A Personalized Text Recommendation Method Based on Deep Learning

Abstract

The present invention relates to a personalized text recommendation method based on deep learning, comprising the following steps: S1: preprocessing historical behavior data and text data of users browsing news; S2: feature extractor modeling, specifically including: S21: hidden layer design ; S22: Output layer design; S3: Personalized recommendation model modeling, including: S31: One-dimensional convolutional network layer design; S32: Classification output layer and loss function design. The present invention effectively solves the problem of operation data sparsity, and enhances the efficiency of model training by using negative sampling technology; introduces browsing time as a global variable, and optimizes the coding effect through the final purpose; by using the coding method of item embedding, and then Effectively solve the problem of cold start of the project; reduce the deep structure, increase the parallel hierarchy, weight sharing in the convolutional layer, and relatively few parameters.

Description

technical field [0001] The invention belongs to the technical field of text recommendation, and relates to a personalized text recommendation method based on deep learning. Background technique [0002] The recommendation system is a connector between people and information, which is used to predict the possible future interaction between users and information content based on user characteristics and past interactions of users. The recommendation system selects the recommendation algorithm according to the historical behavior of different users, the user's interest preference or the user's demographic characteristics, or establishes a recommendation model, uses the recommendation algorithm or model to generate a list of items that the user may be interested in, and finally pushes it to user. [0003] In recent years, with the continuous development of deep learning research, a large number of recommendation algorithm models based on deep learning have been proposed. Recom...

AI Technical Summary

Problems solved by technology

The method described in this patent introduces a global variable to improve the skip-gram method to encode user operation data, and then fuses the convolutional neural network for recommendation, which has not been reported yet.

Images