Method and apparatus of establishing similarity model for retrieving geographic location

Abstract

The present disclosure provides a method of retrieving a geographic location as well as a method and an apparatus of establishing a similarity model including: acquiring training data from a historical click log, wherein the training data contains a search term, a clicked geographic location associated with the search term as a positive sample, and a non-clicked geographic location as a negative sample; expanding the positive sample and the negative sample based on a co-occurrence relationship between geographic locations by using a historical browsing log; and determining a vector representation of the search term by using the first neural network, and determining a vector representation of the positive sample expanded and a vector representation of the negative sample expanded, by using the second neural network; and training the first neural network and the second neural network so as to maximum a difference between a first similarity and a second similarity.