Deus Ex Machina

Deus Ex Machina
Print Friendly

In Italy, for thirty years under the Borgias, they had warfare, terror, murder and bloodshed, but they produced Michelangelo, Leonardo da Vinci and the Renaissance. In Switzerland, they had brotherly love, they had five hundred years of democracy and peace – and what did that produce? The cuckoo clock.
Orson Wells, The Third Man

InnovationOnly two out of the top ten countries in the Global Innovation Index 2012 are Asian, both with a very strong western legacy, Singapore and Hong Kong. The first in ranking is usually considered almost still, Switzerland.

Recently, Raytheon BBN Technologies received an additional $5.9 million in U.S. funding for development of an automated foreign language document translation system, including refinement of a laptop-deployable translation system to automatically convert foreign language text images into English transcripts.

In a recent post on Google’s official blog, Franz Och, the chief researcher for Google Translate, claimed that “most of the translation on the planet is now done by Google Translate,” but he also admits that “for nuanced or mission-critical translations, nothing beats a human translator.”

Recently, Microsoft announced the availability of the Microsoft Translator Hub for commercial use to customized machine translation for specific industries. For the customization, users must feed Microsoft Translator Hub with translation memories, and Microsoft will “use, modify, adapt, reproduce, distribute, publish and display” its customers’ linguistic data, as well as other information supplied with it, to further strengthen the engine and possibly other online services, the refined capacity of which will attract more and more users, thus collecting revenues from ads and recovering from recent decline in investors’ confidence.

A very common assumption among machine translation opponents is that nothing can be translated into another language without understanding what it means in the original language, and that computers don’t understand anything by definition, as they just process instructions. In reality, they do not need to understand anything, even less languages. Statistical machine translation performs at its best by doing exactly what computers are made for, calculations, by picking the most probable translation from a reference pattern derived from a corpus of thousands bilingual sentence pairs.

Another supposedly common pitfall of machine translation is poetry. A recent example has been brought with the proem of the Aeneid. This has been the model of epic poems, and of medieval and Renaissance heroic poems, and many translations of the Aeneid exists on the Web, from John Dryden’s (1697) to Sarah Ruden’s (2008). This makes Google Translate quite brilliant in handling at least the opening lines of Vergil’s poem.

The improvement of machine translation is such as to make an anonymous would-be LSP ask Businessweek’s Karen E. Klein whether online translation services could make this business obsolete. While it is true that online translation services have not dented the market for translation services, it is not true that the translation industry has seen minimal impact from the global economic downturn. For a reliable analysis, the overall translation volume should be considered with respect to the overall turnover, on the basis of the average value.

In doing so, the cliff dive of translation pricing in 2008 and 2009 would look as reasonably due also, if not mostly, to recession and inflation, as well as to many other factors endogenous — and endemic — to the translation industry, such as the delay in renovating obsolete business models. These models are the same the anonymous Businessweek’s reader — and would-be LSP — was probably about to follow.

On the other hand, there’s no evidence whatsoever that machine translation has caused or induced a price drop; the culprits for this drop were unacquainted customers and incompetent LSPs with their reckless use of translation technology, only aimed at compressing prices and sustaining price competition.

For example, Provo’s MultiLing is growing despite Google’s involvement in patent market. MultiLing counts 90 employees in Provo and 120 more throughout the world, and hires immigration attorneys to deal with the finicky process of patent translation.

This case is in line with the trend among law firms of bringing in flocks of attorneys with a flair for languages to settle international business disputes, showing once again that language mastery is still seen as an ancillary skill. Nevertheless, this also supports the belief that “for legal matters, translators and attorneys with foreign-language skills are probably in the same price range.”

More notably, though, MultiLing’s intern/freelancer staff ratio is another important clue to a more effective approach to the translation business. Moreover, the trend from drops to drips requires staffing to deal with projects made up of uninterrupted flows of tiny chunks.

For example, a 28-word job is definitely not profitable, even for € 6,50, especially if a phone call (of even one minute only), a confirmation email (requiring at least five minutes to write), and a PO (requiring ten minutes more to release) are necessary: the revenue is gone nine minutes after starting.

What Google Translate has done — and Google Translator Toolkit could still do — is to favor the commoditization of translation technologies, especially of machine translation, rather than of translation tout court. The commoditization of translation is also due to large translation companies unilaterally imposing their terms and conditions to SLVs and freelancers for their inability to negotiate with powerful customers — with enormous expenditure capacity. Their relationships with these customers are vital as the only means allowing for their gigantic overhead, mostly made up of generous pays and benefits to management and sales staff.

Translators, not only in SE countries, are used to whine, but rarely focus on alternative strategies to improve their business. It’s no surprise that the “American” Chrysler’s and FIAT’s Sergio Marchionne accused Volkswagen of exploiting the crisis to gain market share by offering aggressive discounts: VW is the largest carmaker in Europe and the second in the world after Toyota and has been able to keep costs down by sharing parts and development among a huge stable of brands, from low end and commercial vehicles to luxury cars. The translation “industry” could possibly learn from the car — real — industry to innovate in order to perform better and make better products.

The common attitude in the translation industry, instead, is to seek a deus ex machina. One should be surprised then to hear that translation teachers claim to teach primarily to solve problems!

A common problem amongst translators is represented by contracts. When translators receive a contract, they often censure it for being clogged with clauses, possibly obscure. Is the fear of being suited by customers so big to require vendors to over-guarantee their supply?

Large translation companies and certain agencies are very often counting on translators’ ignorance of their rights, on the geographical distance and the prospect of being suited to dissuade translators from filing claims if the company breaches the contract. In reality, they could incur in the very same problems that could also be harder to solve somewhere, in Europe too. In addition, blatantly illegal terms in a contract or PO do not make it automatically null and void; only the clauses containing those terms are invalid and knowledgeable translators could as much blatantly ignore them.

