The exponential growth of unstructured text data on the internet presents a significant challenge in extracting entity relation entity triplets. Complex overlaps and trans-missions of relations between entities within the same text paragraph further exacerbate this challenge, often leading to contradictory results in extraction models. In this paper, we propose an innovative approach, the Relation Extraction Model based on Overlap Rules and Abductive Learning (ORABL), to address these issues. ORABL integrates an initial classifier with a rule module that identifies and corrects contradictions in the initial predictions. This iterative process continues until the model no longer produces contradictory conclusions, thereby improving the consistency of the extracted relations with natural language habits. Through extensive experimentation, we demonstrate that ORABL outperforms state-of-the-art models in various scenarios, including single entity overlap (SEO), entity pair overlap (EPO), and Subject Object Overlap (SOO). Our results underscore the effectiveness of ORABL in handling the complexities of relation extraction from unstructured text data. The results on FewRel indicate that it also has certain potential in solving the few-shot problem.