DOI: 10.18485/judig.2026.2.ch7 ISSN:

Named Entity Recognition and Relation Extraction in Serbian Football Text

Jelena Lazarević, Ranka Stanković, Olivera Kitanović
Show PDF Cite
The language of football represents a highly dynamic and semantically rich domain for Natural Language Processing (NLP), particularly for tasks such as: Named Entity Recognition (NER), Named Entity Linking (NEL), and Relation Extraction (RE). The discourse of football contains numerous entities and complex semantic relations including: players, clubs, coaches, competitions, transfers, injuries, and match events. This paper focuses on information extraction from the Serbian football corpus SrFudKo , originally developed for contrastive discourse analysis and corpus-based linguistic research. Previous work resulted in the construction of Serbian and Spanish football corpora, while semantic annotation tasks have not yet been systematically explored, making this precious linguistic resource as a basis for the knowledge graph construction as well as semantic search. The goal of this research is to enrich the Serbian football corpus with semantic annotations and structured relational information. The paper investigates the methods for recognizing football-specific named entities, linking them to external knowledge bases, such as Wikidata, and extracting semantic relations between entities. Paying special attention to the football-specific relation types, we included the relations between players, clubs, coaches, competitions, and events, as well as their organization within a domain-oriented ontology. In addition to presenting the corpus and the proposed ontology, the paper provides quantitative analysis and illustrative examples of extracted entities and their relations, based on the Serbian football-themed articles. The presented approach contributes to the development of semantically enriched football resources, suitable for knowledge graph population, and domain-specific NLP applications.

More from our Archive