Evaluation method for model generation text and computer equipment

The invention discloses a model generation text evaluation method and computer equipment, which do not depend on labels and are suitable for a production environment. According to the evaluation method, evaluation is carried out through three indexes of genes, readability and fingerprints, and then comprehensive evaluation is carried out. Wherein the gene index is used for measuring semantic correlation and homology of the model generation text and the input text, and the readability index is used for measuring the degree that the model generation text can be read by people according to the length mean value of sentences segmented from punctuations and the text repetition degradation condition. The fingerprint index is used for measuring the semantic feature distribution consistency degree of the model generation text and the training set label; specifically, a fingerprint extraction network model is trained in advance based on a ternary twin network, a fingerprint database is generated at the same time, then