This book contains a selection of refereed papers presented at the 3rd Workshop onMachineLearningforMultimodalInteraction(MLMI2006), heldinBethesda MD, USA during May 1a4, 2006. The workshop was organized and sponsored jointly by the US National - stitute for Standards and Technology (NIST), three projects supported by the European Commission (Information Society Technologies priority of the sixth FrameworkProgramme)atheAMIandCHILIntegratedProjects, andthePAS- CAL Network of Excellenceaand the Swiss National Science Foundation - tional research collaboration, IM2. In addition to the main workshop, MLMI 2006 was co-located with the 4th NIST Meeting Recognition Workshop. This workshop was centered on the Rich Transcription 2006 Spring Meeting Recognition (RT-06) evaluation of speech technologies within the meeting domain. Building on the success of previous evaluations in this domain, the RT-06 evaluation continued evaluation tasks in the areas of speech-to-text, who-spoke-when, and speech activity detection. The conference program featured invited talks, full papers (subject to ca- ful peer review, by at least three reviewers), and posters (accepted on the basis of abstracts) covering a wide range of areas related to machine learning - plied to multimodal interactionaand more speci?cally to multimodal meeting processing, as addressed by the various sponsoring projects. These areas - cluded humanahuman communication modeling, speech and visual processing, multimodal processing, fusion and ?ssion, humanacomputer interaction, andthe modeling of discourse and dialog, with an emphasis on the application of - chine learning.Third International Workshop, MLMI 2006, Bethesda, MD, USA, May 1-4, 2006, Revised Selected Papers Steve ... performance, with the ROVER result on the MPE outputs being 0.2% better (absolute) than the final ROVER system ... Of course, this inconsistency did notaffect the SDM submission, which achieved a 51.4%WER, since only onemicrophonechannelis used ... This is ofcourse due to cross-talkelimination and correct speaker segmentation in the manual transcripts, which alsoanbsp;...
|Title||:||Machine Learning for Multimodal Interaction|
|Author||:||Steve Renals, Samy Bengio, Jonathan Fiskus|
|Publisher||:||Springer Science & Business Media - 2006-12-22|