Title :
Using Search Engine for Classification: Does It Still Work?
Author :
Govaerts, Sten ; Corthaut, Nik ; Duval, Erik
Author_Institution :
Dept. Comput. Sci., Katholieke Univ. Leuven, Leuven, Belgium
Abstract :
Genre classification is a key aspect of music descriptions. In 2006, Schedl et al. presented a method for genre classification through web-based co-occurrence analysis. We evaluate whether this method is still valid, given the evolution of the web search technologies. We identify some issues with page count as the main parameter for the analysis in relation with the used genre taxonomies, choice of search engine, etc. We show that the results vary over time. Depending on the required response time, we suggest two different strategies for music genre classification using this method. Finally, we discuss future work.
Keywords :
Internet; classification; music; search engines; Web search technologies; Web-based cooccurrence analysis; music genre classification; search engine; Automation; Bars; Computer science; Databases; Delay; Layout; Performance analysis; Search engines; Taxonomy; Web search;
Conference_Titel :
Multimedia, 2009. ISM '09. 11th IEEE International Symposium on
Conference_Location :
San Diego, CA
Print_ISBN :
978-1-4244-5231-6
Electronic_ISBN :
978-0-7695-3890-7
DOI :
10.1109/ISM.2009.103