NEREL-BIO

NEREL-BIO provides nested named entity annotations and bilingual biomedical corpora to support cross-domain and cross-language named entity recognition research.

Key Features:

• Nested Named Entity Recognition: Identifies and annotates nested named entities in biomedical texts, including shorter entities embedded within longer ones that can span boundaries.
• Cross-Domain and Cross-Language Transfer: Supports domain transfer experiments from NEREL → NEREL-BIO and provides parallel English–Russian annotations of PubMed abstracts for cross-language research.
• Extensive Corpus: Contains over 700 Russian and 100 English biomedical PubMed abstracts, with each English annotation having a corresponding Russian counterpart.
• Annotation Scheme: Employs an annotation scheme that integrates general-domain entities and specialized biomedical entity types.

Scientific Applications:

• NER Model Development: Enables training and evaluation of NER models, including transformer architectures and machine reading comprehension frameworks, on nested-entity and bilingual corpora.
• Cross-Domain and Cross-Language Studies: Facilitates computational linguistics and biomedical text mining studies requiring adaptation across domains or analysis of multilingual datasets.

Methodology:

Explicit methods include nested named entity annotation, domain transfer experiments (NEREL → NEREL-BIO), parallel English and Russian PubMed abstract annotations, and evaluation with transformer-based and machine reading comprehension NER models.