Mapping Nonverbal Communication into Social Status: Automatic Recognition of Journalists and Non-journalists in Radio News

This work shows how features accounting for nonverbal speaking characteristics can be used to map people into predefined categories. In particular, the results of this paper show that the speakers participating in radio broadcast news can be classified into journalists and non-journalists with an accuracy higher than 80 percent. The results of the approach proposed for this task is compared with the effectiveness of 16 human assessors performing the same task. The assessors do not understand the language of the data and are thus forced to use mostly nonverbal features. The results of the comparison suggest that the assessors and the automatic system have a similar performance.