Urdu speech corpus for travel domain

Speech corpus is a collection of recorded speech data. It is a fundamental component to develop any speech based applications. This paper presents the design and development of an Urdu speech corpus for travel domain. The corpus consists of a total of 250 vocabulary items including city names, days, time and numbers. The corpus has been used to develop a speech recognition system with a laboratory accuracy of 95.6% and a field accuracy of 87.21%. The corpus can be used for domestic flight queries, domestic flight reservation, train inquiry service and reservation and bus information and reservation.