This is our star service. It is also the most popular requirement from language and localization departments: to have a proprietary SMT system hosted in-house which can be interfaced upon request, fully developed with the client’s own data, style, terminology and linguistic assets and enhanced with PangeaMT’s experience. Our Development Department integrates MT solutions in your translation process with increased efficiency and cost reduction in mind.

With our customized machine translation developments, we can create translation engines that pre-translate your content using your own vocabulary and expressions – optimized with your previous translations. Human post-editing services complete our value proposition to publish more, faster and in more languages and keeping to budget. If you need machine translation for social media reporting, web content generation or just need to get an idea of blogs and news in other languages, you will find that our custom machine translation services and API are extremely useful.

Our Development Department can:

  • Design and customize an MT system according to your specific needs (rule-based or Statistical Machine Translation implementations).
  • Integrate the system in your current workflow.
  • Generate your specific language model.
  • Generate domain-specific terminology databases or dictionaries
  • Provide consultancy, system deployment on-site or hosted at
  • Advice you on post-editing services.

Additionally we can provide

– Training courses.

Tips on customization
a) Statistical Machine Translation systems
The creation of domain-specific databases and dictionaries as training material will undoubtedly help you dramatically increase machine translation quality.

Consider as well training the engine not only your own company’s material, but also all materials that may be relevant for your industry. Point us in the right direction so we can generate as much linguistic material as possible.

We can build domain-specific dictionaries for your specific language domain, but SMT makes use of many sets of equations and probabilities: the more data you have available, the better. Typically, 2M words will suffice to obtain reasonably good results initially. Our job is to increase that figure so your development is based on 5M or even more, for much better results.

If you opt for an SMT system outputting TMX data, you will be purchasing a system that:

  • Provides a better flowing translation.
  • Easily integrates with existing systems (no need to change your current workflow).
  • Integrates our proprietary parser so in-lines and code are placed in the target segment, ready for post-editing (no other SMT system provides parsers for in-lines).
  • Accelerates the translation process without disturbing the formatting (as you are working with TMX data);

b) Rule-based systems
In traditional rule-based MT systems, the creation of domain-specific databases and dictionaries helps to dramatically increase machine translation quality.

We have the experience to build domain-specific dictionaries for your specific language domain.

If you opt for a trained, rule-based system, the implementation with domain-specific dictionaries can:

  • Accelerate the translation process and formatting (from approximately 20% to 60%, depending on the dictionary and user skills);
  • Provide both uniform terminology throughout the documents and uniform styles;
  • Help new employees to get used quickly to terminology and style.