Using Name Spotting in Audio/Video Media Identification to Improve Media Discovery Service in Digital Object Architecture

Digital object repository, a component of digital object architecture, stores large number of audio/video files (as digital objects) and provides access and retrieval to them. Sometimes metadata for audio/video files are almost absent. Lack of enough metadata limits media discovery service from fetching the files containing little metadata. In addition, the media discovery service excludes those files from the result set. Relevant information, such as names, can be extracted from the given content of an audio/video file and appended in metadata of the same audio/video file for enhancing the media discovery service. In this research, we use a Hidden Markov Model and Viterbi algorithm based name spotting module, known as IdentiFinder to extract names. The research will help to make large number of audio/video files visible to the media discovery service with the help of name extraction. Also, it will increase the user satisfaction by improving the search result set.

Published

2012-10

Content type

Original Research

Keywords

Media discovery service, Media Identification, metadata, Hidden Markov Model, Viterbi Algorithm, Name Spotting, Identifinder

DOI

10.5594/M001470

ISBN

978-1-61482-952-2