Relation extraction is an important task in the field of natural language processing, which aims to identify entities from the input text and represent the relationships between entities in the form of (subject, relationship, object) triples. Most of the current research uses deep learning techniques to extract intra-sentence relations, which takes a single sentence as input for relation extraction. However, in real scenarios, text usually consists of multiple sentences to express complex semantic relationships between entities. The existing intra-sentence relation extraction methods cannot consider the dependencies between sentences, so they cannot accurately identify the relations between cross-sentence entity pairs. To this end, this paper proposes a Cascading pointer Network based Inter-sentence Relation extraction method (CNIR), which captures the complex semantic relationships from the text by modeling the dependency between sentences. Specifically, the proposed method first encodes the input text and identifies the entity information contained in it, and then uses the sliding window technique to obtain the local information and context information of the text, to obtain the tail entity (i.e., object) information of each head entity (i.e., subject) in a specific relationship category. Thus, the relationships between entity pairs within and between sentences can be predicted. The proposed method is tested on the C-CLUE and DuIE2.0 datasets to verify its performance on the relation extraction task. Experimental results show that the proposed method outperforms the existing relation extraction methods.