ZAC.PB: An Annotated Corpus for Zero Anaphora Resolution in Portuguese
Simone Cristina Pereira · Recent Advances in Natural Language Processing · 2009
This paper describes the methodology adopted in the construction of an annotated corpus for the study of zero anaphora in Portuguese, the ZAC corpus. To our knowledge, no such corpus exists at this time for the Portuguese language. The purpose of this linguistic resource is to promote the use of automatic discovery of linguistic parameters for anaphora resolution systems. Because of the complexity of the linguistic phenomena involved, a detailed description of the different situations is provided. This paper will only focus on the annotation of subject zero anaphors. The main issues regarding zero anaphora in Portuguese are: indefinite subjects, either without verbal agreement marks or with first person plural or third person plural verbal agreement; position of the anaphor relative to its antecedent, i.e. anaphoric and cataphoric relations; coreference chains inside the same sentence and spanning several sentences; and determining the head of the antecedent noun phrase for a given anaphor. Finally, preliminary observations taken from the ZAC corpus are presented.