My Words Imply Your Opinion: Reader Agent-based Propagation Enhancement for Personalized Implicit Emotion Analysis

TL;DR

This paper tackles Personalized Implicit Emotion Analysis (PIEA) by introducing RAPPIE, a framework that simulates reader feedback via LLM-based reader agents and models reader interaction with a role-aware multi-view propagation graph. By integrating author attributes, simulated reader feedback, and propagation roles, RAPPIE enhances implicit emotion identification beyond author-centric signals. The work provides two new PIEA datasets (English and Chinese) and demonstrates substantial improvements over strong baselines, with macro-F1 gains up to 17.4% on one dataset. This approach highlights the value of reader-driven propagation dynamics for understanding implicit emotions in social media contexts and points to practical applications in personalized emotion-aware content analysis and moderation.

Abstract

The subtlety of emotional expressions makes implicit emotion analysis (IEA) particularly sensitive to user-specific characteristics. Current studies personalize emotion analysis by focusing on the author but neglect the impact of the intended reader on implicit emotional feedback. In this paper, we introduce Personalized IEA (PIEA) and present the RAPPIE model, which addresses subjective variability by incorporating reader feedback. In particular, (1) we create reader agents based on large language models to simulate reader feedback, overcoming the issue of ``spiral of silence effect'' and data incompleteness of real reader reaction. (2) We develop a role-aware multi-view graph learning to model the emotion interactive propagation process in scenarios with sparse reader information. (3) We construct two new PIEA datasets covering English and Chinese social media with detailed user metadata, addressing the text-centric limitation of existing datasets. Extensive experiments show that RAPPIE significantly outperforms state-of-the-art baselines, demonstrating the value of incorporating reader feedback in PIEA.

Figures (9)

• Figure 1: The illustration of the implicit emotion dynamic propagation, interaction, and enhancement between author and readers in social media.
• Figure 2: The framework of our proposed RAPPIE model. The entire study consists of four core modules: (1) utilizing reader information prompts to construct the reader agent for feedback simulation, (2) a global interactive multi-behavior overlapping network is constructed based on the simulated behaviors of reposting and reposting with a comment, as well as following, among different users, (3) a role-aware multi-view GNN is employed to learn users representations from multiple interactive perspectives, integrating reader propagation role during the graph learning process, (4) implicit emotion prediction is carried out by fusing content semantic with propagation role features and reader-feedback enhanced author embeddings.
• Figure 3: Impact analysis of the maximum iteration in multi-view interactive propagation graph learning.
• Figure 4: Influence of different reader scales.
• Figure 5: An example of the prompt template for reader agent creation.