Adaptive eLearning Translation Represents Some Unique Challenges
Context is constantly changing, so translators can’t solely rely on it
Adaptive learning is a hot buzzword in eLearning, deservedly so. It is rapidly becoming the standard model for training and courses that are designed for remote delivery, as opposed to in-classroom experiences. Experienced teachers have learned to assess student progress through a variety of cues, targeted questions, and testing. They can then adapt their teaching to the student’s needs and level of knowledge. But these advantages largely disappear when the learning takes place at a distance via online courses.
When you add additional languages into the mix, there is even more risk of a disconnect between the student and the learning designer due to cultural differences and the quality of the translation. Adaptive learning systems help us bridge that gap. But they present some complexity for translators and localization project managers managing the translation processes. Fortunately, the existing translation technology tools we have help us navigate the increasing complexity of adaptive learning models.
eLearning is already non-linear, but adaptive eLearning is intentionally non-linear
Adaptive learning is a testing mode built into courses designed to test the learner’s actual level of knowledge versus what they think is their knowledge level. It’s that same assessment that in-person teachers learn to make but it is handled by digital processes. These processes are starting to encompass machine learning. The course teaches a concept, tests progress, then asks the learners what they consider their level of expertise to be at that point. It then assesses how mistaken they are about their knowledge level and serves up addition information and examples to improve their actual knowledge. Conversely, if the test reveals that they are, in fact, knowledgeable, they go on to the next level.
As a consequence of these adaptive situations, adaptive eLearning is non-linear, by intent. The course designer must anticipate how the assessment questions could be answered and have prepared a ‘library’ of responses to select from, depending on the situation at any given point in the teaching.
Imagine a database of responses without knowing what they are responding to
On the surface it would look like this presents a significant challenge for a translator to understand the context of what they are translating. Generally speaking to the process looks like this:
The course content is exported out of the courseware platform
The export has ID tags associated with each snippet of content. These tags will tell the courseware platform where the translated content belongs when it is re-imported into the platform
The exported content, with its tags, is run through translation memory which checks it against an existing repository of previously translated content to eliminate redundant translation of previously translated content.
The content requiring translation is sent to the translator and translated.
The process is reversed and the translated content ends up back in the courseware platform, ready for delivery.
This is a simplified workflow. In reality many elements, including video, audio, polling and testing, images, and other elements are part of the package. Often the multimedia elements are handled separately by specialists and then the project managers reassemble the translated package.
Context is determined by metadata and native language examples of the course for reference by the translator
The ID tags, which are meta data, keep things organized. Ideally the translator has little interaction with them. They are there to keep the many hundreds of text pieces and multimedia pieces in context through the workflow. They are not useful to the translator when they require context.
If the context is not a linear, logical progression of information, the translator should have access to a version of the course in the source language it was created in. Typically, this would be an HTML version viewable in a browser. The translator can then reference the original when they run into a confusing context issue. Since adaptive learning is non-linear, this can be a critical requirement in facilitating translation.
Delivering content digitally at distance is far more effective with adaptive learning models
Adaptive learning is, in our view, a critical improvement in the eLearning process that makes it much more effective. It is particularly valuable with localized content that may be delivered across thousands of miles to learners who may not have previously had the opportunity to become conversant in the subject matter. It takes down boundaries by improving the learning experience regardless of location.
Fine-tuning translation workflows for adaptive learning advancements
Because we have developed a business expertise in eLearning and training translation, we have been working with adaptive technology and processes on a daily basis as they come into use. We’ve adapted our workflows to accommodate their unique requirements and worked with our translators, editor/reviewers, and project managers to refine the processes and make best use of the available technology. We have developers who work with the project managers to set up the files and reassemble them. The goal is make adaptive learning translation faster, more accurate, and cost effective.