Ontology-based speech act identification in a bilingual dialog system using partial pattern trees

  • Authors:
  • Jui-Feng Yeh;Chung-Hsien Wu;Ming-Jun Chen

  • Affiliations:
  • Department of Computer Science and Information Engineering, National Chiayi University, Chiayi, Taiwan;Department of Computer Science and Information Engineering, National Cheng Kung University, Tainan, Taiwan;Department of Computer Science and Information Engineering, National Cheng Kung University, Tainan, Taiwan

  • Venue:
  • Journal of the American Society for Information Science and Technology
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

This article presents a bilingual ontology-based dialog systemwith multiple services. An ontology-alignment algorithm is proposedto integrate ontologies of different languages for cross-languageapplications. A domain-specific ontology is further extracted fromthe bilingual ontology using an island-driven algorithm and adomain corpus. This study extracts the semantic words-conceptsusing latent semantic analysis (LSA). Based on the extractedsemantic words and the domain ontology, a partial pattern tree isconstructed to model the speech act of a spoken utterance. Thepartial pattern tree is used to deal with the ill-formed sentenceproblem in a spoken-dialog system. Concept expansion based ondomain ontology is also adopted to improve system performance. Forperformance evaluation, a medical dialog system with multipleservices, including registration information, clinic information,and FAQ information, is implemented. Four performance measures wereused separately for evaluation. The speech act identification ratewas 86.2%. A task success rate of 77% was obtained. The contextualappropriateness of the system response was 78.5%. Finally, the ratefor correct FAQ retrieval was 82%, an improvement of 15% over thekeyword-based vector-space model. The results show the proposedontology-based speech-act identification is effective for dialogmanagement. © 2008 Wiley Periodicals, Inc.