Part 2

Broadly speaking, the translation services utilizing the Internet (including the World Wide Web) already available or at the planning stage appear to fall into two categories. The first category, which encompasses a relatively large number of services already in place, essentially involves modifications or extensions to conventional types of language services which take advantage of some - generally technical - aspect or aspects of the Internet. The second category, with no more than a handful of services evident at present, seeks to offer something new, a service which has either previously not been offered at all, or has been available only to a closed user group.

Many market observers will be aware that the number of 'online' translation services seems to be increasing at a rapid rate. Rather than simply provide a list of the services known at the time of writing, I would like to take a more detailed look at a very small, but in my opinion relatively representative sample of each of the two categories described above.

Addressing a global market

One of the most high-profile translation services to have been launched recently is 'Translate Direct' from Globalink. Announced in 1995, the publicity made available (for instance in the CompuServe FLEFO forum) claimed, among other things, that 'Translate Direct™ is completely different from any current 'online' translation service. It will greatly enhance the way in which much of the world's current and future translation work (for documents in specific electronic formats) will be done, and is intended to increase the size of the market by making it very easy, convenient, and affordable to obtain high quality translations.'

These are, indeed, most impressive claims, but what exactly is Globalink offering here, and how does the publicity advertising match up to the service actually being provided? Basically, 'Translate Direct' is offering customers - initially in the US, and subsequently worldwide - a choice of 'affordable translations of electronic documents', running from low-quality machine translation using Globalink's own low-end MT systems to 'professional quality translation by human translators', with sliding price scales to match. This is clearly an attempt by Globalink to leverage the company's solid market position in the low-end, PC-based MT product segment to achieve a greater degree of vertical integration in the language tools and services. As such, it represents the reverse strategy to that operated by a growing number of conventional translation agencies which are starting to offer clients MT facilities. At the same time, the Globalink service also leverages the opportunities for near-real time, one-stop shopping which the Internet offers.

As far as the human-oriented side of 'Translate Direct' is concerned, this great new 'enhancement' of translation work appears to be nothing grander than a virtual translation agency. Customers send jobs to Globalink, which then farms them out to registered translators and proofreaders, who send back the finished work by e-mail to Globalink for return to the customer. On a smaller scale, but one which is undoubtedly easier to control, similar processes have been in place at agencies and in-house translation services for some time now.

One area in which the Globalink system appears to differ from that of many other language service vendors is that the company clearly states that parts of the process are automated, for instance costing and billing, as well as the management of turnaround time. However, the problems involved here are potentially greater than Globalink may have estimated. To start with, the system appears to base prices (per word) solely on source text length. This could make price comparability difficult, and the company may have difficulty persuading customers that its prices are competitive, particularly when the service is offered on a worldwide basis to customers who may be more familiar with receiving quotations on, for instance, a per line basis. However, this on its own is unlikely to present an insuperable problem.

More fundamentally, the concept of a piece of software which uses a word count and knowledge of the type of service required by the customer to tell the translator when a job has to be completed - without the source text being reviewed by the translator or another competent person - is unlikely to be a successful approach for all but the lower end of the translation market. After all, it does not appear that Globalink is using AI-based tools to analyze source texts for specific parameters - statistical analysis of terminology, terminological homogeneity, neologisms, register, etc. - that would allow a rough estimate of the degree of difficulty and other factors determining the time required for a translation.

In this context, one factor which could prove difficult to control in the sort of online service being offered by Translate Direct™ is quality. With no internationally recognized standards yet available for translation process quality control, and a wealth of so-called accreditation and certification procedures all claiming to rate the output (content) quality of the individual translator, a service provider can only hope to consistently meet defined quality standards by instituting rigorous, transparent in- house quality procedures.

However, it may well be that Globalink too has recognized that in today's world (and even more so in the future) many customers view factors such as time-to- market (the right document at the right place at the right time) as even more critical than some notion of "perfect quality", which will always be subjective and open to dispute. Where many freelances still cling to ideas about delivering "perfect translations", regardless of time and cost considerations, the new breed of service providers is more than willing to give customers what they really need: acceptable quality at a competitive price, and delivered last week. In this respect, it would undoubtedly be of great interest to all translation and localization service providers if Globalink could publish the procedures and criteria employed to select 'Globalink Preferred™ Translators and Quality Assurance Vendors'.

