基于语义增强的多特征融合方面级情感分析OA

Currently,most sentiment analysis models often rely on semantic structure of syntactic dependency trees to extract se-mantic information.However,there exists a gap between syntactic dependency structures and semantic sentiment analysis tasks.To address this issue,this paper proposes a semantic-enhanced multi-feature fusion method for aspect-level sentiment analysis.This method introduces the Abstract Meaning Representation(AMR)structure and combines the extraction methods of global and local features for aspect-level sentiment analysis tasks.Firstly,the relation embedding representation extracted from AMR and the sentence embedding representation extracted from BERT are fused to obtain the semantic information of the input text.Secondly,Bi-LSTM and capsule networks are employed to extract deep global and local features.Finally,a multi-head self-attention mechanism is applied to integrate multi-dimensional features,effectively capturing the associations between aspect terms and contextual sentences.Experiments on multiple public datasets demonstrate the effectiveness of the proposed method.On the Restaurant dataset,the model achieves an accuracy of 87.77% and a recall of 82.60%,while on the Twitter dataset,it at-tains an accuracy of 78.71% and a recall of 77.54%.The experimental results indicate that the proposed method significantly im-proves the performance of aspect-level sentiment analysis.