COTHEC – Unified theory of coreference in Polish and its corpus-based verification

Project factsheet

Project summary

The aim of the project is development of a unified, systemic theory of coreference in Polish, its corpus-based verification and investigation of its computability, i.e. possibility of unambiguous use in the process of coreference resolution by computer. The project is focused on two important topics: verification and extension of existing theories of reference and coreference for Polish and possibility of their application in automated processes. In traditional understanding (e.g. by Topolińska and Padučeva) the theory of reference comprises mainly nominal realisations, since they are regarded as the only constructs featuring referentiality. The newest research, however, indicating e.g. occurrence of chains of indefinite articles carrying common reference, calls for extension of the referentiality notion with its cognitive understanding (resulting in a switch from extra-linguistic world of reference to the world of discourse), refining the typology of coreference and broad corpus-based verification of the theory. The linguistic-computational objective of the project is extension of the developed theory with its computational aspect, i.e. supplementing each coreference type with premises which could be used in the process of automated resolution of coreference. Motivation for this step are charges of incomputability of properties commonly used by existing coreference resolution models (such as Grosz’s Centering Theory).