Leonid Glazychev, CEO
Just as maintaining true physical fitness can’t be achieved by working a single muscle group, providing consistently high-quality, comprehensive, scalable translation services requires an approach that extends beyond any one area of expertise.
Processes include all areas of production in an innovative, flexible and reliable way. A serious translation company should have elaborate, documented processes that cover all major areas (translation, quality assurance, terminology creation and maintenance, etc.) These processes should have no glaring weaknesses and must be strictly followed in all cases. A modern company must also stay current, following industry trends and optimizing or creating new processes in response to emerging needs. The failure the evolve with an evolving industry leads to missed deadlines, quality lapses, costs going out of control, etc.
However polished our processes may be, they would be inconsequential without an apt and experienced in-house production team. Serious end-to-end production that creates added value in the localization business can only be built around a reasonably-sized, well-trained, communicative and agile in-house team comprised not only of project managers at different levels, but also solution architects, engineers, technical, tool, graphic and design experts, language, terminology and quality control specialists, vendor management, and more.It’s a team that cohesively creates and implements strategic solutions and processes. It’s also a team that is available for more than 8 hours a day, and can work in shifts or pass urgent projects and tasks seamlessly from office to office within a single day. Other approaches do not work as well:- A purely brokerage-type translation company with a tiny in-house team has to outsource almost everything and adds little or no value to the end product (unless it provides very specific, boutique-type services)- A less diverse, purely translation-centered team typically cannot efficiently deliver on more complex, less predictable or more challenging projects. A purely technocratic team can do great things in its specialty, but provides sub-par services when it comes to language supplier selection and quality.
This chain needs to be direct, robust, cost-effective and scalable. We must remember that the primary task of translating a text is performed by dozens, hundreds, or even thousands of native speakers all around the world. A high-quality “support network” of suppliers is integral to the production team’s success. Accommodating all project scales, large workload peaks and quick turnarounds requires a diverse, tested and trusted supplier network that combines in-country suppliers or teams and individual freelancers for each language. Everyone must go through training and evaluation protocols. We must ensure not only that they continue to produce high-quality results, but also that they are truly motivated to do so. People are the ultimate resource, and it is important to establish a relationship that encourages the best work. It isn’t possible to work on large projects using standalone freelancers only. While it may be easier on the budget, it typically wreaks havoc on consistency, quality and management for commercial projects. More subtle or specialized tasks like terminology management, however, are typically better executed by dedicated individuals. Direct supply is vital. As soon as a single part of a job is subcontracted downstream, you bid adieu to consistency, quality and reliability.
The quality model is separate from the processes because it describes not how, but what we need to do under the given circumstances to achieve consistent and uniform quality as it relates to the quality metric and, of course, client expectations. The same applies to the measurement of quality for existing translated materials. It goes without saying that simply having “two sets of eyes”(a translator and editor/proofer) on the text is not nearly sufficient to produce good quality. The same applies to blindly trusting certified translators. Being human means being prone to human error. A serious approach requires a model that allows us to define and measure quality, stack up against the client’s expectations, and take corrective measures where necessary. If you want to learn more about the quality model, please read further about its three components: quality methodology, quality metrics and quality process. The quality model starts with methodology — with establishing a concrete definition of quality and how it can be measured. To measure quality, we need to concentrate on factors that are most important for the market or user perception, but universal enough to account for different contexts, subject matters, etc. We also need to take into account a certain inherent level of subjectivity in language quality assessment. Long gone are the days of calculating quality ratings with a simple count of formal errors like typos, broken tags, incorrect terms, etc. This oversimplified approach requires a substantial amount of what might be called “busywork” (merely attempting to count all the imperfections), but often overlooks more important issues, like overall readability or adequacy of translation. This sort of approach can easily result in a technically and terminologically perfect but unreadable translation that overlooks or distorts the source content. We need to embrace a more sophisticated, thorough and universal approach to quantifying quality. Logrus IT’s original Quality Triangle methodology provides one such solution.
