A Study On Speaker Classification Characteristics In Speaker Profiling: A Discourse Information Perspective

Speaker profiling plays an important role in disclosing a suspect’s characteristics and narrowing down police’s investigation scope. The present thesis is devoted to extracting a set of workable linguistic parameters for speaker profiling from the perspective of discourse information analysis.A qualitative approach is taken to conduct this study. First, a theoretical and analytical framework is constructed on the basis of Phil Rose’s classification of parameters for speaker profiling and Du Jinbang’s tree model of discourse information. Then, recording materials from real life situations are collected to extract a set of potential parameters for identifying a speaker’s biological, social and individual characters. Finally, an experiment is designed to test and verify the parameter set, along with the help of spectrographic analysis.It is found that information elements, information units, information development and the flow of information etc are workable parameters for identifying a speaker’s biological, social and individual characteristics. But the effectiveness of these parameters is more salient in real-world recording materials than in simulation experiments.This study provides a workable set of parameters for speaker profiling from a new perspective, i.e. discourse information analysis. It enriches the parameters for speaker profiling and broadens the application of the discourse information analysis as well. It is hoped that this study could help the police narrow the investigating scope and capture the criminal and facilitate the application of forensic linguistics in real-world situations.