Disambiguation of Annotated Text of Audio Using Ontologies
Latifur Khan, Dennis J. McLeod · 2000
To improve the accuracy in terms of precision and recall of an audio information retrieval system we have created a domainspecific ontology. Taking into account the shortcomings of keyword-based techniques, we have opted to employ a conceptbased technique utilizing ontology. The key problem is the identification of appropriate concepts within annotated audio text. In the case of the association of irrelevant concepts with audio objects there is a loss of precision. On the other hand, if relevant concepts are discarded, a loss of recall will ensue. Therefore, we have proposed a novel automatic disambiguation algorithm which prunes as many irrelevant concepts as possible while at the same time retaining the largest possible number of concepts which are relevant. Through the use of techniques for the association of concepts in the ontology, we have devised a method for determining the correlation of concepts and are thus able to employ a correlation factor as a basis for the selection of ...