Paper: A Composite Kernel Approach for Dialog Topic Tracking with Structured Domain Knowledge from Wikipedia

ACL ID P14-2004
Title A Composite Kernel Approach for Dialog Topic Tracking with Structured Domain Knowledge from Wikipedia
Venue Annual Meeting of the Association of Computational Linguistics
Session Main Conference
Year 2014
Authors

Dialog topic tracking aims at analyzing and maintaining topic transitions in on- going dialogs. This paper proposes a com- posite kernel approach for dialog topic tracking to utilize various types of do- main knowledge obtained fromWikipedia. Two kernels are defined based on history sequences and context trees constructed based on the extracted features. The ex- perimental results show that our compos- ite kernel approach can significantly im- prove the performances of topic tracking in mixed-initiative human-human dialogs.