The invention discloses a personalized commodity retrieval method and device based on depth learning attention mechanism. The method comprises the following steps of: constructing a short-term preference model based on the attention mechanism; constructing a long-term preference model based on the attention mechanism; inquiring and re-expressing: wherein the inquiring and re-expressing process comprises the steps of fusing the short-term preference model based on the attention mechanism, the long-term preference model based on the attention mechanism and a current query, learning the interaction relationship among the three through a multi-layer full-connection network to obtain a recombinant query representation, using a distance function to define the degree of correlation between all commodities and the current query; after the multi-layer full-connection network is trained, submitting a new query to each user to obtain the distance values between all commodities and the current query, ranking all the distance values from high to low, and returning the commodities corresponding to the previous n distance values to the user, wherein n is a positive integer.