Phoneme based spoken document retrieval
conference paper
Since speech recognition technology has become more and more mature, retrieval of spoken documents has become a feasible task. We report about two cases, which aim at scalable and effective retrieval of broadcast recordings. The approach is based on a hybrid architecture, which combines the speed of off-line phoneme indexing and precision of wordspotting while maintaining a scalable architecture, which allows for frequent updates of the database where out-of-vocabulary (OOV) words are abundant. A pilot experiment has been done on a small database of recordings of a Dutch talkshow. A more extensive evaluation took place in the framework of the Spoken Document Retrieval track of TREC7 on English broadcast news.
Bij Spoken Document Retrieval zijn twee technieken vergeleken: foonherkenning met indexering, en wordspotting. Het blijkt dat een hybride techniek de beste resultaten geeft.
Topics
TNO Identifier
12359
Source title
Proceedings of Twente workshop on Language Technology (TWLT14): Language Technology in Multimedia Information
Retrieval, December 1998
Retrieval, December 1998
Pages
1-12
Files
To receive the publication files, please send an e-mail request to TNO Repository.