Session recommendation method based on self-adaptive user interest dual correction

The invention discloses a session recommendation method based on self-adaptive user interest dual correction, and belongs to the technical field of session recommendation. The method mainly comprises the following steps: initializing high-dimensional spatial semantics and position representation of each article in a user click sequence, splicing to obtain an article representation matrix, and generating an article representation matrix containing time sequence and context information; the current user interest representation is initialized by using a first self-adaptive correction mechanism, and then the user interest representation of the next moment is output by using a user interest representation distillation module containing a second self-adaptive correction mechanism; and predicting the probability that all the articles in the article dictionary are clicked at the next moment. According to the method, a self-adaptive correction mechanism is adopted, the problem that the current interest of the user can