Text Sentiment Classification Model based on Fusion of DualChannel Features of CNN and BiLSTM
Meng Zeng, Zhonglin Zhang · 2023
Text sentiment analysis is an important task in natural language processing (NLP), which aims to determine people's emotional tendency towards a certain topic or event by analyzing the language and emotion in the text. Aiming at the traditional emotion classification model can't fully capture the semantic information implied in short text comments, a two-channel emotion classification model based on CNN and BiLSTMl is proposed.Dynamic allocation weights introduced since the attention mechanism, build fusion BiLSTM and CNN's dual channel neural network architecture, and extract the bureau of emotional characteristics and emotional characteristics as global pay attention to the input feature fusion layer, through your emotions full text feature fusion module integration characteristic information and emotional polarity to break..Compared to the experimental results show that the model of emotion classification performance of the optimum Transformer model, this model (CNN-BiLSTM-AFF) on a public data set senti_weibo_100k accuracy, F1 value, the recall rate of 1.034%, 1.265% and 1.045% respectively.