Text object accurate pushing method for mining deep features based on neural collaborative filtering

Abstract

The invention discloses a text object accurate pushing method for mining deep features based on neural collaborative filtering. The method comprises a policy description document feature extraction method based on a convolutional neural network and a hybrid pushing method for mining deep features. The invention provides specific steps of a policy description document feature extraction method based on a convolutional neural network. The specific steps of the hybrid pushing method for mining the deep features are provided. Compared with an existing policy pushing method, the method has the advantages that the convolutional neural network can be used for automatically extracting the local features of different word ranges of the semantic level contained in the text from the policy description document; and meanwhile, the extracted features are fused into the neural collaborative filtering policy pushing method in a more flexible manner, a nonlinear interaction relationship between the user and the policy is established in the neural collaborative filtering policy pushing method, and deeper interaction features between the user and the policy are mined, so that higher pushing accuracycan be achieved.

Description

technical field [0001] The invention relates to the field of push technology, in particular to a method for accurately pushing text objects based on neural collaborative filtering to mine deep features. Background technique [0002] In recent years, with the rapid development of mobile communication networks, users can obtain text objects in electronic forms such as policies and news more and more easily. Quickly and targetedly obtain the content of the text object, and a message push method about the text object appears, so that a solution can be proposed for the user to quickly make a satisfactory choice. Therefore, it is very important how to construct a push method that can more accurately and sensitively capture the user's preferences and needs for searching for desired text content (such as policies). [0003] Existing methods for pushing messages about text objects such as policy and news mainly have the following deficiencies: [0004] 1) There are limited ways to ...

AI Technical Summary

Problems solved by technology

[0005] 2) The linear modeling method between the user and the text object to be pushed cannot capture the complex structure in between, thus affecting the accuracy of policy push

[0006] 3) Most of the hybrid push methods cannot effectively and automatically capture the contextual features in the text objects to be pushed

Images