8th International Conference on Spoken Language Processing

Jeju Island, Korea
October 4-8, 2004

Mining Customer Care Dialogs for "Daily News"

Shona Douglas, Deepak Agarwal, Tirso Alonso, Robert Bell, Mazin Rahim, Deborah F. Swayne, Chris Volinsky

AT&T Labs-Research, USA

As real deployments of spoken dialog systems become more common, a wealth of information about their operation becomes available from their system logs. This paper describes the "VoiceTone Daily News" data mining tool for analyzing this information and presenting it in a readily comprehensible form suitable for use by either system designers or call center businesses. Relevant features are extracted from the logs of caller-system interactions and tracked by a trend analysis algorithm. Features that move outside their expected bounds on a given day generate headlines as part of a web site generated completely automatically from each day's logs. A "drilldown" facility allows headlines to be investigated all the way to viewing logs of individual interactions behind the headline and listening to the audio for individual turns. Some initial experiments with automated measures of dialog success are described as possible additional features to track.

Full Paper

Bibliographic reference.  Douglas, Shona / Agarwal, Deepak / Alonso, Tirso / Bell, Robert / Rahim, Mazin / Swayne, Deborah F. / Volinsky, Chris (2004): "Mining customer care dialogs for "daily news"", In INTERSPEECH-2004, 225-228.