Prior research on Aspect-based Sentiment Analysis (ABSA) has mainly centered on sentence-level analysis, with document-level ABSA garnering limited attention. However, document-level sentiment analysis is crucial as it provides a more comprehensive understanding of the overall sentiment towards different aspects in long-form texts. In this paper, we propose a context interaction awareness method for document-level ABSA, employing a bi-graph reasoning mechanism to establish connections among sentences within long documents. Specifically, we create a structural reasoning chain for long documents by linking mentions within sentences. Meanwhile, a semantic reasoning chain is built based on semantic role relationships. Ultimately, by forming a document structure-aware graph and a document semantic-aware graph, we enhance the comprehension of context interaction within the document. Experimental results on public datasets demonstrate that our method surpasses advanced models. Furthermore, the rationality and effectiveness of our model structure have been thoroughly analyzed and validated.