Building Multilingual Pathways: Glossary-Informed Machine Translation
General description of the project
The Glossary-Informed Machine Translation and Multilingual Access Initiative at SUNY Empire State University was developed to address linguistic and accessibility barriers faced by Spanish-dominant students in online programs. Funded through an internal innovation grant, the project created a scalable translation workflow that integrates Microsoft Translator with a custom bilingual glossary and structured ChatGPT prompts, aligning translation with institutional terminology, tone, and accessibility standards (WCAG 2.1; UDL).
The initiative’s core success lies in combining human expertise with AI-driven efficiency. Two part-time linguists – both PhDs in Linguistics and certified translators – collaborated with the Digital Accessibility Coordinator to design and evaluate the process. Their glossary of more than 500 academic and administrative terms ensures consistency across course materials, website content, and student communications.
Evidence of success:
– Translations produced through the glossary-informed workflow required 50% less post-editing time than traditional manual translation.
– The workflow enabled the rapid localization of Brightspace course shells, reducing turnaround from 4–6 weeks to less than two.
– Research emerging from this initiative has been presented the ALBUS conference and won an award for best student paper.
The initiative reduced outsourcing costs by integrating Microsoft Azure’s existing enterprise license and leveraging open-source AI tools for post-editing support.
Lessons learned:
– Human oversight is essential: MT accuracy can increase dramatically when linguists supervise and train the system using curated glossaries.
– Terminology consistency builds trust: Institutionally coherent translations as more credible and welcoming.
– Sustainability requires integration: Embedding translation into existing workflows, rather than treating it as an afterthought, ensures continuity and scalability.
– Collaboration is key: Success depended on cooperation between accessibility staff, linguists, and instructional designers, reflecting the interdisciplinary nature of inclusive design.
When accessibility, language, and AI intersect intentionally, institutions can achieve cost-efficient, equitable, and replicable multilingual models that directly enhance student belonging and academic success.
Technologies
Microsoft Azure Translator – Custom Glossary Integration:
The bilingual glossary (Spanish–English) was uploaded as a CSV file to Azure’s Translator Document Translation service.
The glossary “locks” approved translations for academic and administrative terms—ensuring that phrases like credit hour, financial aid, or academic integrity are rendered consistently across all materials.
This customization directly improved translation accuracy and reduced revision time by 50%.
Explain project results
Since the Glossary-Informed Machine Translation initiative launched, SUNY Empire has translated over 280 course documents and student-facing resources into Spanish, including syllabi, assignment descriptions, academic integrity policies, financial aid guides, and orientation modules. Each document passes through the glossary-informed MT workflow and a human-review cycle led by the university’s linguists.
Why it should be considered best practice?
The Glossary-Informed Machine Translation and Multilingual Access Initiative represents a best practice because it transforms AI translation from a reactive accommodation into a proactive, equity-centered design model. It aligns technology, linguistics, and accessibility to ensure that multilingual learners can engage with course and support materials in a way that is accurate, culturally respectful, and cost-efficient.
Highlights of your proposed presentation
Live demonstration of glossary-informed translation in Microsoft Translator
Practical workflow guide for creating and maintaining a shared bilingual glossary
Human oversight is indispensable: Glossary-informed MT only succeeds when linguists guide the machine; automation alone cannot ensure equity or accuracy.
Accessibility and translation are inseparable: Treating linguistic access as part of digital accessibility fosters belonging and compliance simultaneously.
Iterative improvement sustains quality: Regular glossary updates, feedback loops, and student input keep translations current and authentic.
The Evaluation Committee will evaluate submitted proposals based on the following criteria. Each area will be rated on a scale from 1 to 5 (1= non-satisfactory; 5 =outstanding), for a maximum of 45 points.