Towards a Representation for Understanding the Structure of Multiparty Conversations

TitleTowards a Representation for Understanding the Structure of Multiparty Conversations
Publication TypeTechnical Report
Year of Publication2011
AuthorsEllsworth, M.
Other Numbers3222
Abstract

Dialog is a crucially important mode of communication. Understanding its structure is vital to a great number of technological innovations. Within the domain of meeting dialog applications alone, there is a need for navigating, summarizing, and extracting action items in recorded meetings and automated facilitation of meetings. These efforts, however, are hampered by a lack of agreement on how best to annotate meeting dialog to serve downstream applications.In this technical report, I describe of some of the efforts at representation and annotation of dialog acts, surveying some of the vast differences between approaches. In addition, in order to more directly compare different representations of discourse structure, two small, pilot annotations were carried out on a subset of the ICSI Meeting Corpus using different annotation schemes. The results support the idea that skew between different annotation systems renders them to some degree incompatible.

Acknowledgment

This work was partially supported by funding provided to ICSI through National Science Foundation grant CI-P: 0958561 (“Towards a Consensus Representation for Understanding Structure of Multiparty Conversations”). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors or originators and do not necessarily reflect the views of the National Science Foundation.

URLhttp://www.icsi.berkeley.edu/pubs/techreports/TR-11-003.pdf
Bibliographic Notes

ICSI Technical Report TR-11-003

Abbreviated Authors

M. Ellsworth

ICSI Research Group

Speech

ICSI Publication Type

Technical Report