CONF Favre_ACMMULTIMEDIA_2008/IDIAP Role Recognition for Meeting Participants: an Approach Based on Lexical Information and Social Network Analysis Favre, Sarah Salamin, Hugues Vinciarelli, Alessandro Hakkani Tür, Dilek Garg, N. P. EXTERNAL https://publications.idiap.ch/attachments/papers/2008/Favre_ACMMULTIMEDIA_2008.pdf PUBLIC https://publications.idiap.ch/index.php/publications/showcite/salamin:rr08-57 Related documents ACM International Conference on Multimedia Vancouver, Canada 2008 October 2008 This paper presents experiments on the automatic recognition of roles in meetings. The proposed approach combines two sources of information: the lexical choices made by people playing different roles on one hand, and the Social Networks describing the interactions between the meeting participants on the other hand. Both sources lead to role recognition results significantly higher than chance when used separately, but the best results are obtained with their combination. Preliminary experiments obtained over a corpus of 138 meeting recordings (over 45 hours of material) show that around 70% of the time is labeled correctly in terms of role. REPORT salamin:rr08-57/IDIAP Role Recognition for Meeting Participants: an Approach Based on Lexical Information and Social Network Analysis Garg, N. P. Favre, Sarah Salamin, Hugues Tür, D. Hakkani Vinciarelli, Alessandro EXTERNAL https://publications.idiap.ch/attachments/reports/2008/salamin-idiap-rr-08-57.pdf PUBLIC Idiap-RR-57-2008 2008 IDIAP To appear in Proceedings of ACM International Conference on Multimedia (2008) This paper presents experiments on the automatic recognition of roles in meetings. The proposed approach combines two sources of information: the lexical choices made by people playing different roles on one hand, and the Social Networks describing the interactions between the meeting participants on the other hand. Both sources lead to role recognition results significantly higher than chance when used separately, but the best results are obtained with their combination. Preliminary experiments obtained over a corpus of 138 meeting recordings (over 45 hours of material) show that around 70 percents of the time is labeled correctly in terms of role.