Having chosen the quality measurement methodology, we can begin constructing quality metrics. These metrics allow us to actually assign quality a numeric value (quantify it) and see how well the results align with client expectations. It is vital to create metrics that are flexible enough to utilize across a wide spectrum of real-life cases with different quality priorities and expectations. We can’t forget that quality is relative. What is rated substandard in one situation may be a lifesaver in another – it’s why we complain about the glaring translation errors on the Chinese take-out menu, but would be grateful for any semblance of English text while dining in a rural Chinese village. A modern translation environment requires a more diverse and flexible set of metrics. Some provide a comprehensive, albeit more costly, quality evaluation. Others are designed specifically for an agile translation process and are less exhaustive and more targeted, with emphasis on the fundamental factors, like overall readability and translation adequacy.
Last, but certainly not the least, is the quality process, which guarantees that all translated materials are evaluated using contextually suitable and relevant metrics that adhere to the client’s expectations.
A well-designed quality process needs to account for agile localization. There is often insufficient time to carry out a complete set of quality checks or to achieve final quality between updates. This calls for designing a process that accommodates quick, affordable, iterative, feedback-driven fixes and telemetry-based adjustments. The latter means that the frequency and depth of quality checks can be increased if serious problems are discovered, and decreased if the project is progressing more smoothly.
Technology is the fifth and final major pillar supporting the holistic approach to translation. It is crucial for project management, translation, quality measurement, processes, etc. 20 years ago, translation was just that — predominantly manual, with little or no automation, done in a simple word processor or spreadsheet. In the age of innovation, translation memories (TM), translation management systems (TMS) with online localization, tracking, job management, machine translation (MT) and other technologies are simply a fact of life.
While tools are easier to use and require less heavy engineering, but they don’t eradicate all problems and challenges.
– In reality, tool compatibility leaves much to be desired, and results in continued “tool lock-in”. In these situations, localization companies need to be nimble, open-minded, and well-versed in a range of tools to accommodate the client’s preferences.
– Tool integration is also far from perfect. Some tools are compatible, others are not. If a company has its own proprietary ERP system or has standardized on a certain project management or translation tool, it will inevitably face compatibility problems with many other tools and will need to improvise and/or invest in technology that will link everything together.
– Quality checks are integrated into most translation tools, but the level of sophistication varies dramatically. Passing internal tool quality checks with flying colors does not guarantee error-free material. Most built-in checks as well as third-party tools miss a lot of typical issues (we’ve proved it), and in order to release high-quality materials, the company needs to invest in more sophisticated quality tools that uncover a much greater volume of issues and do so better than built-in or popular third-party tools.
– Tool functionality and overall usability and quality also vary dramatically.
– The jury is still out on MT quality. It strongly depends on the language pair, subject area, corpus quality, MT engine type and training history, etc. Nothing is guaranteed there yet, and we need to innovate to integrate MT into the process efficiently and without sacrifices.
To summarize: a modern localization company needs to apply a client-centric, tool-neutral approach and provide modular, flexible, highly-customizable solutions embracing a variety of technologies and tools.
As for translation tool compatibility, we have less freedom of choice than we think. Take GSM phones and SIM cards, for example. Theoretically, any SIM card is supposed to work in any smartphone provided it physically fits. In reality, various providers around the globe use different frequencies, channels and protocols for data transfer, which most phones only partially support, limiting actual compatibility to making calls and sending SMS messages, not to mention coverage issues. It’s a gray area.
The same is more or less true for translation tools. Ideally, they should all support the latest version of the bilingual document format (XML Localization Interchange File Format, XLIFF 2.x) as well as common formats for exchanging TMs and glossaries (TBX), so that we could change tools without thinking about it. The reality is that each tool only supports a limited or customized subset of XLIFF, let alone variance in XLIFF versions. TM exchange is also complicated by the fact that fuzzy match definitions are tool-dependent and not completely compatible.