Image Logic and Semantics of User Motion Interaction Language Based on Deep Learning

With the rapid development of science and technology, exploring the image logic semantics of users’ Chinese language using deep learning is crucial for understanding users’ interactions in virtual space. To address the challenges of recognizing existential sentences in Chinese language images, this research introduces a Tree-LSTM encoder, combined with an improved Transformer attention mechanism, to construct a logical semantic recognition model. The model was validated using a dataset from the Modern Dictionary database. Results showed that the proposed model achieved a recognition accuracy of 90.01%, surpassing other models by at least 2.99%. After incorporating Wikipedia data, the model’s performance in handling complex sentence structures and technical terminology was particularly outstanding. When all the extended data sources were integrated, the model’s accuracy reached 92.31%, representing a 2.3% improvement over the original dataset. This success is due to the Tree-LSTM’s ability to capture hierarchical and logical relationships within sentences. The model also effectively learns and recognizes low-frequency new words, enhancing overall performance.