A framework for annotating animal bioacoustic data

Abstract

Annotation graphs (AGs) provide an efficient and expressive data model for linguistic annotations of time–series data [Bird and Liberman, 2001]. The term annotation here refers to the association of structured symbolic information with an extent of signal data. This paper demonstrates that AGs also provide an excellent framework for annotating time–series animal bioacoustic data. The AG model has several desirable properties for annotating animal bioacoustic signals. Partial or incomplete annotations can be represented naturally. Hierarchical structures, such as internal structures of animal communicative signals, can be stored and manipulated effectively. The model permits efficient query using the relational database technology. The spreadsheet-like editor component described in the paper provides an intuitive user interface for creating annotations of bioacoustic signals, such as monkey vocalizations, as AG data. The paper describes the tool used for annotating the vocalization of vervet monkeys, and the preliminary findings of the project are presented. Finally, the paper lays out a complete infrastructure for supporting the rapid development of annotation tools for animal bioacoustic data [S. Bird and M. Liberman, ‘‘A Formal Framework for Linguistic Annotation,’’ Speech Commun. 33, 23–60 (2001)].