Information recommendation method, computer device, and storage medium

Patent metadata

Field	Value
Publication number	US-11514260-B2
Application number	US-201916592085-A
Country	US
Kind code	B2
Filing date	Oct 3, 2019
Priority date	Sep 8, 2017
Publication date	Nov 29, 2022
Grant date	Nov 29, 2022

Abstract

Information recommendation methods are provided. Image information corresponding to an image is obtained by processing circuitry. The image is associated with a user identifier. A user tag set corresponding to the user identifier and the image information is generated. A feature vector corresponding to user tags in the user tag set and the image information is formed. The feature vector is processed according to a trained information recommendation model, to obtain a recommendation parameter of to-be-recommended information. A recommendation of the to-be-recommended information is provided to a terminal corresponding to the user identifier according to the recommendation parameter.

First claim

What is claimed is:

1. An information recommendation method, comprising:
obtaining, by a processing circuitry, image information corresponding to an image, the image is associated with a user identifier;
generating, by the processing circuitry, a user tag set corresponding to the user identifier and the image information, the user tag set being generated based on a first user tag set and a second user tag set, the second user tag set being generated based on the first user tag set;
forming, by the processing circuitry, a feature vector corresponding to user tags in the user tag set and the image information;
processing, by the processing circuitry, the feature vector according to a trained information recommendation model, to obtain a recommendation parameter of to-be-recommended information; and
providing, by the processing circuitry, a recommendation of the to-be-recommended information to a terminal corresponding to the user identifier according to the recommendation parameter.

2. The method according to claim 1 , wherein
the image information includes image content information and image acquisition information, the image content information including a plurality of images; and
the generating includes
classifying, by the processing circuitry, the images according to the image content information, and determining the first user tag set corresponding to the image content information based on a result of the classification according to the image content information; and
classifying, by the processing circuitry, the images according to the image acquisition information, and determining the second user tag set corresponding to the image acquisition information based on a result of the classification according to the image acquisition information.

3. The method according to claim 1 , wherein the forming comprises:
performing, by the processing circuitry, matching with standard user models according to the user tags in the user tag set and the image information corresponding to the user tags when a scale of the user tag set corresponding to the user identifier is less than a preset scale;
determining, by the processing circuitry, a target standard user model of the standard user models matching the user identifier; and
obtaining, by the processing circuitry, a standard user feature vector corresponding to the target standard user model as the feature vector corresponding to the user identifier.

4. The method according to claim 3 , wherein the performing comprises:
calculating, by the processing circuitry, degrees of matching between a user corresponding to the user identifier and the standard user models according to the user tags in the user tag set and the image information corresponding to the user tags; and
selecting, by the processing circuitry, a standard user model of the standard user models with a highest degree of matching of the calculated degrees of matching as the target standard user model matching the user identifier.

5. The method according to claim 4 , wherein the calculating comprises:
obtaining, by the processing circuitry, image quantities corresponding to the user tags in the user tag set;
determining, by the processing circuitry, current scores corresponding to the user tags according to the image quantities;
for each of the standard user models,
obtaining, by the processing circuitry, standard scores corresponding to standard user tags that are in the respective standard user model and that are the same as the user tags,
calculating, by the processing circuitry, degrees of similarity between the user tags in the user tag set and the standard user tags according to the standard scores and the corresponding current scores, and
obtaining, by the processing circuitry, the degree of matching between the user corresponding to the user identifier and the respective standard user model according to the degrees of similarity.

6. The method according to claim 1 , further comprising:
obtaining, by the processing circuitry, training image information;
generating, by the processing circuitry, a training user tag set according to the training image information;
forming, by the processing circuitry, a training feature vector according to training user tags in the training user tag set and the training image information corresponding to the training user tag set;
obtaining, by the processing circuitry, a standard output result corresponding to the training feature vector; and
performing, by the processing circuitry, model training by using the training feature vector and the standard output result as a training sample, to obtain a target information recommendation model.

7. The method according to claim 1 , wherein the generating comprises:
determining, by the processing circuitry, the first user tag set corresponding to the user identifier and the image information;
generating, by the processing circuitry, the second user tag set based on extracted features of the first user tag set; and
forming, by the processing circuitry, the user tag set corresponding to the user identifier according to the first user tag set and the second user tag set.

8. The method according to claim 1 , wherein
each piece of the to-be-recommended information has a corresponding information recommendation model; and
the processing includes
processing, by the processing circuitry, the feature vector according to the corresponding information recommendation models, to obtain a corresponding recommendation parameter set, each recommendation parameter in the recommendation parameter set being used for determining a recommendation probability of one piece of the to-be-recommended information,
generating, by the processing circuitry, an information recommendation list corresponding to the user identifier according to the recommendation probabilities corresponding to the pieces of the to-be-recommended information, and
determining, by the processing circuitry, target to-be-recommended information corresponding to the user identifier according to the information recommendation list.

9. The method according to claim 1 , wherein the providing comprises:
providing, by the processing circuitry, the to-be-recommended information to the terminal corresponding to the user identifier in a form of a picture when the recommendation parameter is greater than a preset threshold.

10. An information recommendation method, comprising:
obtaining, by processing circuitry, image information, the image information corresponding to an image, the image is associated with a user identifier;
generating, by the processing circuitry, a current user tag set corresponding to the user identifier and the image information, the user tag set being generated based on a first user tag set and a second user tag set, the second user tag set being generated based on the first user tag set;
obtaining, by the processing circuitry, to-be-recommended information;
obtaining, by the processing circuitry, an expected user tag set corresponding to the to-be-recommended information;
calculating, by the processing circuitry, a degree of similarity between the current user tag set and the expected user tag set; and
providing, by the processing circuitry, a recommendation of the to-be-recommended information to a terminal corresponding to the user identifier according to the degree of similarity.

11. The method according to claim 10 , wherein
the generating includes
generating, by the processing circuitry, the first user tag set corresponding to the user identifier and the image information,
generating, by the processing circuitry, the second user tag set based on extracted features of the first user tag set, an

Assignees

• Tencent Tech Shenzhen Co Ltd

Inventors

• Shi Zhiling
• Lin Li

Classifications

• G06N20/10 Primary

  using kernel methods, e.g. support vector machines [SVM] · CPC title

• G06F18/22

  Matching criteria, e.g. proximity measures · CPC title

• G06F18/2113

  by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation · CPC title

• G06F18/2148

  characterised by the process organisation or structure, e.g. boosting cascade · CPC title

• G06Q30/0631

  Recommending goods or services · CPC title