There are several different types of information inherent in conversations: speech features, participants, and content. By recording conversations in organizations and aggregating this information, we claim that high-potential collaborations and expertise can be identified. Preliminary results for learning context from computer transcribed conversations show encouraging results. Current research involves collecting a larger conversation dataset to be used for social network analysis, knowledge management and as training data for a simulation to model organizational disruptions.
PDF . Full list of tech reports