The AMI Meeting Corpus is a multi-modal dataset consisting of 100 hours of meeting recordings, created as part of a project developing meeting browsing technology. It includes both naturally occurring and elicited meetings, with participants playing roles in design teams. The corpus is recorded using various devices, including microphones, cameras, and whiteboards, with all signals synchronized. It is also hand-annotated for various phenomena, including speech, discourse, emotions, and gestures. The dataset is being recorded, transcribed, and annotated to support research in speech, language, gesture, information retrieval, and organizational psychology. The AMI project is a multi-site, multi-disciplinary initiative aiming to develop meeting browsing technologies that improve work group effectiveness. The corpus includes meetings with non-native English speakers, providing a high degree of variability in speech patterns. The dataset is expected to be valuable for various research communities, including those working on speech, language, gesture, information retrieval, and organizational psychology. The corpus includes both elicited and naturally occurring meetings. Elicited meetings are designed to control variables, allowing for clearer analysis of group behavior. Naturally occurring meetings are used to validate findings from the elicited data. The dataset is being collected in parts, with the first part consisting of elicited material using a design task, the second part including less controlled elicitations for different tasks, and the third part consisting of naturally occurring meetings of various types. The goal is to combine data from disparate sources to better understand group behavior.The AMI Meeting Corpus is a multi-modal dataset consisting of 100 hours of meeting recordings, created as part of a project developing meeting browsing technology. It includes both naturally occurring and elicited meetings, with participants playing roles in design teams. The corpus is recorded using various devices, including microphones, cameras, and whiteboards, with all signals synchronized. It is also hand-annotated for various phenomena, including speech, discourse, emotions, and gestures. The dataset is being recorded, transcribed, and annotated to support research in speech, language, gesture, information retrieval, and organizational psychology. The AMI project is a multi-site, multi-disciplinary initiative aiming to develop meeting browsing technologies that improve work group effectiveness. The corpus includes meetings with non-native English speakers, providing a high degree of variability in speech patterns. The dataset is expected to be valuable for various research communities, including those working on speech, language, gesture, information retrieval, and organizational psychology. The corpus includes both elicited and naturally occurring meetings. Elicited meetings are designed to control variables, allowing for clearer analysis of group behavior. Naturally occurring meetings are used to validate findings from the elicited data. The dataset is being collected in parts, with the first part consisting of elicited material using a design task, the second part including less controlled elicitations for different tasks, and the third part consisting of naturally occurring meetings of various types. The goal is to combine data from disparate sources to better understand group behavior.