mechanical and routine, it frees human beings for what is essentially human." [32]

[3] Bar-Hillel, Yehoshua. Language and Information: Selected Essays on their Theory and Application, Addison-Wesley Publishing Company, Inc., 1964.

[4] The Emerging Role of Translation Experts in the Coming MT Era, by Zhuang Xinglai, Translation Journal, Volume 6, No. 4 October 2002

[6] Melby, Alan K,Terry C. Warner. The possibility of Language: A Discussion of the nature of language, with implications for human and machine translation. Amsterdam and Philadelphia: John Benjamins, 1995

[8] Lynn E. Webb, Advantages and Disadvantages of Translation Memory: A Cost/Benefit Analysis(http://www.webbsnet.com/translation/thesis.html, accessed 10 Feb, 2006)

[11] Isabelle, P.(1992b).Bitexual Aids for Translators. Screening Words: User Interfaces for Text, Proceedings of the Eight Annual Conference of the UW Centre for the New OED and Text Research (Waterloo, October 18-20,1992),76-89

[13] Parallel Text Processing, Alignment and Use of Translation Corpora, Edited by Jean Véronis,Kluwere Academic Publishers, Dordrecht/Boston/London,2000,pp2

[14] Translators Through History, edited and directed by Jean Delisle and Judith Woodsworth, John Benjamins Publishing Company, UNESCO Publishing, 1995, preface

[15] The translator as information user, Pamela Mayorcas-Cohen, Commission of the European Communities, Luxemburg. Translating and the computer, edited by Catriona Picken, Aslib, 1986

[16] Translators Through History, edited and directed by Jean Delisle and Judith Woodsworth, John Benjamins Publishing Company, UNESCO Publishing, 1995

[18] Translating and the Computer 10, The Translation Environment Ten Years On, Edited by Pamela Mayorcas, 1990, Aslib and contributors

[19] Machine Translation Summit VII, 13th-17th , September 1999, Kent Ridge Labs, Singapore. Proceedings of MT Summit VII "MT in the Great Translation Era" (Tokyo: AAMT), pp.30-44

[21] Computer-based translation tools, terminology and documentation in the organizational workflow: a report from recent EAMT workshops. Proceedings of the International Conference on Professional Communication and Knowledge Transfer, Vienna, 24-26 August 1998, vol.ll: 4th Infoterm Symposium: Terminology work and knowledge transfer? Best practice in terminology management and terminography. (Vienna: TermNet, 1998), pp. 255-268.)

[26] Translation and Project Management by Celia Rico Pérez, Ph.D. Translation Journal Volume 6, No. 4 October 2002 (http://accurapid.com/journal/22project.htm accessed 10 Feb, 2006)

[27] Swiss Participation in European Research Programmes (http://www.sbf.admin.ch/htm/services/publikationen/international/frp/eu-abstracts/html/fp/fp4/t98.0122.html, accessed 9 Feb, 2006)

[28] John Hutchins, ITI conference 11: international conference, exhibition & AGM. Proceedings compiled by Catherine Greensmith & Marilyn Vandamme.

[29] From the Rosetta stone to the information society, A Survey of parallel text processing, Jean Véronis, Parallel Text Processing, Alignment and Use of Translation Corpora, Edited by Jean Véronis, Kluwere Academic Publishers, Dordrecht/Boston/London,2000

[30] Ten Years of Machine Translation Design and Application: from FAHQT to Realism, Juan Sager, Translating and The Computer 10, Edited by Pamela Mayorcas, Aslib, 1990

[31] Machine translation and human translation: in competition or in complementation?, John Hutchins, International Journal of Translation, vol.13, no.1-2, Jan-Dec 2001, pp. 5-20.