Combining Evidence for Improved Speech Retrieval

The goal of my dissertation research is to investigate the combination of new evidence sources for improving information retrieval on speech collections. The utility of these evidence sources is expected to vary depending on how well they are matched to a collection’s domain. I outline several new evidence sources for speech retrieval, situate them in the context of this domain dependency, and detail several methods for their combination with speech recognition output. Secondly, I highlight completed and proposed work for the production of this evidence.

• Type: Conference
• Year: 2007
• Where: NAACL
• Authors: J. Scott Olsson
• Keywords: Computational Linguistics, Dissertation Research, Evidence Sources, NAACL 2007, Speech Collections