Barriers keeping post-edited machine translation from being more widespread

Barriers keeping post-edited machine translation from being more widespread

It’s a sad reality. We live in a world where instant gratification is a demand, not a luxury. When a job needs finishing, you, your project, and your clients cannot afford to wait.

In spite of the high demands for projects to get finished, wait times still exist. With only 24 hours in a day, jobs that rely on human intellect and skill are often slowed when a translator reaches capacity.

As the demand for immediate results increases, so does the technology to help companies like yours. This is certainly the case in the translation industry today.

A look at machine translation

Machine translation was developed to speed the process of creating vital documents in multiple languages. Still, no matter how advanced of a service machines are able to provide, mistakes happen.

To alleviate the inevitable inaccuracies of machine translated text, companies such as BURG use “post-edited machine translation” (PEMT). This makes translating a document significantly faster while still upholding the expected levels of accuracy. It works by starting the translation project using a machine translation. Then, a human translator reworks the machine-translated document to make it sound natural.

Barriers to PEMT

On the surface, it seems that this style of translation is win-win. Companies like yours win because of the faster turnaround and quality results. Translation companies win because it speeds the process, keeping your project moving along at a faster clip.

Still, the adaptation of PEMT has not reached the expected levels. There are a few reasons why companies are not utilizing this approach.

A recent Common Sense Advisory (CSA) survey of PEMT buyers exposed some of the top reasons why this approach is not becoming more widespread. The top concerns were surprising: quality problems, technical complexity, and limited integrations.

Quality problems

For complex industries, such as the legal industry or pharmaceutical industry, quality is paramount. There is zero room for error. One mistake could cost people their health or livelihood.

The CSA survey showed that 65% of all respondents were concerned over quality problems with PEMT. This was by far the biggest concern felt by the respondents.

Upon closer look at the breakdown of responses, something interesting came to light. The survey exposed that the group far more concerned with quality were those who used machine translations in-house. A full 73% of this group said quality problems were a concern. However, the companies that outsourced entire projects to language service providers were less concerned with quality problems. Only 52% of this group was concerned by quality.

The primary reason for this large disparity is the quality checks and balances process used by language services providers. In-house teams have far fewer resources to ensure that the translated text is up to snuff.

Technical complexity

Another interesting response was that technology acted as a barrier to PEMT. Overall, 31% of respondents cited technology complexity as a concern.

Once again, there was a clear disparity between the primary groups of respondents when broken down between in-house PEMT users and outsourced PEMT buyers. 36% of in-house machine translation users were concerned with the complex technology. Only 24% of respondents that outsourced translation services were concerned with technology.

This shows that the widespread adoption is halted more by people working directly with these systems rather than those who are isolated. The people accessing translation projects and updates through familiar communication platforms are more inclined to feel at ease with PEMT.

Limited integration with other systems

It is perhaps not surprising, but the third largest barrier to adopting PEMT is the lack of integration between platforms: 26% of respondents cited this as a concern. This breaks down to 30% of in-house machine translation users and 24% of companies that outsource being concerned about how well each system is able to link together to create a high-quality translation.

This again plays to the notion that people prefer comfort and familiarity, as well as convenience. Shifting between unfamiliar platforms is not ideal in any work environment, let alone one that is so sensitive.

As developers create more systems that streamline the process, this concern will likely fade.

These barriers are ones that will be overcome as more experience and development takes place with machine translations. Until then, many people seem more comfortable and at ease working with a familiar and trusted language service provider.

Image: PhotoSpin