E.g., 11/20/2019
E.g., 11/20/2019

Translation Quality Models and Tools - Is There Room for Improvement?

18 July 2013

As a follow-up to our last post on Mapping Today's Translation Evaluation Practices and Needs, we wanted to tell you more about translation quality models and tools currently being used, and ask if it's time for improvements? Almost 500 translation and localization buyers and vendors voiced their opinions to us via an online survey conducted in May 2013 by GALA in partnership with QTLaunchPad.

Which Models are Being Used?
While some overlap exists, there is a clear preference for internal models, some of which are modified models that originally came from external tools and standards. Tool-specific models, i.e. those already included in CAT tools and existing workflows, are also popular. External standards such as EN15038 and ISO900 are prevalent too, while a large portion of participants had no formalised model at all. Figure 1: QA Models Being Used
Figure 1: QA Models Being Used

What about QA Tools? Given the close link between models and tools, it's perhaps unsurprising that most respondents use the built-in QA tool in their existing CAT tool(s), while many have implemented their own in-house tools. As with models, more than one can be used, a spread of external tools shows usage of ApSIC XBench as a lead. Yet, once again, many do not use, or do not need to use, any QA tool at all – see Figure 2.

Figure 2: QA Tools Being Used

Figure 2: QA Tools Being Used

Is There Room for Improvement? The survey participants were asked whether they saw the need to change their quality assessment processes in the next two years. Out of the 313 individuals who answered this question, the majority (70%) reported a need for gradual improvements, e.g. by developing better quality evaluation software or adopting standard metrics. An additional 9.9% regarded the need to improve their quality evaluation processes as urgent, while 15% reported being satisfied with them, and seeing no need for change in this area in the near future. Only 5.1% of the respondents did not have an opinion on this issue. Specific comments focused especially on the need to improve quality assessment processes concerning in particular terminology and the quality resulting from MT post-editing; finally, a number of comments emphasised that the respondents are constantly pursuing enhanced quality assessment, underscoring the crucial importance of this aspect within the industry.

Figure 3: Is There a Need for Improvement in Your QA Processes?

Figure 3: Is There a Need for Improvement in Your QA Processes?

Conclusion This industry-driven survey has shown the crucial importance in not falling behind in quality assessment. The feedback provided by the almost 500 participants highlighted, in particular, the need for constant improvement in this key area, and a general willingness on the part of translation and localization buyers and vendors to enhance their quality assessment processes from what they currently use. One of the main aims of the QTLaunchPad project is to help overcome quality barriers in machine and human translation, especially by promoting widely adopted, flexible and effective Multidimensional Quality Metrics that can meet widely felt needs in the industry.

Interested in Learning More about QA Models and Tools?
QTLaunchPad is a European Commission funded collaborative research initiative dedicated to overcoming quality barriers in machine and human translation and language technologies. We invite all interested parties to learn more about QTLaunchPad's Multidimensional Quality Metrics: http://www.qt21.eu/launchpad/content/training

Get in touch with us at [email protected] on Twitter www.twitter.com/qtlaunchpad and LinkedIn http://www.linkedin.com/groups/QTLaunchPad-4807518

randomness