Text this: Modeling audio dynamics using hierarchical assisted K-means model for structured speaker profiling in TED talks.