Multimodal Adversarial Learning Video Recommendation Method and System

Abstract

The invention relates to the fields of computers and artificial intelligence, and provides a multi-modal confrontation learning type video recommendation method and system. The method of the present invention introduces the image information of the recommended item, integrates key technologies such as hierarchical kernel descriptor features, "cross-modal semantics", and adversarial learning into the Bayesian personalized ranking model to construct a MVABRP model, based on the MVABPR model A preferred set of most relevant items is recommended to the user. According to the method or system of the present invention, the recommendation task can be completed based on heterogeneous data (user rating matrix, image), alleviate the "data sparse" problem in the recommendation to a certain extent, and improve the degree of personalization of the recommendation.

Description

technical field [0001] The present invention relates to the fields of computers and artificial intelligence, and more specifically, to a video recommendation method and system. Background technique [0002] In recent years, with the rapid development of Internet technology, Internet application products emerge in an endless stream. They have the advantages of strong interactive ability, convenient operation, and easy dissemination, and can carry rich network information. However, this has also promoted the explosive growth of the amount of information on the network, causing Internet users to be submerged in massive data and unable to extricate themselves, resulting in the problem of "information overload" (also known as "information explosion"). The recommendation system is one of the effective means to deal with the problem of "information overload". [0003] With the advent of the era of artificial intelligence (AI), recommendation systems integrating advanced technologi...

AI Technical Summary

Problems solved by technology

[0007] (1) The degree of personalization of the recommendation is not high, and the recommendation results that are diversified and close to their real preferences are not given according to the actual needs of users;

[0008] (2) Since the number of items browsed or rated by users is very limited, there is a serious "data sparse" problem in the user rating matrix, which seriously affects the recommendation performance;

[0009] (3) Only rely on matrix decomposition to obtain item feature representation, but lack of deep semantic description of recommended items;

[0010] (4) Mainly consider the explicit interaction between the user and the item, while ignoring those key implicit interactions, lacking the analysis of the potential interest of the user;

[0011] (5) The recommendation results are more inclined to long-term users, and the actual recommendation effect for new users is not good, that is, there is a "cold start" problem in the recommendation, and it is impossible to make relevant recommendations to new users

Images