The relationship of a complex sentence refers to the logical connection and semantic relationship between clauses in the sentence. Implicit complex sentences have certain relationships between clauses, but lack obvious conjunctions, requiring inference and identification through retrieval of other semantic information between the clauses. Therefore, the identification of complex sentence relationships is a challenging problem in the field of natural language processing. This paper proposes a model based on deep enhanced representation for recognizing the relationships between implicit binary clauses. The model utilizes the BERT model to transform the input sequence into word vectors, which are then fed into a multi-layer encoder module to capture the contextual features between sentences. This module is composed of stacked Bi-LSTM layers, with residual connections between each layer. In addition to being fed into the next layer encoder, the output of each layer also enters an interaction attention layer, aiming to better capture the relevance between two arguments. The outputs of the interaction attention layers are concatenated and then passed through a fully connected layer and softmax layer to perform the task of identifying implicit complex sentence relationships. The model was tested on a dataset extracted from the Chinese Complex Sentence Corpus (CCCS) and the Chinese Discourse Treebank (CDTB), achieving a highest accuracy of 79.34% and a highest macro-average F1 value of 78.77%, indicating the effectiveness of the method proposed in this paper.