| WHITE PAPER |
Memsource: AI doesn’t have
to be showy to be effective
Memsource: AI doesn’t have to be showy to be effective
Introducing AI-Powered Automatic Linguist Selection
In 2017, Memsource established a dedicated AI team, and since then has released several features that have enhanced translation workflows. These include patented non-translatable recognition and most recently Memsource Translate, a dynamic machine translation management solution that selects the optimal engine based on the user’s language pair and domain.
Sometimes the best innovation is one that integrates so seamlessly with existing processes, that you barely notice it’s there.
In November 2019, we quietly launched our new Automatic Linguist Selection feature to help project managers make the most effective decisions when selecting translators for projects. The change was not visibly signaled in the UI, so some users might not have noticed any difference at first.
What does it do? Automatic Linguist Selection uses a machine learning algorithm to recommend the best possible translator for any given project based on their past experience.
The recommendation is powered by sophisticated domain detection. Whenever an organization uploads a document for translation, Memsource runs a keyword analysis to identify the nature of the content to establish its domain. Similar technology powers Memsource Translate, which sorts documents into one of 11 domains, such as Travel & Hospitality, Medical, and Legal & Finance documents.
The domain identification for Automatic Linguist Selection is far more detailed, recognizing over a thousand different domains, with each document being tagged with an appropriate combination of domains. This granular insight allows us to very carefully identify what makes each document unique, but also the many ways in which it relates to other documents processed through Memsource.
In addition to determining domains, Memsource keeps track of which linguists worked on which document. Over time, each linguist will develop a unique profile, with experience working with specific domains, which may be useful for future projects.
So, when a project manager creates a new project and uploads a document, our algorithm will analyze the domain of the content and create an ordered list of linguists with relevant experience. Using linguists with specific domain experience is always beneficial, as it ultimately leads to faster and better translations.
The AI recommends changes — but it doesn’t make the final call, the project managers do. However, our data shows that they rarely disagree with the algorithm. In approximately 70% of cases the project managers will choose a linguist from the list recommended by the AI.
Automatic Linguist Selection is already helping businesses translate faster and with better results, but it has the potential to develop further in several interesting directions.
For instance, Automatic Linguist Selection could be used to find a qualified translator from all of the users that work within Memsource, and not just with a given organization. This would be especially useful for businesses that need to quickly find qualified freelance translators.
Memsource’s AI technology is focused on ensuring that all users of the platform can work as effectively as possible.
You can learn more about how Memsource AI can help your business at: Memsource.com/AI.