In the field of natural language processing, sentiment analysis is a key task to understand the emotional tendency of text, but polysemy and insufficient semantic feature extraction often lead to difficulties in analysis. This study proposes a BERT-BiLSTM-KAN (BBK) model aimed at solving these problems. First, BERT model pre-training technology is used to convert Chinese text into high-dimensional matrix vectors to fully capture the contextual information of words, phrases,and sentences. Subsequently,the bidirectional semantic features of the text are further extracted through the BiLSTM model to enhance the model's sensitivity to time series information. On this basis, the KAN model is introduced to replace traditional linear weights with learnable activation functions at the edges, which improves the model's ability to handle data fitting and complex feature representation. Experimental results show that the BBK model has significantly improved precision, recall and F1 score, verifying its effectiveness and superiority in Chinese text sentiment analysis.