Towards a new assisted automated translation model
Throughout our last blog posts, we have been explaining the current state of the translation industry, the improvements on Artificial Intelligence applied to translation, and the key roll corporate proprietary and validated data (terminology bases) will have in order to make the leap to the next level of automated translation.
We have seen how the roll of professional translators goes beyond post-editing to verifying text accuracy and reliability. Their roll is quite relevant when it comes to validating and giving feedback to the results of automated translation systems.
Why is this so?
Because the true challenge to automatize translations with reliable results lies on what is known as domain specific translations. That is to say, specialty area translations with vocabulary and content scarcely used outside technical materials inherent of a given field. In such areas, machine translation engines don’t have enough knowledge base to generate accurate results.
Technological improvements on machine translations are based on working with assisted automated and custom learning systems (custom machine translation). This means teaching and training automated translation applications with multilingual corpus data validated by expert translators. Thanks to such curated input, machine decision-making models are developed on the basis of industry data combined with corrections and feedback from professional translators.
And as long as custom machine translation engines feed from validated terminology bases –not only vocabulary but also technical phrases, expressions and texts– and train their results with human post-editing, these systems will be more accurate and will allow to streamline processes and reduce costs.
The importance of data
All that said, once the technological frame is developed, we still find that this new domain specific translation systems require a large information volume to “learn”, as well as structured data so “the machine can understand them”, not to mention an experienced team to manage such technology. Therefore, it is difficult for a company to train its own domain specific translation system as translations are usually not their core business at all.
One example: The machine translation environment of Microsoft -Microsoft Custom Translator- requires a minimum of 10.000 phrases to be able to start training the engine. Nowadays this can be applied to companies owning a large validated terminology base, with many expressions, phrases and vocabulary.
In order for companies to profit from this new technology, they need to be aware that most of them aren’t considering their translations as corporate data, nor are they saving their information. This means right now they will not be able to use domain specific translations. Moreover, they aren’t generating data for future use.
Gear Translations technology is prepared for domain specific translations
We’ve established that a future asset is having a consistent multilingual content database by domain, (having a DB of terms and phrases in a number of languages for each sector, area or specialty sector).
At Gear Translations we are generating such assets for our clients. On the one hand, we identify and organize expressions, phrases and specific terms by technical area in a Terminology Base. We create translation memories (TM) that we call Translation Libraries.
On the other hand, we are structuring the information and creating the necessary registers so our clients benefit from improvements on custom automated translation in the future, as well as creating their own specific translation engine by domain.
As new technology breakthroughs occur in machine translation, Gear Translations is ready to train automated systems, integrating personalized and specific translation memories in platforms based on Artificial Intelligence.
At Gear Translations we help our clients accelerate and maximize their translation projects ROI, bringing together the best of human and machine translation. We continue to research and develop our artificial intelligence based platform.
Contact our team if you want to know more about turning your content into assets with our intelligent algorithm.