Home // eKNOW 2026, The Eighteenth International Conference on Information, Process, and Knowledge Management // View article

Persona-Conditioned Emotion Classification of Conversation Using LLMs

Authors:
Israel Cuevas
Andrew Mackey
Susan Gauch

Keywords: natural language processing; emotion analysis; large language models.

Abstract:
Large Language Models (LLM) have demonstrated success across a wide range of tasks in the field of natural language processing, including within the emotion classification task of language. With the recent advancements of agentic workflows and conversational chatbots in the field of artificial intelligence, it is fairly common to employ the use of personas to bias LLM interactions toward domain-specific applications. In this study, we investigate the impact of persona-conditioned models for the task of emotion classification along with model confidence of performance under these persona-conditioned settings. Our statistically-significant results ($p < 0.001$) demonstrate that persona-conditioned models affect model performance while also demonstrating the performance differences between each of the personas. Furthermore, through our experiments we observed variations in model confidence between both open and closed LLMs for the Emotion Recognition in Conversation (ERC) task.

Pages: 36 to 43

Copyright: Copyright (c) IARIA, 2026

Publication date: May 24, 2026

Published in: conference

ISSN: 2308-4375

ISBN: 978-1-68558-387-3

Location: Venice, Italy

Dates: from May 24, 2026 to May 28, 2026