G2P-DLP: An Open-Source System for Grapheme-to-Phoneme Conversion

Grapheme-to-phoneme (G2P) systems convert written text into phonetic representations, building the bridge between orthography and speech. The existing G2P tools for European Portuguese (e.g., Grafone, Convert Text to IPA Transcription) remain unreachable, limiting access to phonological data and thus hindering phonological research and education. This paper addresses this gap by introducing an open-access G2P system initially designed for 730 lemmas belonging to level 1 of the Vocabulário Fundamental of the Dicionário da Língua Portuguesa of the Academy of Sciences of Lisbon (DLP-ACL). Developed using open-source, widely supported technologies such as Python and regular expressions, this system provides a transparent, replicable, and extensible framework for large-scale translineation/syllabic division, stress marking, and phonetic transcription. It ensures technical sustainability through a flexible architecture that supports easy updates and adaptation to new corpora or languages (e.g., Mirandese, Spanish, etc.). This system’s criteria, regex rules, and processing pipeline are publicly available under an open license (GitHub), ensuring full transparency and accessibility and therefore supporting reproducible research. Applied to level 1 of the Vocabulário Fundamental, the system achieved 100% accuracy in translineation, 99.58% in stress marking and 97.5% in phonetic transcription, with the remaining inconsistencies mainly involving exceptional cases. Its scalability has been tested through its expansion to levels 2 and 3 of the Vocabulário Fundamental, and future work will extend testing to the more than 110,000 lemmas of the DLP-ACL. Openness ensures knowledge equity, making linguistic resources transparent, democratized, and usable across institutions and projects, thereby enabling inclusive research. By combining linguistic precision with open data principles, this project offers a findable, accessible, interoperable and reusable resource that illustrates how open linguistic infrastructures can strengthen education, foster participation, and promote inclusion in language research and documentation.