You’re probably already familiar with content management systems like Adobe Experience Manager and Salesforce Commerce Cloud. But what about the marketing technology that allows brands to effectively adapt content for international audiences?
Computer-assistant translation (CAT) tools are software applications that linguists use (whether working in-house
or for a language service provider) to translate content from one language to another.
CAT tools typically have a range of databases that automatically aid linguists with their translations, including translation memories (from specific source language and target language datasets), glossaries, term bases, advanced search functionalities, segmentation data, and frequency information.
Then there’s the Translation Management System (TMS). This tool supports file formats and the exchange of information between team members. The TMS is designed to support the efficiency of the translation process from project launch to delivery without users needing any additional technical skills.
These tools allow for the automation of manual tasks so teams can focus on activities that are harder to automate and the more creative aspects of adapting content to different markets.
We’ve also developed a propriety technology platform, STREAM, that has the best of both TMS and CAT tools. Linguists work in our secure and centralized ecosystem with access to translation memories, glossaries, term basis, and QA tools, while project managers can track and manage the translation workflow.
The evolution of intelligent language automation
From our research, it’s clear that automation certainly has enormous potential, especially when it comes to machine translation. Brands expect it to help them deliver increasingly accurate interpretations of content in the future. And companies like Facebook are already dipping their toes in the water of automated localization.
While platforms like Google Translate aren’t sufficiently robust to tackle complex translations for critical website content, we trained Neural Machine Translation (NMT) networks to analyze large volumes of translation memories to advise clients if machine translation is suitable for their projects.
In fact, through our research with eConsultancy, Ralph Aoun, Global Marketing Manager, at Facebook, discussed how the infamous social media platform uses machine learning.
“Facebook’s personalization engine utilizes machine learning fueled by understanding people’s interests, preferences, and behaviors. The personalization engine can then suggest automated content translation to marketers and deliver the right language to the right person at the right time.”
Typically, we’ve found success with MT for large volumes of content with lots of repetition, such as user-generated content, video transcripts, or even product descriptions. Of course, sensitive content such as financial statements or certain legal documents should always be handled by human translators.
NMT quality is decided by three key factors:
- Size of available training data with which to train the custom NMT models
- Language distance – how similar the languages are to each other
- Domain closeness – the level of match between the source translations used to train the model and the content you will translate using machine translation
There are also different forms of machine translation depending on the purpose of the content that’s being translated.
- Raw MT – internal use. Quick general understanding of content that’s not consumer-facing.
- Post-edited Machine Translation (PEMT) – Consumer-facing content where style and tone of
voice are not critical.
- Post-edited Machine Translation + Revision (PEMT+R) – large volumes of content meant for public consumption that need to be produced quickly at a lower cost than professional translation but with high levels of accuracy and fluency.
While machine translation will certainly not take the place of professional translators, especially for more complex translations, it will certainly be used as a tool to aid translators and marketing teams with
workflow efficiencies and reduce costs.
We began training Neural Machine Translation engines in-house and became one of the first UK providers to get ISO 18587 for machine translation post-editing.
Our propriety technology platform, STREAM, has been compatible with major eCommerce platforms since 2011, and we design bespoke processes based on your critical content need with automated functionalities to improve efficiencies, ROI, and time to market.
To find out more about STREAM and how you can implement automated processes into your translation workflow, click the link below: