Automobile network public praise viewpoint label extraction method

The invention provides an automobile network public praise viewpoint label extraction method. The automobile network public praise viewpoint label extraction method comprises the following steps that S1, new words in a public praise text are found through mutual information and left and right entropies, a professional lexicon is further enriched, and the mutual information can calculate the aggregation degree; s2, data preprocessing: mainly adding extraction results of new words and phrases into a dictionary, and carrying out operations such as word segmentation and stop word removal on a text; and S3, constructing a deep learning calculation framework. According to the automobile network public praise viewpoint label extraction method provided by the invention, two automatic viewpoint label extraction schemes based on the expert knowledge base and the deep neural network are realized at the same time, and a Bootstrapping algorithm is adopted to support automatic updating and confidence analysis of the expert