Workshop Foreword
Multimedia content often contains spoken audio as a key component. Although speech is generally acknowledged as the quintessential carrier of semantic information, speech tracks are often ignored or not fully exploited by multimedia retrieval systems. In particular, the potential of speech technology to improve information access has not yet been successfully extended beyond multimedia content containing planned speech, such as broadcast news. The Searching Spontaneous Conversational Speech (SSCS) workshop series is dedicated to providing a forum for speech retrieval research as it expands into spoken content domains involving spontaneously produced, non-scripted, less-highly conventionalized, conversational speech characterized by wide variability of speaking styles and recording conditions. SSCS workshops aim to foster the development of robust, scalable, affordable approaches for accessing multimedia collections with a spoken component and to encourage sustained collaboration of researchers in the areas of speech recognition, audio processing, multimedia analysis and information retrieval.
The papers and demonstrations presented at SSCS 2009 report on research being carried out on techniques for information access in the domains of broadcast, meetings, interviews, telephone conversations, podcasts and speech tagging. A wide range of approaches are investigated, including using subword units, exploiting dialogue structure, fusing retrieval models, modeling topics and integrating visual features.
We would like to express our appreciation to the Program Committee for their work reviewing submissions for the workshop. We would also like to extend thanks to the research groups that submitted papers this year. Finally, a special word of thanks goes to the SSCS 2009 sponsors: AMI (Augmented Multi-party Interaction) Consortium the Netherlands Institute for Sound and Vision (Beeld en Geluid) and the PetaMedia Network of Excellence.
