Medieval Spanish (12th-15th centuries) named entity recognition and attribute annotation system based on contextual...

Area of Science:

• Computational Linguistics
• Digital Humanities
• Historical Linguistics

Background:

• Spanish medieval texts exhibit complex morphosyntax, inconsistent orthography, and diachronic variations, complicating named entity recognition.
• Named entities in this period often include complex structures like nicknames, social roles, and geographic origins.
• Existing named entity recognition (NER) systems struggle with the unique challenges posed by historical Spanish documents.

Purpose of the Study:

• To develop and implement a sophisticated named entity recognition and classification system tailored for Spanish medieval texts.
• To address the specific challenges of morphosyntactic complexity, orthographic variation, and diachronic/geographical linguistic changes.
• To accurately identify and categorize named entities, including their associated attributes, within historical Spanish corpora.

Main Methods:

• A context-aware system utilizing semantic cues for entity detection and type assignment.
• Parsing of entity contexts to determine type-specific dependencies for attached attributes.
• Implementation of a variant generator to manage phonetic and morphosyntactic evolution of medieval Spanish terms.
• Iterative enrichment of lexica, dictionaries, and gazetteers.

Main Results:

• The system achieved high performance on a corpus of over 3,000 manually annotated entities.
• F1 scores for named entity recognition ranged from 0.74 to 0.87 across different entity types and periods.
• Attribute annotation, specifically for person and role names, yielded an overall F1 score of 0.75.

Conclusions:

• The developed system demonstrates significant effectiveness in recognizing and classifying named entities in challenging Spanish medieval texts.
• The approach successfully accounts for linguistic variations and complex entity structures inherent in historical documents.
• This work provides a valuable tool for digital humanities research and the computational analysis of historical Spanish literature.