Code for my MSc project on "Morphosyntactically constrained sentence generation for vocabulary practice".
The repository has two separate parts, matching the distinction in the report:
| Part | Role | Location |
|---|---|---|
| Experimental framework | CLI pipeline that runs generation methods on benchmarks, scores outputs with a shared metric suite, and stores results in SQLite experiment databases. This is what produces the technical-method and transfer results. | research/ |
| User-facing learning prototype | Interactive vocabulary / sentence-practice app. Demonstrates how the research could sit inside a learning product. It is not connected to the experimental framework and is not used for the reported experiments. | frontend/ + backend/ |
research/ Experimental framework (benchmarks, methods, evaluation, cluster scripts)
frontend/ Next.js learning prototype UI
backend/ FastAPI service for the prototype
docs/ Specs, experiment notes, and report-writing material
Setup for each part is in its own README: research, frontend, backend.