Using AI to Create Multi-lingual Glossaries for Engineering Students

This case study has two dimensions. The first is student-facing: a structured glossary was created for each module to help all students engage with technical vocabulary. For international students, translations into their first language added a further layer of accessibility. The second is staff-facing: the pipeline and process were shared with colleagues through a 90-minute workshop, so that other teaching staff could replicate the same approach for their own modules.

The glossaries were produced using a five-step AI pipeline that any teaching staff member can follow without programming knowledge. The output is a searchable, interactive HTML file that can be published directly on Blackboard. Glossaries were developed across several modules taught by the author, covering programming, machine vision, robotics and autonomous systems, and control engineering, with cohorts ranging from approximately 15 to 90 students.

Aim of the case study

The student-facing aim was to improve vocabulary support across modules. A well-designed glossary helps all students engage with technical terminology more effectively, supporting comprehension, self-study, and exam preparation. For students whose first language is not English, providing a translation adds an additional layer of accessibility that standard module materials rarely offer. The aim was not to replace English-language learning, but to give students another route into understanding the same content.

Benefits of the case study

For students, a structured glossary provides a clear, searchable reference for technical vocabulary that is useful throughout the module, not only at revision time. For students who are not native English speakers, having definitions and terms available in their first language reduces the cognitive load of learning new concepts in a second language and gives them a more direct route to understanding. Both groups benefit; the translation simply adds an extra layer for those who need it.

Approach Taken

The pipeline has five steps, all carried out through a standard chat conversation with an AI tool such as ChatGPT or Microsoft Copilot.

Step 1: Generate the term list. The staff member provides the AI with their module name, student level, and a brief description of the main topics. The AI produces an initial list of 25 to 30 key terms.

Step 2: Review and refine. The AI is asked to explain its choices, identify gaps, and reorganize terms into logical categories. The staff member then removes terms that are too basic or irrelevant and adds any that are missing.

Step 3: Write definitions. The AI writes a short definition for each term (two to four sentences) at the appropriate student level. The staff member reviews and corrects these before proceeding.

Step 4: Build the HTML glossary. The finalized terms and definitions are passed back to the AI with a request to produce a self-contained HTML file. Features such as search, filters, collapsible definitions, and dark mode are added progressively in follow-up prompts.

Step 5: Translate. The completed English glossary is passed to the AI with an instruction to translate all visible text into the target language, leaving the HTML structure untouched. For right-to-left languages such as Arabic or Persian, the AI is also asked to adjust the layout direction.

The whole process, from a blank page to a published Blackboard glossary in two languages, was demonstrated live during a 90-minute staff workshop on 30 April 2026. Participants followed along and produced their own glossaries during the session.

Outcomes and Impact

For students, glossaries are now available across several modules in multiple languages. The Robotics and Autonomous Systems glossary alone covers six languages: Arabic, Burmese, French, Hindi, Polish, and Telugu. Student feedback has been informal but positive, with students from relevant language backgrounds reporting that the glossary helped them connect technical terms to concepts they already understood in their first language. Students studying in English as their first language also found the structured glossary a useful reference.

For staff, the workshop demonstrated that the full pipeline, from a blank page to a published Blackboard glossary in two languages, is achievable within a 90-minute session. By the end of the workshop, most participants had produced a working glossary and understood how to publish it on Blackboard. Several expressed interest in developing glossaries for their own modules.

Future Developments

The next steps are to share the pipeline and templates with the wider school and university, collect structured feedback from workshop participants as they use the resources with their students, and publish the findings as a case study or short paper by mid-2027. A presentation at a SHU learning and teaching event is also considered.

Longer term, the approach could be extended to other interactive learning formats. The same AI pipeline used for glossaries can produce flashcard activities, concept maps, and simple self-assessment tools, all as self-contained HTML files that work on Blackboard without any technical support.

Recommendations

1. Start with English. Produce and review the English glossary fully before translating. Errors in the source carry through to all translations.

2. Translate one language at a time, in a fresh conversation or a new message thread. This keeps the AI focused and reduces the risk of mixing languages in the output.

3. Use Gemini for translation where possible. Its translations of technical terminology tend to be more accurate than those from ChatGPT, likely because it draws on a broader multilingual index.

4. Always review AI-generated definitions before publishing. Definitions are generally accurate, but edge cases and module-specific nuances do occur. A short review before publishing is sufficient for most modules.

5. Add features one at a time. Asking for too many features in a single prompt often produces a less reliable result. Start with a basic file and add search, dark mode, and print functionality in separate follow-up prompts.

6. Involve students in review. Where possible, ask a native-speaking student from the cohort to check their language version. This takes very little time and significantly improves accuracy and trust.