Capturing Chat: Annotation and Tools for Multiparty Casual Conversation.

Capturing Chat: Annotation and Tools for Multiparty Casual Conversation. Emer Gilmartin, Nick Campbell Speech Communication Lab Trinity College Dublin gilmare@tcd.ie, nick@tcd.ie Abstract Casual multiparty conversation is an understudied but very common genre of spoken interaction, whose analysis presents a number of challenges in terms of data scarcity and annotation. We describe the annotation process used on the d64 and DANS multimodal corpora of multiparty casual talk, which have been manually segmented, transcribed, annotated for laughter and disfluencies, and aligned using the Penn Aligner. We also describe a visualization tool, STAVE, developed during the annotation process, which allows long stretches of talk or indeed entire conversations to be viewed, aiding preliminary identification of features and patterns worthy of analysis. It is hoped that this tool will be of use to other researchers working in this field. Keywords: multimodal corpora, casual conversation, annotation 1. Introduction Casual conversation is a fundamental feature of human so- cial life, seemingly present in most situations where peo- ple congregate. In order to better understand the workings of casual conversation, suitable data are needed to inform studies of the various aspects of this most common of hu- man activities. Although there are increasing numbers of high quality annotated multimodal corpora of spoken in- teraction available, there remain very few collections of longer stretches of casual talk among multiple participants. While it is tempting to use some of the rich resources cur- rently available, it is unclear that results obtained from them would apply to structurally different casual talk. Therefore, we have undertaken a programme of collection, annotation, and analysis of multiparty casual talk data. In this paper, we outline the annotation process and tools developed dur- ing our ongoing programme of research into multiparty ca- sual conversation, which aims to deepen understanding of this very common modality of spoken interaction, and in- form the design of human-machine interaction. We briefly discuss models of social or casual conversation from the literature, provide an overview of corpora currently avail- able and their limitations for our requirements, describe the annotation procedures we have developed in our work on two corpora of long (one hour or more) multiparty conver- sations, and introduce a visualisation tool developed to aid in preliminary examination and analysis of these conversa- tions. 2. Social Talk Casual social conversation is described as ‘talking just for the sake of talking’(Eggins and Slade, 2004), and its sub- genres include smalltalk, gossip, and conversational narra- tive. Aimless social talk or ‘phatic communion’ has been described as an emergent activity of congregating people, and viewed as the most basic use of speech (Malinowski, 1923). Researchers in fields including anthropology, evo- lutionary psychology, and communication have theorized that the such talk functions to build social bonds and avoid unfriendly or threatening silence, rather than simply to ex- change information or express thought, as postulated in much linguistic theory. Instances of these views are found in the phatic component in Jakobson’s model of communi- cation (Jakobson, 1960), distinctions between interactional and instrumental language …

Content truncated to 500 words