Coreferential Relations in Basque: The Annotation Process

Area of Science:

• Computational Linguistics
• Corpus Linguistics
• Pragmatics

Background:

• Coreference is a pragmatic phenomenon influenced by context and language-specific patterns.
• Basque, a non-Indo-European language, possesses distinct grammatical features compared to its neighbors.
• Previous studies have automated mention detection, improving the efficiency of corpus annotation.

Purpose of the Study:

• To present the coreferential tagging of a portion of the EPEC Corpus of Basque.
• To explain the language-specific features of Basque relevant to coreference.
• To describe the criteria and annotation process for coreferential tagging in Basque.

Main Methods:

• Utilized a morphologically and syntactically annotated corpus for easier identification of reference chains.
• Employed a multi-annotator approach with independent tagging and a judge for resolving disagreements.
• Leveraged automated mention detection from prior research to streamline the annotation workflow.

Main Results:

• Successfully tagged a segment of the EPEC Corpus for coreference.
• Established clear criteria for coreferential tagging tailored to Basque.
• Developed an automated process for coreferential tagging, enhancing efficiency.

Conclusions:

• The developed methodology provides a robust framework for coreferential tagging in Basque.
• The annotated corpus serves as a significant resource for computational treatment of coreference resolution.
• This work facilitates future research in Basque natural language processing and computational linguistics.