Kurnia Cahya Febryanto, Izzat Aji Androfaza, Lalu Aldo Wadagraprana, Riyanarto Sarno, Kelly Rossa Sungkono, Yeni Anistyasari, Joko Siswantoro, A Min Tjoa
Heterogeneous Business Process Model and Notation (BPMN) platforms present critical integration challenges, as approximately 75% of large enterprises employ multiple modeling tools lacking unified transformation capabilities. Existing solutions address only single-format conversions or provide limited cross-platform compatibility without comprehensive validation. This paper presents a production-ready multi-format BPMN parser library uniquely integrating intelligent format detection, dual-tier validation, and optimized graph transformation within a unified architecture. The library utilizes specialized parsers for BPMN 2.0 XML, XML Process Definition Language (XPDL) 2.2, native formats, and Microsoft Visio diagrams through a plugin-based architecture. Multi-criteria detection algorithms automatically identify source formats with 99.2% accuracy by analyzing file signatures, XML namespaces, structural patterns, and content heuristics. The dual-tier validation framework ensures structural BPMN 2.0 compliance through rule-based constraints derived from official OMG specifications and semantic consistency through metadata quality assessment based on established process modeling guidelines, surpassing existing tools that perform only syntactic validation. The transformation pipeline generates standardized Cypher queries optimized for process mining workflows. Evaluation across 127 real-world business process models demonstrates 98.7% overall parsing accuracy, with format-specific performance ranging from 97.2% (Visio) to 99.8% (BPMN XML), achieving 85% reduction in transformation time compared to manual approaches. Released as open-source software via the Python Package Index with complete documentation, the library establishes foundational infrastructure for cross-platform business process intelligence, enabling unified graph-based analytics across heterogeneous modeling ecosystems without format-specific preprocessing. Copyright © 2026. Published by Elsevier B.V.
Department of Informatics, Institut Teknologi Sepuluh Nopember, Surabaya, Indonesia; Faculty of Engineering, Universitas Negeri Surabaya, Surabaya, Indonesia; Faculty of Engineering, Universitas Surabaya, Surabaya, Indonesia; Faculty of Computer Science, University of Vienna, Vienna, Austria
