Research on the application of deep learning-based sentiment analysis models to classical Chinese literary texts

Classical Chinese literary texts, such as poetry and prose, are rich in emotional and cultural expression but pose significant challenges for computational analysis due to their archaic language, symbolic structures, and lack of punctuation. Traditional sentiment analysis models often fail to capture the depth and nuance of these texts, limiting their effectiveness in literary interpretation and digital humanities research. This research aims to develop a deep learning (DL)-based RoCoSenti-Classical framework tailored to classical Chinese literature, capable of accurately identifying and classifying sentiments despite the language’s complexity and historical nature. A comprehensive corpus of classical Chinese literary texts was gathered from reputable digital sources. The text data underwent pre-processing steps, including text normalization, tokenization, and stopword removal. Domain-specific language representations were created by fine-tuning RoBERTa embeddings on the classical corpus. These embeddings were then input into a ConvLSTM model, which combines convolutional layers for local feature extraction with LSTM layers for progressive sentiment modeling. Implemented in Python, the findings show that the RoCoSenti-Classical framework performs better than multimodal baseline architectures, achieving superior results with accuracy, F1-score, recall, and precision ranging from 90% to 97%. The suggested RoCoSenti-Classical framework effectively captures complex emotional cues in classical texts, demonstrating robustness across different literary genres and significantly improving sentiment analysis performance for classical Chinese literature. These findings support the broader integration of Artificial Intelligence (AI) in digital humanities and open new avenues for the computational interpretation of ancient texts.