To sum up, the sort of service being offered by Globalink is a natural extension to the traditional agency concept. While its ease of use may lead to a limited overall market expansion (above all at the low end of the market) and to a greater market share for Globalink, it is unlikely to affect the higher end which depends to a much greater extent on customer relations. In addition, it should be borne in mind that Globalink is not the only organization taking this approach.

In Part 1 of this article, I mentioned LINGO. In its original form, LINGO was a consortium headed by Siemens-Nixdorf and Deutsche Telekom. It aims to provide a wide spectrum of language-related services ranging from translation and terminology to distance learning courses. In the pilot phase, services and contents were sourced from a number of companies located in Germany. LINGO uses ISDN to provide high-bandwidth communications (particularly important for multimedia distance learning, but less critical for translation services) and X.400 as the messaging standard, with gateways to the Internet and other protocols. System rollout was scheduled for Q2 1996, but this has apparently been postponed (for an undefined period). Although technically sophisticated and relatively easy to use, it may well be that its complexity and high entry threshold (in terms of the facilities which customers and service providers would have to install) have tended to deter the attractiveness, and thus the potential growth, of the network.

Another venture which seems to have met a similar fate is TELELANG, a scheme originally co-funded by the European Commission. Although the initial pilot project produced a blueprint for commercial implementation, this distributed language services network has not got off the ground. This may be due in part, one suspects, to the decision by the consortium (which itself has seen some realignment recently) to await the outcome of the Commission's decision on whether funding would be made available for the implementation phase, instead of actually building the network and hoping that some public-sector funds will provided at some point in the future. This would appear to point at a lack of commitment and/or belief in the viability of the project on the part of the consortium - an old adage states that you should only apply for EU funding for something you are going to do anyway.

LINGO, TELELANG and other projects along a similar line (such as OTELLO, which will be presented at the LISA Forum in County Clare) also suffer from the same generic drawbacks - in particular in terms of quality control - as Globalink's 7Translate Direct—. Although some of them claim to be novel services, they are often no more than translation agency-type schemes broadly similar to the services provided by established agencies and translation companies, such as Logos in Italy, which have also installed Internet gateways as part of the process of the "Netification" of the high end of the language services market.

CompuServe: a network operator moving into contents provision

Whereas the status of the LINGO project involving Deutsche Telekom appears to be unclear at the moment, the CompuServe Document Translation Service (CDTS) has been operating on a commercial basis for almost one year now. It was presented by Dr. Mary Flanagan at last year's MT Summit in Luxembourg, and was recently profiled in Language International. Mary Flanagan kindly provided an update on the status of the CDTS for this article. Despite being restricted to subscribers to CompuServe's online service, the CDTS (Go CDTS for CompuServe users) appears to have enjoyed remarkable success during its first year of operation. The basic CDTS service is raw MT using a modified Intergraph engine, with post-edited translations (provided by an established translation company) available as an additional, higher-priced option. The ratio between raw and post-edited jobs appears to have shifted slightly in favour of the latter, and now stands at approximately 85% raw/15% post-edited. However, with post-edited jobs generally much larger than raw MT, the word count ratio is closer to 60% raw/40% post-edited.

Data protection aspects and confidentiality agreements mean that the CDTS does not analyze job contents or subject areas. However, they suspect that based on job sizes, the system is being used mostly to handle business documents and the occasional manual. Neither has CompuServe conducted any quantitative market research on the type of customer using the service, but user communications lead them to believe that most CDTS users are taking advantage of the service for business purposes.

