基于BERT-Bi-IAN的方面级情感分析模型

Aspect-based sentiment analysis model based on the BERT-Bi-IAN

为进一步提高情感识别的准确率,提出了一种基于BERT预训练模型的交互式情感分析模型——BERT-Bi-IAN模型。首先利用BERT预训练模型分别动态表示评价对象及其上下文,然后通过Bi-LSTM网络提取二者的语义信息,随后输入交互式注意力模块,得到二者的交互关系与提取到的语义信息重组评价对象及其上下文的最终表示。模型在中英文数据集中的实验表现证明了与现有经典模型相比,其准确率和F1值有所提升,表明该模型有较大的适用领域。

In order to further improve the accuracy of emotion recognition, this paper proposes an interactive sentiment analysis model based on the BERT pre-training model, the BERT-Bi-IAN model. Firstly, the BERT pre-training model is used to dynamically represent the evaluation object and the context, and then the semantic information of the two is extracted through the Bi-LSTM network, and then it is input into the interactive attention module to obtain the interaction between the two, then it is recombined with the extracted semantic information to evaluate the final representation of the object and context. The experimental performance of the model in the Chinese and English datasets proves that the accuracy and F1 value of the model proposed in this paper have improved compared with the existing classic models, reflecting that the model has a large application field.