Thought chains represent coherent sequences of value-driven reasoning paths derived from interconnected beliefs and knowledge. In this paper, we propose a novel framework for extracting such thought chains from a Value-Belief Network using relation similarity. Our approach computes semantic similarities between relations connecting value and belief nodes to infer and extract plausible reasoning paths. To enrich the psychological realism of these chains, we incorporate emotion vectors based on the VAD model into relation embeddings. The framework integrates vector-based embeddings, relation similarity metrics, and graph traversal strategies to uncover implicit thought chains. We propose a Relation Similarity Based Thought Chain Extraction Mechanism and provide formal definitions, present the algorithmic structure, and validate the model through simulation with synthesized and ConceptNet-based data. The proposed method demonstrates robust coherence in extracted chains and outperforms baseline approaches in accuracy and interpretability.