in All Things Localization

In a world of streaming entertainment and serious innovation, where do game localization and experimental language AI cross paths? This week, one common theme appears to be the distant past, as yet another natural language processing (NLP) algorithm tackles an ancient language without modern corollaries to see what AI can reveal about its remnant hieroglyphs – in this case, with ties to the game development world, by way of the relatively uncharted territory of ancient Egyptian.

Those keeping up with the CSOFT Blog know that NLP is a subject of particular interest to us as providers of technology-driven translation solutions. As the same fundamental technology that delivers machine translation, NLP has made headlines for its ability to generate original speech and text outputs based on its understanding of a language it has trained at predicting over and over, often with tremendous value to areas such as medical research. For people more interested in language itself than practical uses for it, NLP is a curiosity not easily depleted, adding to already fierce discussions of what language is and how it works. (Preeminent linguistics scholar Noam Chomsky weighed in just last month on the question of what the most famous of these algorithms, GPT-3, does and does not demonstrate.) Amid so much theory and practice, one of the more tangible breakthrough in 2022 has been the use of NLP to shed light on languages so lost to history that only patterns in the relics remain for scholars to sift through, as in our recent look at how researchers deciphered an ancient Greek precursor called Linear B. Now, in a sign of just how much these kinds of capabilities can yield in creative hands, Egyptologist Bree Kelly is reporting on her work with a Google AI model called Fabricius and developers at Ubisoft to enable people with minimal expertise to render authentic translations of ancient Egyptian artifacts, scaling the labor-intensive work that started with the Rosetta Stone.

Related:  Eating Your Words: How LSPs Can Help Agriculture

As Kelly writes, the push to leverage and advance Fabricius for Egyptologists’ use started with Ubisoft’s commitment to creating an authentic linguistic and cultural environment for its Egyptian-themed title, Assassins Creed: Origins. Recognizing the herculean effort that people adept at translating Egyptian go through when validating the meanings of glyphs, Ubisoft launched the Hieroglyphics Initiative not only to leverage machine learning in its own development process, but also to empower the outside work of experts who continue to discover secrets of the past. In some cases, machine learning was able to identify markings in Egyptian script that the naked eye could not. In others, it furnished a kind of workbench that helped the scholars better take stock of complex tasks involving symbols that different semantic meanings in different contexts. While not a generalized translation engine, Fabricius in this respect offers the same kind of support that machine translation offers to linguists engaged in tasks such as, appropriately, game localization. By allowing machines to do some of the heavy lifting, linguists and developers can turn their attention to the more challenging aspects of game localization and testing, when it comes to delivering a final product that feels integrated and consistent, rather than a sum of many separately translated elements. While Kelly is quick to identify the engine’s shortcomings, in all it reflects the most interesting elements of language AI and practical efforts in translation, particularly in their complex relationship as vital components in the delivery of cross-cultural media.

Whether it is a gaming title or streaming services, leveraging technology and human insight in tandem to tell stories across languages is at the heart of CSOFT’s approach to entertainment localization, as well as our complete range of AI-powered translation solutions in over 250 languages. Learn more about our services at!

[dqr_code size="120" bgcolor="#fff"]