Recently, a large number of studies of spoken document retrieval and IR using by spoken queries have been reported. Most of them utilize a large vocabulary speech recognition system to transcibe spoken documents, then the transcribed documents are converted into an index for retrieving. In this paper, we describe a retrieval model to estimate retrieval performance in case of taking into acount various factors such as speech recognition rate and so on. We confirm the validness of results of simulation experiments using the retrieval model and estimated the retrieval performance under various conditions.
展开▼