Information retrieval method and device

The invention relates to an information retrieval method and device, and the method comprises the steps: obtaining all modal data, and obtaining a multi-modal data set; performing feature extraction on data in the multi-modal data set according to data types, fusing obtained feature vectors, constructing a recurrent neural network model, and training the recurrent neural network model through the feature vectors; performing multi-modal semantic analysis on the multi-modal data through a recurrent neural network model; semantic representations of different modal data are mapped to entities and relations of the knowledge graph, cross-modal semantic association is achieved, and semantic reasoning and association are conducted on cross-modal data; and performing graph construction and query on the entities and the relationships by using a graph database or a knowledge graph storage engine. According to the method, data of different modes can be synthesized to form integrated knowledge representation, so that the