abstract |
The present invention proposes a sentiment classification method using graph convolutional neural network and Chinese syntax, including the following steps: S1, acquiring social network text, taking the acquired social network text as the text to be processed, and preprocessing the text to be processed; S2 , learn the context information of sentences and attribute clauses, and obtain the corresponding feature representation; S3, generate the semantic tree of the sentence according to the obtained dependencies and grammar information of the given text; S4, according to the dependency tree embedded with the feature vector, use A graph convolutional network generates sentiment feature representations for a given text; S5, utilizes the Softmax classifier to construct a conditional probability distribution for each sentiment label, and outputs the final sentiment label of the text. The present invention can classify the sentiment of the acquired social network text, and provide more detailed and in-depth sentiment analysis for the short text in the social network platform. |