Every great conversation pulls you into a world you thought you knew—then flips the perspective. That’s exactly what happens in this episode of Localization Today, where host Eddie Arrieta sits down with Alexandra Feeley, Director of Market Development at Oxford Languages, part of Oxford University Press. From her decade at TransPerfect to her current role shaping Oxford’s language data strategy, Alexandra shares how one of the world’s most trusted linguistic institutions is tackling some of the toughest challenges in NLP and localization.
👉 Don’t just read about it—listen to the full conversation with Alexandra Feeley right here.
From Translation to Language Data
Alexandra’s career started on the ground floor of translation project management at TransPerfect. Over ten years, she witnessed seismic shifts: the adoption of neural machine translation, the rise of AI-powered interpretation, and the challenge of balancing speed with quality. Those experiences gave her a unique perspective: technology evolves quickly, but the foundation—language data—remains essential.
That insight led her to Oxford Languages. Unlike most LSPs, which rely on translation memories or proprietary data, Oxford has spent centuries curating authoritative linguistic resources. Today, that means digitalizing under-resourced languages, from Assamese to Catalan, ensuring they’re not left out of the AI revolution.
As Alexandra explains, when users in India or Africa open their phones, they shouldn’t be forced into English-only interfaces. Predictive text, spell check, and autocorrect exist in English by default—but without digitalized resources, other languages are locked out of those everyday conveniences. Oxford’s mission is to close that gap.
Building for the Future
What makes Oxford’s work stand out is its mix of tradition and innovation. Hundreds of lexicographers maintain and update corpora, definitions, and word metadata, ensuring that language data reflects how people actually speak today. These resources power everything from translation models to voice assistants, giving developers the tools to build culturally relevant user experiences.
The initiatives go beyond dictionaries. Oxford is developing sensitivity datasets to help AI models recognize not just vulgar terms but also contextually sensitive words—crucial for education, healthcare, and marketing. They’re also advancing industry-specific dictionaries, such as medical pronunciation datasets, designed to make AI interpretation safer in high-risk environments.
And then there’s the concept of World Englishes. If a voice assistant can understand and respond in a Nigerian, Indian, or Caribbean accent, it doesn’t just improve UX—it makes the technology feel like it belongs to the user.
A Mission-Driven Path
For Alexandra, the transition from a tech-heavy role to business development has been about connecting mission with market. Oxford’s work may be rooted in preservation, but it also feeds directly into partnerships with tech companies, LSPs, and global brands eager to reach new markets.
The takeaway? Language data is future-proof. AI models will change, but without high-quality linguistic input, the outputs will never reach the standard global users deserve.
Want more conversations like this one? Discover other episodes of Localization Today on Spotify, Apple Podcasts, or directly at Localization Today online.
