From archive to corpus: transcription and annotation in the creation of signed language corpora

From archive to corpus: transcription and annotation in the creation of signed language corpora* Trevor Johnston Department of Linguistics, Macquarie University, Sydney, Australia trevor.johnston@mq.edu.au Abstract. The essential characteristic of a signed language corpus is that it has been anno- tated, and not, contrary to the practice of many signed language researchers, that it has been transcribed. Annotations are necessary for corpus-based investigations of signed or spoken languages. Multi-media annotation software can now be used to transform a recording into a machine-readable text without it first being necessary to transcribe the text, provided that linguistic units are uniquely identified and annotations subsequently appended to these units. These unique identifiers are here referred to as ID-glosses. The use of ID-glosses is only possible if a reference lexical database (i.e., dictionary) exists as the result of prior foundation research into the lexicon. In short, the creators of signed language corpora should prioritize annotation above transcription, and ensure that signs are identified using unique gloss-based annotations. Without this the whole rationale for corpus-creation is un- dermined. Keywords: corpus linguistics, corpora, annotation, sign language, language documentation, Auslan (Australian Sign Language). 1 Introduction A modern linguistic corpus is something more than just a reference dataset of written or tran- scribed texts of a particular language on which a description of a language is based. If anything, this is an old fashioned sense of corpus. Rather, a corpus in the modern sense is a collection of spoken and spoken texts in a machine-readable form that has been assembled for the purposes of studying the type and frequency of lexical items and grammatical structures and construc- tions in a language (McEnery and Wilson, 2001). A modern linguistic corpus contains linguistic annotations and appended sociolinguistic and sessional data (metadata) that describe the partici- pants and the circumstances under which the data were collected. With the development of digi- tized video recording and multi-media annotation software, signed language corpora can now be described as sub-types of ‘spoken’ (i.e., ‘face-to-face’) language corpora. Signed language corpora promise to vastly improve peer review of descriptions of signed lan- guages and make possible, for the first time, a corpus-based approach to signed language analy- sis. Corpora are important for the testing of language hypotheses in all language research at all levels, from phonology, through lexis, morphology and syntax to discourse (Baker, 2006; Halli- day et al., 2004; Hoey et al., 2007; McEnery et al., 2006; Sampson and McCarthy, 2004; Sin- clair 1991). There are several reasons why this is especially true of deaf signing communities. First, signed languages—inevitably young minority language communities—lack written forms * Acknowledgments: research towards this paper was supported by a Hans Rausing Endangered Lan- guages Documentation Project (University of London) grant #MDP0088 to Trevor Johnston. Additional archived corpus material was contributed through Australian Research Council (ARC) grant #LP09346973 to Adam Schembri and Trevor Johnston. Additional corpus annotation were partially sup- ported by Australian Research Council grant #DP0665254 to Trevor Johnston and Louise de Beuzeville. I also wish acknowledge the support …

Content truncated to 500 words