Replication Package — From Natural Language to Executable Filter Code: LLM-Assisted Behavioral Customization in Trigger-Action Platforms

Replication package for the article "From Natural Language to Executable Filter Code: LLM-Assisted Behavioral Customization in Trigger-Action Platforms" (Behaviour & Information Technology, 2026). Contents:- data/ — IFTTT-derived rule datasets (train/val/test), trigger/action catalogs, and the representativeness/API-validation CSVs.- data/user_study/ — raw questionnaire data (SUS, NASA-TLX, TOAST), scenario sessions, participants, and a self-contained pipeline (compute_metrics.py, analysis.ipynb) reproducing every user-study metric reported in the paper.- scripts/ — prompt templates, LoRA/SFT fine-tuning, and representativeness analysis.- figures/ — architecture and interface figures. Reproduce the user-study results: pip install numpy pandas scipy cd replication_package/data/user_study python compute_metrics.py