PORTLEX_LLM_didactic_modules_EURALEX2026

This deposit contains the research artefacts associated with the EURALEX 2026 paper “Can Lexicographic Entries Embed Didactic Modules? A Case Study on LLM-Generated Outputs for a Dictionary Interface”, by Carlos Valcárcel Riveiro, María Teresa Sanmarco Bande, Iván Arias Arias and María José Domínguez Vázquez.

The archive includes 24 interactive HTML didactic modules generated from structured PORTLEX dictionary entries. PORTLEX is a multilingual valency-oriented dictionary of the nominal phrase. The modules were produced as experimental outputs in order to evaluate how large language models can transform curated lexicographic data into pedagogically oriented activities for language learners.

The dataset covers three languages — German, French and Spanish — and four nominal entries per language. For each entry, two modules were generated: one with a GPT-based custom system and one with a Gemini-based system. Each HTML module is autonomous and self-contained, with internal CSS and JavaScript and no external dependencies.

The archive also includes the system prompt, the extended master prompt used as operational specification, a README file and a module inventory file. These materials are provided to document the experimental conditions and to support reproducibility.

The generated modules should be interpreted as experimental research artefacts, not as validated teaching materials ready for classroom use. They may contain linguistic, pedagogical or technical limitations, which are precisely part of the object of evaluation in the associated study.

Archive structure:
- modules/chatgpt/: HTML modules generated with the GPT-based system.
- modules/gemini/: HTML modules generated with the Gemini-based system.
- prompts/: system and master prompts used for generation.
- metadata/: module inventory.
- README.md: documentation of the archive contents.