TextRank Chinese abstract generation method based on multiple features of text

The invention relates to the technical field of computers, in particular to a text multi-feature-based TextRank Chinese abstract generation method, which comprises the following steps of: extracting word vectors from a text through a Word2vec model, fusing the word vectors into sentence vectors, calculating the similarity between sentences, constructing a TextRank graph model by taking the similarity as a weight of an edge and taking each sentence as a node, and calculating the TextRank value of each sentence, the method comprises the following steps: firstly, extracting sentiment word features, keyword features, turning word features and punctuation features from a text, calculating multi-feature values of sentences in the text, combining the multi-feature values with TextRank values of the sentences, finally obtaining weights of the sentences, sorting the sentences, and extracting the sentences sorted in the front as candidate abstract sentences. On the basis of filtering the semantic features of the text,