Title: Improving Speaker Identification using Network Knowledge in Criminal Conversational Data

Abstract: Criminal investigations rely on the collection of conversational data. The identity of speakers must be assessed in order to build or improve the accuracy of an existing criminal network. Investigators use social network analysis tools to identify the most central character and the different communities within the network. We introduce Crime Scene Investigation (CSI) television show as a potential candidate for criminal conversational data. We also introduce the metric of conversation accuracy in the context of criminal investigations. In this paper, a speaker identification baseline is improved by re-ranking candidate speakers based on the frequency of previous interactions between speakers and the topology of the criminal network. The proposed method can be applied to conversations involving two or more speakers. We show that our approach outperforms the baseline speaker accuracy by 1.3\% absolute (1.5\% relative), and the conversation accuracy by 3.7\% absolute (4.7\% relative) on CSI data.