A new method of location detection with knowledge transfer learning

Abstract

This paper propose a new method of location detection with knowledge transfer learning(KTL-LD). Processing Framework was developed and compared with existed experimental results. In this Framework, the label extraction model is used to extract the Weibo user tags, using the existing word embedding model as the source domain, and applying the knowledge transfer learning model to transfer the knowledge to the target domain, which obtain the target domain vector, then the classification model is used to train and predict the user’s location. This method does not require manual tagging data, the training time of this model is effectively reduced, and the ability of generalization is improved as well.