Personality Recognition in Conversations using Capsule Neural Networks

Abstract

Automatic identification of personality in conversations has many applications in natural language processing, such as community role identification (e.g., group leader) in online social media conversations as well as meeting transcripts. Conversation utterances provide a lot of information about the parties involved in a conversation such as cues to the participants’ personality traits, one of human’s most distinguishable attributes. However, traditional computational personality assessment models rely on limited domain-knowledge and various psychometric indicators. In this paper, we propose a novel model based on capsule neural networks to extract meaningful hidden patterns from conversations and use them to assess the personality of individuals. Our experimental results on a real-world dataset reveals evidence that personality can be captured from conversation utterances outperforming traditional approaches. CCS CONCEPTS • Computing methodologies → Neural networks; • Applied computing → Psychology.