How translation quality assessment is changing

The translation process is constantly evolving to adapt to the changing market, but no matter what the latest trend is, translation quality is always in the spotlight. The last month has seen the SDL Translation Productivity team delving into all things quality and I have been focusing on quality evaluation. Faced with a vast amount of information on the subject, the main thing that struck me is that quality assessment is evolving. So, I hope this blog can shed some light on what can be quite a complex topic.

Download ‘The Pursuit of Perfection in Translation’ eBook to find out more>>

Ticking boxes

How often have you heard translation vendors say they provide ‘high quality’ translations? And what about customers who claim, ‘I know good quality when I see it’. Quality is an ambiguous concept – one that has no set definition of ‘high quality’, nor one approach for assessing translation quality.

Translation quality assessment models provide a structured way to assess quality. There are existing industry models which are mostly based on error typology. Put simply, different types of error are given a score, these scores are totaled and if the total is over a pre-defined threshold, the translation is not good enough quality.

If you want to use an existing model, then you should choose the model best equipped for the type of content you are evaluating. The SAE J2450 model suits technical documentation for example, whilst the MQM model is a good all round framework which provides flexibility for customizing metrics. These models provide a great framework for industries where technical translation is required to ensure translations are of the highest quality.

A more flexible approach

In the digital era, the way that people consume content has changed, and with this so has the type and frequency of content production. Digital content that will only exist for a month may not need the same level of translation quality as a printed brochure, for example. Common Sense Advisory (CSA Research) states that “one-size fits all quality models are insufficient to meet the variety of needs companies face today. A flexible model will allow you to tune processes without having to retool them every time."

This is where the models that offer more than one way of assessing quality can deliver benefits. For example, the TAUS DQF model includes ways of assessing fluency, MT systems, post-editing tasks, adequacy and productivity for example. Attila Görög, Director of Business Development at TAUS explains “There are now ways to provide different levels of quality.

Before, 20 to 25 years ago when these technologies weren’t available yet, translators were given a source file and translated it to a very high quality. So there was basically one quality level, the human translation quality. Nowadays with new technology, new translation workflows and budgetary restrictions on the buyer’s side, translators can be required to deliver lower levels of quality. So I believe that that the difference in the level of quality that is now required is something which will determine the way we evaluate quality."

For a comprehensive look at the models, the analysts at CSA Research provide a great overview of assessment methods in their recent brief “How to Assess Translation Quality“.

Quality assessment in practice

As the type of metrics by which quality is measured evolve, so too does the way that these metrics are accessed and used by translation professionals. The launch of SDL Trados Studio 2015 saw the addition of the Translation Quality Assessment (TQA) tool which gives user a simple way to perform quality evaluation either by using existing industry models or by creating their own framework within the tool. Users can also go one step further by taking advantage of the open Studio 2015 API, to develop their own quality assessment plug-ins, which make the possibilities for designing bespoke quality evaluation endless. The flexibility of the open API enabled TAUS to develop the TAUS Quality Dashboard plugin for example, which enables Studio 2015 users to analyse their productivity and quality data – you can download it for free from our SDL OpenExchange App Store.

The future

It is clear that over the last few years, quality assessment processes have become more flexible to accommodate changing needs of content owners, but what’s next? I asked Attila what he thought about the future of translation quality evaluation, “I think transparency and benchmarking will become major trends. It’s not enough now to just shout “I am the best, I have the best quality". Everybody is shouting that. There should be a way to prove it. One way is to share your results and contribute to creating industry averages."

So whether the tick box approach suits your business or you are looking for something a bit more flexible, the technology exists to build a quality evaluation process that is specific to your needs.

For a detailed look at the subject of translation quality you can read our new eBook, ‘The Pursuit of Perfection in Translation’. Or if you are interested in finding out how this works in reality, why not register for our live webinar on ‘Measuring quality in Studio 2015 with TQA’. Finally, don’t forget to keep an eye on our ‘Quality Matters‘ webpage over the next few weeks, as we release a host of new resources on the topic, including infographics and a how-to video.