Towards a Multi-dimensional Taxonomy of Stories in Dialogue

Towards A Multi-Dimensional Taxonomy Of Stories In Dialogue Kathryn J. Collins(1), David Traum(2) (1) Swarthmore College (2 )Institute for Creative Technologies, University of Southern California kcollin3@swarthmore.edu, traum@ict.usc.edu Abstract In this paper, we present a taxonomy of stories told in dialogue. We based our scheme on prior work analyzing narrative structure and method of telling, relation to storyteller identity, as well as some categories particular to dialogue, such as how the story gets introduced. Our taxonomy currently has 5 major dimensions, with most having sub-dimensions - each dimension has an associated set of dimension-specific labels. We adapted an annotation tool for this taxonomy and have annotated portions of two different dialogue corpora, Switchboard and the Distress Analysis Interview Corpus. We present examples of some of the tags and concepts with stories from Switchboard, and some initial statistics of frequencies of the tags. Keywords: Story, Narrative, Identity. 1. Introduction Stories have long been recognized to be an important part of many genres of dialogue, including casual conversation (Sacks, 1992; Labov and Waletzky, 1967; Polanyi, 1985). Previous analyses have explored the structure of the nar- rative, e.g., (Labov and Waletzky, 1967), how stories are introduced and co-constructed in dialogue (Sacks, 1992; Jefferson, 1978), and how stories relate to cultural iden- tity (Polanyi, 1985). Schank describes some dimensions toward classifying stories with respect to retrieval (Schank, 1991). In this paper, we introduce a taxonomy of stories that can be used for classifying and annotating how sto- ries are told in dialogue. Our ultimate goals are to be able to automatically recognize aspects of stories and be able to retrieve and re-tell stories in appropriate places in dia- logue. The taxonomy includes multiple dimensions relat- ing the story and the way it is told to the narrator and di- alogue function. We have used this taxonomy to annotate several dialogue corpora, and we report some preliminary results from the Switchboard corpus (Godfrey and Holli- man, 1993). 2. Corpora The Switchboard: Telephone Speech Corpus for Research and Development corpus was collected between 1990 and 1991. It was sponsored by DARPA and collection was au- tomated and completed by Texas Instruments. It was col- lected for use in research. Some possible areas of study on the corpus suggested at time of publication were speaker verification and large vocabulary speech recognition (God- frey et al., 1992). Since its collection, the Switchboard corpus has been re- leased several times with updated annotations. It has addi- tionally been used by researchers for a wide range of speech annotation projects. The release of the corpus which we use in our study had been previously annotated by dialogue act. Although these annotations sometimes correspond with our use of the corpus, we primarily annotate the corpus on a higher level, with multiple utterances in each block. Our goal was to identify groups of utterances by a single speaker which constituted a story. 3. Identifying stories Identification of stories to be annotated by this scheme is based on Labov’s narrative structure (Labov and Waletzky, …

Content truncated to 500 words