A relatively in-depth discussion of Support Vector Machines, including detailed and diagrammed breakdown of the mathematics of different types of SVMs and kernel functions. A comprehensive reference that is cited in numerous more recent works as a source of key background knowledge.
Chen, Lei, Sule Gunduz, and M. Tamer Oszu. 2006. “Mixed type audio classification with support vector machine”. Paper presented at the International Conference on Multimedia and Expo, July 9-12, 2006, Toronto, Canada. http://www3.itu.edu.tr/~sgunduz/papers/ICME06.pdf
The authors describe a SVM-based classifer trained on audio samples taken from video materials, classifying input as music, speech, environmental sounds, speech/music together, and music/environmental sounds together. The SVM system is found to perform better than classifers based in K-NN, Neural Net, and Naive Bayes algorithms. This kind of classification could have practical application in many tasks of audio content labelling, an area of concern in both the MIR and general LIS field.
Hearst, Marti A., Susan T. Dumais, Edgar Osuna, John Platt, Bernard Scholkopf. 1998. “Support vector machines.” IEEE Intelligent Systems 13(4): 18–28. http://www.cs.cmu.edu/~guestrin/Class/10701/readings/hearst98.pdf
A special "Trends and Controversies" issue on Support Vector Machines, comprising a series of short essays beginning with an introduction explaining the technology, followed by summaries of research on text categorization, face detection, and general application of SVMs. Though over ten years old, the publication clarifies SVMs and provides an interesting and accessable snapshot of the state-of-the-art at the time.
Manning, Christopher D., Prabhaker Raghavan, and Hinrich Schutze. “Support Vector Machines and Machine learning on documents”. In Introduction to Information Retrieval, Cambridge University Press. 2009. 319–348. http://nlp.stanford.edu/IR-book/pdf/irbookonlinereading.pdf
In this chapter, the authors provide a thorough and coherant overview of how SVMs operate and provide a comprehensive summary of how these algorithms have performed in text-classifying applications such as literature database search engines.
Temko, Andrey, Climent Nadeu, and Joan-Isaac Biel. 2007. “Acoustic event detection: SVM-based system and evaluation setup”. Paper presented at CLEAR’07.The researchers demonstrate a system that is based on a SVM trained to identify approximately twelve typical "meeting room sounds" such as speech, footsteps, shuffling paper, and stirring a tea cup. The output of the SVM is mirrored in an animation showing a character performing the events indicated by the detected sounds.
Video demonstration available at TALP Labs website, Technical University of Catalonia: http://www.talp.cat/talp/demos/AEDLdemo_UPC.avi