Quan Zou
Pronouns: she/her
Research Mentor(s): Jonathan Brennan
Research Mentor School/College/Department: Linguistics / LSA
Program:
Authors:
Session: Session 3: 11:00 am – 11: 50 am
Poster: 46
Abstract
Conversation is the basis of human communication and plays a vital role in knowledge and information sharing. Alignment in conversation facilitates mutual understanding or connection between interlocutors to conversational success (Meneti et al., 2012). Studies in natural language processing (NLP) use computational methods to quantify multidimensional (i.e., lexical, semantic, affective) psycholinguistic alignment between interlocutors based on conversation transcripts (Ulichney et al., 2023). Variations in the affective valence of words can signal shifts in the emotions of interlocutors, while changes in semantic distance within exchanges can indicate shifts in conceptual topics. Both aspects have the potential to influence the effectiveness of a conversation. Understanding how these variations in affective and semantic dimensions align between interlocutors during conversations may provide insights into how they either facilitate or compromise human communication. To investigate this, we analyze seven text transcripts transcribed from audio recordings of previously collected conversation data, each representing a 60-minute pairwise, undirected naturalistic interaction. Affective variation is defined by the change in word affective values, which are derived from the pairwise comparison of each word in a language transcript to a specific anchor word from an emotion database (AffectVec) using a vector-based embedding approach. Semantic distance is quantified by converting each turn to a semantic vector and comparing the cosine distance between vectors of the two participants per exchange. We examine whether there is a correlation between the variations in affective and semantic distance alignment during conversation. The study provides an analytic perspective for future research examining the psycholinguistic alignment between interlocutors in natural conversations.
