Datasets and Models for Ontology-aligned Information Extraction in Portuguese Cultural Heritage

The datasets include annotations of entities and relations mapped to classes and properties from ArchOnto, an ontology designed for the Portuguese archives. The collection comprises both general-domain and domain-specific archival datasets, including manually transcribed and OCR-extracted documents from 20th-century Portuguese archives.

In addition to the annotated datasets, the repository provides fine-tuned NER models based on BiLSTM-CRF architectures with Portuguese language embeddings, supporting the development and evaluation of information extraction methods for archival records. The resources facilitate research on semantic enrichment, ontology-driven information extraction, and natural language processing for cultural heritage collections. **