Dynamic dictionary library generation method based on feedback self-learning

The invention discloses a dynamic dictionary library generation method based on feedback self-learning. The method comprises the following steps: S1, defining a dictionary library; S2, classifying a system based on a dictionary database; S4, gradually iterating and optimizing the model by matching with model parameter adjustment on the basis of annotation data exported from the corpus; and S5, backfilling the prediction data generated in the step S4 into the dictionary database. The method has the beneficial effects that the basic dictionary library is generated from original corpus concept pattern definition through concept pattern definition, a natural language processing technology and a full-process scheduling mechanism, automatic construction, iterative training and correction of the labeling model are carried out on the basis of the basic dictionary library, and finally, new corpus data labeling is carried out by utilizing the labeling model, the dictionary database is fed back and updated, a closed-loo