Media resource processing method and device, storage medium and electronic device

Abstract

The invention discloses a media resource processing method and device, a storage medium and an electronic device. The method comprises the following steps: acquiring a first label, a first title and afirst picture of a first media resource, and a second label, a second title and a second picture of a second media resource to obtain a first text feature of the first media resource and a first picture feature of the first media resource; obtaining a second text feature of the second media resource, and obtaining a second picture feature of the second media resource; determining the relevancy between the first media resource and the second media resource according to the first text feature, the first picture feature, the second text feature and the second picture feature. The purpose of determining the relevancy between the media resources according to the labels and titles of the media resources and the content of the media resources is achieved, and then the technical problem that in the prior art, the accuracy of pushing the multimedia resources is low is solved.

Description

technical field [0001] The present invention relates to the field of multimedia, in particular to a method and device for processing media resources, a storage medium and an electronic device. Background technique [0002] In the recommendation system scenario of related news, the commonly used algorithm is mainly the text feature extraction method of news tags or titles. The methods in the prior art are all by extracting the text features of news tags or titles, and then by cosine distance or Euclidean The distance calculates the correlation between news, and finally recommends related news. [0003] The related news recommendation algorithm based on the correlation model that generates word vectors cannot distinguish the different semantics of polysemous words. At the same time, the feature calculation method of vector addition often ignores the position information between different words in the label or title, which will be to a certain extent result in loss of informat...

AI Technical Summary

Problems solved by technology

[0003]The related news recommendation algorithm based on the related model of generating word vectors cannot distinguish the different semantics of polysemous words, and the feature calculation method of vector addition often ignores tags or Position information between different words in the title, which will cause information loss to a certain extent

The relevant news recommendation algorithm based on Chinese text, due to its unique convolution kernel pooling operation, makes it difficult to capture semantic information such as negation and antisense in the text sequence

[0004] For the above problems, no effective solution has been proposed

Images