Bidirectional Curriculum-Industry Alignment Framework: Replication Code for a Transformer-Based Framework with Domain Disparity Measurement for Indonesian Higher Education

This repository contains the replication code for the paper "Bidirectional Curriculum-Industry Alignment in Bilingual Contexts: A Transformer-Based Framework with Domain Disparity Measurement for Indonesian Higher Education". The code implements a bidirectional transformer-based framework for measuring semantic alignment between university curricula and industry job requirements in bilingual (Indonesian–English) settings. It introduces the Domain Disparity Index (DDI), a novel metric for quantifying representational balance across both alignment directions (Job→Curriculum and Curriculum→Job). Six sentence-embedding models were benchmarked against a TF-IDF baseline using 427 curriculum learning-outcome statements and 777 IT job requirement sentences. The pipeline covers bidirectional alignment scoring, code-mixing robustness evaluation, bootstrap statistical validation, human expert validation, and efficiency profiling. Contents: src/ — Five modular Python scripts reproducing Tables 2–7 from the paper notebooks/full_experiment.ipynb — Complete experiment notebook (Google Colab compatible) figures/framework_flowchart.py — Script to generate the framework diagram Data (not included here) must be downloaded separately from Mendeley Data: Curriculum corpus: https://doi.org/10.17632/dkrrdjs262.2 Job requirements corpus: https://doi.org/10.17632/zzy8v7ggkz.2