Literary works, as the carriers of social thoughts of the times to which they belong, are important references for social and cultural research. This paper focuses on ancient Chinese literature and uses corpus as a research tool to explore the social phenomena and cultural connotations reflected in literary works. Based on the different textual characteristics and connections between ancient and modern texts, we analyze the methods of selecting paragraph alignment features of the two texts. We also analyze the alignment process of ancient and modern texts, so as to complete the construction of the Ancient Literature Annotated Corpus. Then the technical framework of K-BERTopic model is designed to analyze and extract the themes of the texts in the corpus. It also combines the improved EDR with the Kmeans++-based dimensionality optimizer to dynamically reduce the dimensionality of the input documents and enhance the model effect. Subsequently, ancient poems are selected as the research object to analyze the social and cultural connotations of ancient Chinese literary works, extract the central keywords of ancient poems, statistically analyze the frequency of intonation auxiliaries, and collate the information of temporal changes of ancient poems. The K-BERTopic model is used to analyze the first data in the Tang Dynasty corpus, in which the probability of the theme of “landscapes and fields” is 0.2142 and the probability of the theme of “wandering to the ends of the earth” is 0.5768, which is in line with the corresponding thematic content of the poems. It shows that the K-BERTopic model can realize the thematic analysis of literary works by aligning ancient literary works with modern textual features, and assist in analyzing the social and cultural connotations.
