Making Sense of Italian Legacy Resources: Integrating ItalWordNet and PAROLE-SIMPLE-CLIPS for Sense Annotation

This paper discusses the creation of the Italian sense-annotated sub-corpus of a parallel corpus developed in the context of a large European initiative, and focuses on the reuse of legacy computational lexical resources to produce high-quality semantic annotation data for Italian. The work is situated within the current shift in Language Technology towards data quality, where expert-curated resources are increasingly valued for model evaluation, fine-tuning, and explainability. In the absence of an open, comprehensive digital dictionary for Italian, we adopted a bottom-up strategy based on the integration of two authoritative openly licensed legacy lexical databases, namely ItalWordNet and PAROLE-SIMPLE-CLIPS. These resources were semi-automatically merged and curated to construct a unified initial sense inventory, which formed the backbone of the manual sense annotation of a parallel Italian corpus carried out by expert linguists, and was subsequently refined and enriched as a direct outcome of the annotation process. Beyond the resulting dataset, the paper reflects on the challenges involved in merging heterogeneous legacy resources, the role of data reuse in countering digital obsolescence, their usefulness and limitations when applied to annotation tasks. In addition it reinforces reflections and best practices for sustainability, interoperability, and documentation, strongly advocated by Open Science and especially within large research infrastructures. Overall, the experience demonstrates how legacy lexical knowledge can be revitalized and made actionable for contemporary NLP and Digital Humanities research, provided consistent human work is dedicated to carry out or supervise the process. It also highlights the limitations of existing resources and the need for the continued creation of new, high-quality lexical data.