Deep emotion analysis and multi-source recommendation view fusion hybrid recommendation method based on user comment

Abstract

The invention provides a deep emotion analysis and multi-source recommendation view fusion hybrid recommendation method based on a user comment. The method comprises the steps that S1 a collaborativefiltering recommendation model based on the comprehensive grade of a user is realized by mining the emotional tendency of the user comment; S2 a word vector and a convolutional neural network are usedto calculate the similarity of the content information of an item; and S3 a collaboratively trained policy is used to fuse multi-source recommendation views to construct a hybrid recommendation model. A solution is provided for using short text information such as the user comment to mine the emotional tendency and solving the problem of the authenticity of the grade of the user. A theoretical basis and technical means are provided for understanding the short text description of the content of the item and fusing the recommendation models of multiple views.

Description

technical field [0001] The invention relates to the field of computers, in particular to a hybrid recommendation method based on deep sentiment analysis of user comments and fusion of multi-source recommendation views. Background technique [0002] Today, with the rapid development of e-commerce, social networks, and the sharing economy, discovering user needs, understanding user behavior, and screening out the most relevant information and items for users are a core part of Internet services. Personalized recommendation, as a tool to help users quickly search for useful information, is increasingly favored by people. [0003] Traditional recommendation algorithms use collaborative filtering to recommend items based on user ratings. However, this recommendation method is restricted by the authenticity of user ratings, and the recommendation results based on user ratings cannot accurately reflect user preferences. By comparing the emotional tendencies of user ratings and co...

AI Technical Summary

Problems solved by technology

Traditional natural language processing techniques (such as the bag of words model) represent text as a W-dimensional one-hot vector, but this one-hot representation assumes that all objects are independent of each other and is susceptible to data sparsity problems

[0005] On the other hand, model-based recommendations often lack sufficient labeled data due to the sparsity of ratings relative to items.

Images