Data formats for phonological corpora

The annotation of linguistic resources has long-standing traditions (see Cole et al., 2010). The other chapters of this book make clear that the production of annotated resources is a laborious, time-consuming, andexpensive task. In theory, we want to make these resources available in such a way that they can be re-used by as many scholars as possible (see Ide&Romary, 2002). However, a largevariety of annotation formatshave been developed in the previous decades, each one created for a specific research task. Consequently,the resulting resources are frequently only usable by members of the individual research projects. The goal of the present chapter is to explore the possibility of providing the research and industrial communities that commonly use spoken corpora with a set of well-documented standardised formats that allow a high re-use rate of annotated spoken resources and, as a consequence, better interoperability across tools used to produce or exploit such resources. We hope to identify standards thatcoverall possible aspects of the management workflow of spoken data, from the actual representation of raw recordings and transcriptions to high-level content-related information at a semantic or pragmatic level. Most of the challenges here are similar to those for textual resources, except for, on the one hand, the grounding relation that spoken data has to illocutionary circumstances (time, place, speakers and addressees), and, on the other hand, the specific annotation layers that correspond to speech related information (e.g. prosody), comprising multimodal aspects such as gestures. We should also not forget, as is well illustrated in this book, the importance of legacy practices in the spoken corpora community, most of them resulting from the existence of specific tools at various representation layers, ranging from basic transcription tools (Transcriber, PRAAT) to generic score-based annotation various tools do not have the same maintenance rate and capacity and it is therefore essential to think about standardised formats as offering the possibility to be embedded with existing practices. This implies that we have two basic scenarios in mind:  We want to be able to project existing data into a range of standardised representations that bear as little specificity to the original format as possible but as much faithfulness as necessary;  We want standardised formats to havethe capacity to be used for the development of new technical platforms, thus allowing the integration of new requirements and new features. These two general requirements both imply standards that can incorporate features and …