Optimized Semantic Recognition and Pattern Analysis in English Text Using Enhanced DeBERTa with Word Embedding
Xin Ma · 2025
This study theoretically proposed an English text semantic recognition and pattern analysis algorithm based on enhanced DeBERTa combined with the word embedding. This algorithm can improve the understanding ability of current human-computer text dialogue systems. First, a word embedding model is built at the algorithm design level to represent text features with high-dimensional vectors. At the same time, data density calculation and PageRank algorithm are used to extract keywords. This operation ensures efficient expression of text features and provides a basis for semantic analysis. Based on this model, the decoding mechanism of DeBERTa is improved. Specifically, by introducing label perception layer optimization and adjusting the association weights between words through the attention mechanism. This operation improves the model's ability to understand the semantics of complex text structures. The experimental part uses a self-developed data set for testing, and conducts comparative tests on the precision rate, recall rate and the F1 indicators. Experimental results show that the proposed model performs best and outperforms the comparison algorithms in all indicators.