Studying The Linguistic and Psychological Markers in the Speech of Serial Killers

Serial killers have long fascinated criminologists, psychologists, and forensic experts due to their complex psychological profiles and violent behaviors. While extensive research has explored their psychological traits, the linguistic patterns in their speech and writing remain an under-explored area of forensic analysis. This study investigates the psycholinguistic and behavioral markers in serial killer discourse using Natural Language Processing (NLP), sentiment analysis, emotion classification, and forensic linguistics.

The research leverages a comparative linguistic analysis between serial killer interview transcripts and non-criminal speech samples from the MediaSum dataset. Key linguistic dimensions such as deceptive language, emotional tone, cognitive complexity, and thematic structures are analyzed to identify distinguishing characteristics of serial killer communication. The study employs RoBERTa-based models for fine-grained sentiment and emotion tagging, Linguistic Inquiry and Word Count (LIWC) for psycholinguistic profiling, and topic modeling for thematic extraction. Expert psychological profiles are incorporated to validate and triangulate the computational findings.

Findings indicate that serial killers exhibit distinctive linguistic patterns characterized by reduced emotional expressiveness, increased self-referential language, higher cognitive load, and markers of manipulation and detachment. Compared to non-criminals, their language also reflects greater thematic focus on control and dominance. These insights not only contribute to the growing field of forensic linguistics but also illustrate the potential of AI-driven methods in augmenting criminal profiling and investigative strategies.

Finally, the thesis addresses ethical considerations associated with the computational analysis of criminal language, particularly in contexts involving high-stakes forensic applications. This work underscores the interdisciplinary relevance of NLP, psychology, and criminology and proposes a scalable methodology for understanding violent criminal behavior through linguistic data.