Emotion analysis method and device fusing syntactic information and feature interaction

The invention relates to a sentiment analysis method and device fusing syntactic information and feature interaction, and the method comprises the steps: carrying out the coding of an inputted global text and a local text through two channels, and carrying out the construction of a syntactic dependence tree through the text input of the local channel, the obtained matrix can copy one part as one of two parts of feature interaction to interact with the global feature vector, and the other part is extracted through the graph attention neural network, so that syntactic information contained in the text can be well extracted. Besides, a vector obtained by a local channel through a Bert embedding layer and a vector obtained after a local context focusing mechanism serve as the other part of feature interaction to interact with the global features, and then the vectors of the two parts are added, so that the information features of the text can be well reserved, and the accuracy of text polarity prediction is impro