The CompuServe model is thus similar in many ways to the Globalink service, the differences lying mainly in the fact that the service is being offered by a service provider in cooperation with a limited number of contents providers, and in the style of the initial presentation. CompuServe itself is not attempting to assume the position of middleman between customer and supplier in the same way as Globalink, nor does it pretend to be revolutionizing the market. It is offering a service which it realistically describes as producing "for your information" translations, as opposed to camera-ready copy. According to Mary Flanagan, 'the chief things MT buys you are speed and consistency of terminology—. With the CDTS considering support for end-user custom dictionaries, the attractiveness of the CDTS to business customers with speed considerations or large volumes of standardized or standardizable, often repetitive text is likely to grow still further. The possibility of a serious localization customer trying out a service such as the CDTS for a major documentation translation job looms ever closer. This may go some way to explaining why CDTS has met with initial success despite the monotonously venomous reaction of the vast majority of (individual) translators in the CompuServe FLEFO forum. Far from accepting the fact that an increasing number of customers are quite happy to accept quality deficits - and even the occasional garbage - in translations they need urgently so as to gain a rough idea of what is being said in a foreign language, this is yet another instance of the blinkered attitude of so many human translators, remote as they are from what the market really wants. One of the most frequent criticisms of MT voiced by translators is that it turns them into nothing more than (post-) editors, it 'destroys the creative, artistic nature' of translation. I think that the point will come where translators have to decide between changing the way they work and make a living, or moving into a different line of business. Dogmatic, purist beliefs will not, I'm afraid, stop technological progress.

The broker model - A new breed of language-related service?

Typical of the small number of Internet-based services which are not easily classifiable as extensions of the agency concept is the 'LinguaServe' database (available in Germany at http://www.bodan.net.linguaserve, and in Switzerland at http://spectraweb.ch/peconweb). What makes LinguaServe so different is that it is not providing language services such as human or machine translation directly, but acts merely as a broker. One of the first services to start realizing the full potential of the Internet and World Wide Web as a mature business medium, the LinguaServe database modules contain information on a wide range of language and subject-area services and topics, such as

• Language services: registered translators, interpreters, terminologists and translation companies
• Dictionaries, lexica, monographs and other specialist literature, glossaries, catalogues and other publications
• Tools: MT, CAT and TMS programs, communication and conversion hardware and software
• Documentation: technical writers and editors, graphic artists and documentation companies
• Legal: national and international laws, regulations, standards, guidelines, European Directives and so on
• Periodicals and learned journals
• Agency business (offline only)

As far as service providers are concerned, a relatively modest annual fee secures them a record in the database containing quite detailed information about their offering. Unlike directories issued by professional associations, for instance, the information can be updated regularly. Publishers can also store detailed information about dictionaries, glossaries, etc. in the relevant module.

Users too pay a subscription fee to access the information contained in the database modules. Unlike the Globalink or LINGO model, LinguaServe plays no part in transactions between service providers and users, who then establish business relations independently. Its success therefore stands and falls with the degree of maturity and expectations of its users (a familiar problem to translation profetral information broker, it cannot be held liable either for the information stored in its database or for the quality of the services which it is brokering, the uncertain nature of the relevant legislation, which often establishes a chain from the end-user back to the original manufacturer or service provider, could potentially pose a number of problems for a broker such as LinguaServe.

And next?

Part 1 of this article highlighted some of the trends affecting the language services industry as a consequence of the expansion of the Internet and related systems, for example online access as an indispensable requirement for market competitiveness, the shift away from 'art-form— translation dominated by the skills (or lack of them) and idiosyncrasies of individual translators and towards a defined business process, and the requirement for deeper subject area knowledge, and the constant updating of such knowledge, on the part of the individual service providers.

Part 2 has outlined some of the widely differing approaches adopted by companies wishing to exploit what they perceive to be the business advantages of the Internet. Their long-term success will depend to a large extent on whether their assessment of these advantages proves to be accurate and sustainable, and on whether they are able to sell these benefits, not only to their customers, but also to the other service providers on whom the success of their venture ultimately depends.

This increasing interdependency between language service providers could well prove to be one of the most striking effects of the Internet. Translation service providers at the lower end of the market, already facing the great upheavals described in Part 1, could well find themselves ever more at the mercy of full- service vendors able to dictate prices and other terms to a much greater extent that today. At the same time, however, translators with more tightly focused specialist skills can take advantage of brokerage services to increase their own market visibility, win new customers and gain a greater degree of independence. Their success in doing so will depend to a large extent on their level of confidence in their own abilities.

---

Robin Bonthrone
Fry & Bonthrone Partnerschaft
Language Consultancy and Services
Rochusplatz 10
D-55252 Mainz-Kastel
Tel +49 6101 22504
Fax +49 6101 22860
E-mail 100277.3467@compuserve.com