Technology
Resources

Machine Translation

Over the last 30 years Machine Translation has made huge progress and transformed the localization industry – and Alpha has been there every step of the way.

While early Machine Translation was either rules-based (RBMT) or statistical (SMT), utilizing large corpuses of translated documents, developments in technology over the last decade have led to the application of deep learning and Neural Machine Translation (NMT). Put simply, this is a series of ever more powerful and complex algorithms combined in a neural network based on the way in which human brains function.

The most recent development is Deep Neural Machine Translation; it harnesses the power of AI and machine learning to process multiple neural network layers, rather than the one layer used in NMT. The result is fast, accurate translation from one language to another.

All of these different types of MT have their pros and cons, and at Alpha our engineers and linguists work together to combine them in the way that best suits the needs of our clients. For example, one of our biggest clients needed over 8 million words translated in just 4 days… and we did it! Thanks to Machine Translation, content connectors and well-maintained Translation Memory (TM), client glossary and bespoke algorithms, we managed to get this monumental workload translated to a standard the client was delighted with.

Machine Translation can have its drawbacks, but here at Alpha we’ve learned how to use it to our advantage, and we’re constantly improving our systems with more data, client insights, trials and tests. Depending on the project, we can offer a range of services and workflows: from raw MT (unedited/unproofread content) to a full linguistic translation with MT assistance. 

We have always taken a pragmatic approach to technology; our linguists are experienced at post-editing machine translated content, while our engineers are experts in customizing MT engines to suit our clients’ demanding requirements. And it doesn’t stop there. We can help with:

  • Customized engine creation with your own data and terminology
  • API access to MT engines with translation tools
  • Consultancy on which MT system is best suited to your domain, languages, and budget
  • A proprietary reverse analysis process that ensures fair pricing for post-editing projects
  • Quality evaluation of MT output following TAUS DQF or customized models

We respect and champion the work of our human linguists, first and foremost, but we also understand the importance of AI, and its role in both NMT and Deep NMT. In a world of ever-growing volumes of content, technology’s place within the industry is always supporting and working with human linguists.