A machine translation (MT) system is a programme that uses natural language processing technology to automatically translate a text from one language to another. Google Translate is the most commonly used MT system.
The systems we build or help select for our clients are very different from Google’s general-purpose engines. Our MT systems are custom built for a specific domain, client or even product. The translations produced by customised MT systems will be markedly better than those of general-purpose engines for the content that they are built to translate.
The main commercial applications of MT are for real-time or near real-time interaction, understanding/getting the gist of texts, and as a productivity tool. There are significant differences in the extent of customisation offered by different solutions on the market. When you read about providers offering custom MT they are often talking about different things.
In most cases customisation refers simply to the ability to use your own data (language assets) and try different combinations of your data with other openly available data. These providers all use the same open source deep-learning frameworks and engineered them to work for commercial use, adding user interfaces and management reporting in the process. These are typically SaaS-based offerings.
A far smaller group of providers actually build custom engines tuning the systems using a combination of artificial intelligence together with sophisticated domain adaptation to help increase the precision of the output.
We help clients identify which solution is the best for them. When it makes sense, we build tailor-made customised MT for our clients leveraging eighteen plus years of continuous applied research. We help clients ensure their people, process and technology work harmoniously to reap the full benefits of translation automation.