Question and answer reasoning method and system based on cross-modal implicit local learning

The invention provides a question and answer reasoning method and system based on cross-modal implicit local learning, and relates to the technical field of natural language processing. The method comprises the steps that training feature data sets are obtained, the training feature data sets are paired, a plurality of samples are obtained, and each pair of samples is composed of an image, a question description text corresponding to the image and an answer corresponding to the question. Extracting image features of the image, and masking the problem description text to obtain a mask problem description text; performing feature extraction on the mask problem description text by using a Transform encoder to obtain mask problem description text features; and constructing a cross-modal fusion module, taking the image features and the mask problem description text features as inputs of the cross-modal fusion module, and predicting characters replaced by masks by a mask prediction module. And calculating and predi