Text sentiment analysis method

The invention relates to a text sentiment analysis method. The method comprises the following steps: S1, converting words in text sentences into word vectors; s2, mapping the word vector into an emotion feature vector; s3, splicing the word vectors and the sentiment feature vectors into sentiment feature word vectors of each word, wherein a plurality of sentiment feature word vectors in the whole sentence are used as an input feature matrix; s4, inputting the input feature matrix into a convolutional neural network for training to obtain a first abstract emotion feature; s5, inputting the input feature matrix into a bidirectional long-short-term memory network for training to obtain a second abstract emotion feature; and S6, fusing and splicing the first abstract emotion features and the second abstract emotion features, inputting the fused and spliced features into a full-connection neural network, and finally outputting emotion classification of sentences through a Softmax layer. Compared with the prior art