Localization
Localization is the process through which a piece of content – text, image, video, audio – is translated and adapted to fit the target market’s look and feel. Localization is much more than just simple translation. The end purpose is that the localized piece of content feels as if it was created for the person reading the message in that exact place and at that exact time.
When managing large volumes of content that need to be localized, localization becomes problematic from an operational point of view. In most cases, the localization team is not in-house, which results in inefficient communication workflows that lead to content duplication, delays in returning localized content, or even losing content. Structured content can offer an excellent solution to streamlining content localization efforts. With a translation management system (TMS), especially when combined with a structured content management platform, internal and external content teams can work smoothly without ever losing or duplicating translation and localization efforts.
Example use cases
- Increase speed to market for products and support investor communications
- Multilingual access to field service documents
- Create personalized content globally across the entire content lifecycle of a product
- Improve customer experience with customized aftersales support
Translation memory
A translation memory (TM) is a database that stores segments, paragraphs, or sentences of content you have translated before, and it is usually implemented in a translation management system (TMS). As the name implies, its purpose is to enable the usage of previously translated content pieces into the current translation project or document. Using a translation memory enables translators to focus only on pieces of content they have not translated before while automating the translation of everything else. Each entry or segment stored in the TM includes the version in the original language (the source) and its translation (the target). A source and a target pair is called a translation unit or a ‘TU’.
A TM works by parsing the source content pushed for translation and checking whether any content there has been translated before. When coupled with structured content, a translation memory and a TMS can significantly decrease the time needed to translate content.
Example use cases
- Field service documentation
- Medical device regulatory submission assets
- Medical writing
- Consistent customer support across all language versions
Key benefits
- Shorter time to market
- Reduced costs
- Improved translation quality
- Consistent content
Machine translation (MT)
Machine translation (MT) refers to a content translation method that uses algorithms and machine learning models to translate natural language text. The algorithm breaks down text into words and phrases that are translated into the target language. There are three different types of machine translation: rules-based machine translation, statistical machine translation and neural machine translation.
It is important to mention that not all content types are suitable for MT. MT works best with structured content and unambiguous content, like a user manual or a clinical evaluation report (CER). Machine translation alone cannot accurately translate creative content like novels or marketing content – as it cannot capture the nuances of such content types. Another significant mention is that it’s best to have a human-in-the-loop approach when using MT – meaning that a human translator is involved in checking the accuracy of the translation in the post-editing phase.
Example use cases
- User-generated content such as reviews
- Multilingual technical files for medical devices
- Complex multi-regional user manuals
- eDiscovery or other reasons for parsing large sets of multilingual content
- Enable higher volumes of translation, faster
- Drastically reduce translation costs
- Translate into more languages with the same localization budget
Neural machine translation (NMT)
Neural machine translation is one of multiple approaches to implementing machine translation. In this case, machine translation is enabled by using an artificial neural network. In the beginning, rule-based systems were used, then they were replaced with statistical methods, and nowadays, we benefit from NMT. Compared to previous approaches, NMT has gained more popularity in recent years due to reduced time and memory consumption and better quality output.
NMT models can be trained in various ways based on a corpus of information. When using componentized content, the NMT model can ingest more granularly defined and specific content, due to the how structure content is defined. Being trained on granular data, as opposed to unstructured documents, allows organizations to develop, launch and integrate fully functional NMTs faster. Going one step further, if this structured content is also meta-tagged, the NMT model now has an even higher context level about the content it parses. This means the model will get to an appropriate translated form even faster.
Key benefits
- Increased machine translation precision for commercial or industry-specific content
- Reduced time for finalizing translations