New Trends in MT

communicate-Globally---in-any-language

Machine Translation (MT) has experienced an impressive growth in the rate of adoption over the last two years and today is being used by a growing number of enterprises and Language Service Providers (LSPs). Increasing numbers of MT vendors are now coming to market offering solutions which open MT up to a much wider audience and with a broader range of usage possibilities.

The motivation to use Machine Translation arises from many factors:-
  • Improvements in MT quality and the introduction of SaaS based platforms.
  • Exponential growth in digital content production and a growing customer expectation for information in their own language.
  • An increase in online social community activity and a desire to interact with international users.
  • Increasing competition within global markets and a motivation to provide more value to customer and stakeholders.

 

New Trends in Machine Translation

 

Google and Microsoft are probably the most well know MT providers for the general public, however, when we talk about enterprise users or the Language Industry, different suppliers of MT services and technology take precedence.

Traditionally, enterprise Machine Translation was supplied through high-cost consultancy organisations – with lengthy and complex deployment models and high project costs. Add to this, maintenance and pay-per-word annualised contracts and these costs mount up fast! In this scenario, small to medium-sized enterprises weren’t able to afford MT and so the majority couldn’t embrace MT within their organisations.

The Dawn of a New Era!

So how has Machine Translation changed in the last few years and what direction is the technology moving towards?

Well, as companies become more aware of the cost reductions that MT can deliver, they are beginning to search for sustainable MT solutions which will allow them to control the MT development process, and the integration and deployment process, at a reasonable price. These progressive organisations are searching for a solution that will protect their data and their client’s data, will provide maximum Return on Investment (ROI) and will leverage their existing translation assets and add velocity to their localisation workflows.

Understanding these challenges was the central motivation behind the development of KantanMT.com; a cloud-based platform used to customise, improve and deploy Machine Translation within translation workflows and software applications. KantanMT’s Founder and Chief Architect, Tony O’Dowd, a well-known localization industry entrepreneur became more aware of the language challenges facing international companies and LSPs around 2010 and started to develop an industry solution that would make MT available to a much wider audience.

Automating processes that consultancy companies were doing by hand would maximise efficiencies during the MT development process, reducing overall costs. It would also help to tackle complexity; a scary thought that plagues most who consider MT. Tony added an intuitive, easy to navigate user interface and extremely powerful analytics to make it easier for users to visualise what was happening to their data and what affect it was having on their MT engines during the building and customizing process.

Later, after a platform was developed that could not only match but also outperform traditional systems, KantanMT decided to address the challenge of translation quality evaluation.

how-to-measure-machine-translation-quality

For years translation companies and organisations battled with post-editors about Machine Translation output evaluation. The basic metrics that tell you how your engine is likely to perform, do not always offer a good indication about how much post-editing will be needed. Together with a team from The CNGL Centre for Global Intelligent Content, KantanMT developed an algorithm which assigns quality scores for each segment (or sentence) translated by a KantanMT engine. This means that Project Managers can accurately scope a project and determine how much of the output needs to be post-edited (if any).

KantanMT’s approach to MT has shaken up the language industry and left many companies questioning why they paid so much for MT in the past. Traditional MT providers are now quickly following in KantanMT’s footsteps and developing online portals for Machine Translation that try to mimic KantanMT.com. In the past six months alone, five vendors have released SaaS based SMT solutions. However, many still are pushing their consultancy services and saying that the same quality will never be achieved using online portals. KantanMT’s customers say different, indeed all of these have moved from traditional services to the web based approach. I guess time will tell which approach stay ahead…