Some LSPs do not even open their customer’s file before quoting and asking their vendors to quote. They act as they firmly believe that one size does fit all, while it does patently not. Some LSPs do not even run a reliable word count, and do not make any reference material available to translators even when they have plenty of it, know some exists or could even only ask their customer for it.

The rivers of ink flooding online and paper journals pretending to debunk translation myths and misconceptions are possibly addressing the wrong audience.

And while things have been changing over the centuries in translation, the job is essentially the same of 10, 20, or even 30 years ago. And those who come after us will do the same.

Over the last 30 years, vendor management has characterized the translation industry. A vendor manager should be responsible for identifying and qualifying suitably experienced translators and ensuring they are available when required, for evaluating projects and ensuring they are consistent with specifications and price policies and models by negotiating rates. Last but not least, a vendor manager should also be responsible for communications, service level agreements, etc.

Yet, leaving aside any consideration on the confusion of roles, for example between VM and PM, most LSPs are still struggling with absolutely inefficient models.

Any translator can virtually tell of at least an email from a prosperous LSP asking him to solve a problem with a (translation) software tool, a conversion problem, a cleaning problem, a compiling problem, etc.

This usually happens because the core business of LSP is no language service, but job intermediation. Therefore most LSPs are not adequately staffed to deal with the typical sort of problems of translation business.

It should come as no surprise, then, that the number of certified LSPs is increasing, especially as EN 15038. Quality standards do not require applicants to recruit staff with any specific skills. As a matter of fact, EN 15038 requires the TSP to have a documented procedure in place for selecting the people involved in translation projects, to ensure they have the requisite skills and qualifications, as well as the requisite equipment for the proper execution of the translation projects. What’s more, in delivering the translation service, the TSP must only provide for the technical aspects appropriate to the project requirements, making sure that the technology required is available, and carrying out any necessary technical tasks. In drafting the standard, nearly all the delegates to the working group BT/TF 138 claimed a section on value added services, but categorically refused to introduce any reference to metrics and measures, or set the relationship with contractors, even as to contract requirements only. The introduction of job ticket was severely hampered, and accepted only as an informative annex, i.e. not binding, as well as the technical processing.

Again, it should come as no surprise that more and more LSPs are increasingly recruiting freelancers only through portals, such as and TranslatorsCafé.com, although these led to a three-layered resource market, with a very wide low-end layer, a narrow middle layer and a very thin high profile layer, each one virtually impermeable to the adjacent one.

It seems obvious then that most LSP are interested in quality standards only as a marketing vehicle, as they do not know what these standards intrinsically mean and imply. The leading Italian company in this industry, founded in 1979, more than twenty years ago, repeatedly announced to be about to start the certification process to quality standards, but they never really did, and they are not certified yet. They were the first, however, to exploit the Web in its business model. They used then the profits— some said made also by not paying suppliers — to diversify, thus creating an escape pathway to sustain price competition and growing in volumes.

Overhead is less of a problem for the so-called “mom & pop agencies” and the small translation firms, that yet must give up large orders because they cannot afford vendor management and can only act as subcontractors. Therefore, price competition with the consequent compression of profit margins makes their business less and less sustainable. The lack of financial resources and the credit squeeze do the rest: these micro-businesses use freelancers as they were banks, paying them as late as possible, typically much after having been paid. Sometimes they do not pay them at all, possibly relying on the cumbersome and burdensome practices of credit collection. And if they do lose a business partner, they do not think too much about it: in accordance with Gresham’s law, they are not used to use the best resources, and they know they can always find a desperate or naive translator just around the corner, maybe through the training events they organize and/or sell to supplement their income.


AKA "il barbaro", Luigi Muzii is a founding member and associate at sQuid. He has been working in the language industry since 1982 as a translator, localizer, technical writer and consultant. He spent 12 years in several departments of a major Italian telecommunications company, and two years in a broadcasting service company. In 2002 he started his own consulting firm to act as an information design and delivery consultant. He was visiting professor of terminology and localization at the LUSPIO University in Rome for almost ten years, and is the author of a book on technical writing, and of many papers and articles. He was one of the founders of the Italian association for terminology (ASSITERM) and of Gruppo L10N, a group of localization professionals volunteering in educational programs. At present he is in charge of strategies at sQuid.

No comments yet.

Leave a Reply

− nine = 0

Issue 4 – Business Model Innovation

Issue 4 (January, 2013) focuses on “Business Model Innovation in the GILT Industry.” A business model defines the key elements of […]


Welcome to “The Big Wave”, a quarterly multilingual journal specialising in language technology. “The Big Wave” consists of five sections: Technology […]

The Big Wave Interviews Tilde

Andrejs Vasiļjevs, co-founder and chairman of the board of LetsMT maker Tilde, answers the questions prepared by The Big Wave’s review […]

A Three-Legged Table

By Luigi Muzii (This Article has been first published on the December, 2013 issue of MultiLingual) Over the last few […]

New Business Models for the Translation Industry

By: Diego Bartolomé The manufacturing industry produces goods. It tries to maximize the throughput while at the same time reducing […]

TBW interviews CrossLang

CrossLang NV, headquartered in Ghent, Belgium, is a privately-owned company comprising three divisions (CrossLang, CrossCheck and CrossCast) that offers consultancy, […]

ASD-STE100 – What’s new in Issue 6

By Orlando Chiarello This article summarizes and briefly analyses the new features of the ASD Simplified Technical English (STE) specification, […]

A Chat with Safaba

Safaba is an enterprise-level machine translation solution provider founded by Alon Lavie, a famous expert in the field of machine […]