CN115687745A

CN115687745A - Multimedia data recommendation method and device, storage medium and computer equipment

Info

Publication number: CN115687745A
Application number: CN202110864794.6A
Authority: CN
Inventors: 朱家卫
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2021-07-29
Filing date: 2021-07-29
Publication date: 2023-02-03

Abstract

The embodiment of the invention discloses a multimedia data recommendation method, a device, a storage medium and computer equipment, wherein the method comprises the steps of determining first multimedia data which are interesting to a user and correspond to identity data of each user on the basis of operation records by acquiring user information; acquiring second multimedia data matched with each first multimedia data in a multimedia database, and generating target multimedia data corresponding to each user identity data according to the first multimedia data and the second multimedia data; training a preset recommendation model by using training sample data to obtain a trained recommendation model; and recommending the multimedia data to the target user by adopting the trained recommendation model. Therefore, the positive sample of the recommendation model is automatically obtained from the multimedia database by the machine learning technology and matched with the multimedia data interested by the user to expand, so that the recommendation model is trained more fully, and the accuracy of multimedia data recommendation is improved.

Description

Multimedia data recommendation method and device, storage medium and computer equipment

Technical Field

The invention relates to the technical field of artificial intelligence, in particular to a multimedia data recommendation method, a multimedia data recommendation device, a storage medium and computer equipment.

Background

In recent years, with the development of internet technology, information is rapidly increasing, and how to rapidly and effectively screen information, so that personalized content (such as commodities, advertisements, news information, APP, and the like) suitable for a user is accurately recommended to the user is an important research topic at present.

The recommendation algorithm can perform machine learning according to the existing user information, and further recommend personalized contents which may be interested to the user. The multi-e-commerce platforms adopt a recommendation system based on a propulsion algorithm to analyze browsing and purchasing behaviors of users and try to carry out personalized recommendation on the users who browse or purchase commodities once, so that the sales of the e-commerce platforms are effectively increased, and the recommendation system is more and more widely used.

However, when a recommendation system is adopted on a multimedia data providing platform to recommend multimedia content, the amount of active clicks of a large number of users on the multimedia content is small, so that the sample size of a recommendation model during training is insufficient, and the training effect of the recommendation model is poor.

Disclosure of Invention

The embodiment of the application provides a multimedia data recommendation method, a multimedia data recommendation device, a storage medium and computer equipment.

A first aspect of the present application provides a multimedia data recommendation method, including:

acquiring user information, wherein the user information comprises a plurality of user identity data and an operation record of multimedia data corresponding to each user identity data;

determining first multimedia data which are interesting to the user and correspond to each user identity data based on the operation records;

acquiring second multimedia data matched with each first multimedia data in a multimedia database, and generating target multimedia data corresponding to each user identity data according to the first multimedia data and the second multimedia data;

training a preset recommendation model by using training sample data to obtain a trained recommendation model, wherein the training sample data comprises attribute data corresponding to each user identity data and target multimedia data corresponding to each user identity data;

and recommending the multimedia data to the target user by adopting the trained recommendation model.

Accordingly, a second aspect of the present application provides a multimedia data recommendation apparatus, comprising:

the device comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring user information which comprises a plurality of user identity data and operation records of multimedia data corresponding to each user identity data;

the determining unit is used for determining first multimedia data which are corresponding to each user identity data and are interesting to the user based on the operation records;

the generating unit is used for acquiring second multimedia data matched with each first multimedia data in a multimedia database and generating target multimedia data corresponding to each user identity data according to the first multimedia data and the second multimedia data;

the training unit is used for training a preset recommendation model by adopting training sample data to obtain the trained recommendation model, wherein the training sample data comprises attribute data corresponding to each user identity data and target multimedia data corresponding to each user identity data;

and the recommending unit is used for recommending the multimedia data to the target user by adopting the trained recommending model.

In some embodiments, the generating unit includes:

the calculating subunit is used for calculating the similarity between each multimedia data in the multimedia database and each first multimedia data;

and the determining subunit is used for determining second multimedia data matched with each first multimedia data based on the similarity.

In some embodiments, the computation subunit includes:

the first acquisition module is used for acquiring a first vector corresponding to each multimedia data in the multimedia database;

the second acquisition module is used for acquiring a second vector corresponding to each first multimedia data;

the first calculating module is used for calculating the cosine similarity of each first vector and each second vector to obtain the similarity of each multimedia data and each first multimedia data.

In some embodiments, the first obtaining module includes:

the generating submodule is used for generating a first matrix according to the operation record, and the first matrix indicates the score corresponding to each user identity data and each multimedia data;

the first decomposition module is used for decomposing the first matrix into a second matrix corresponding to the user identity data and a third matrix corresponding to the multimedia data;

and the second decomposition module is used for decomposing according to the third matrix to obtain a first vector corresponding to each multimedia data.

In some embodiments, the determining subunit includes:

the sequencing module is used for sequencing the multimedia data in the multimedia database from high similarity to low similarity between the multimedia data and each first multimedia data to obtain a sequencing sequence corresponding to each first multimedia data;

and the determining module is used for determining the first preset number of multimedia data in the sequencing sequence corresponding to each first multimedia data as the second multimedia data matched with each multimedia data.

In some embodiments, the determining module comprises:

the generating submodule is used for generating a similar multimedia data set corresponding to each multimedia data according to the front preset number of multimedia data in the sequencing sequence corresponding to each first multimedia data;

and the determining submodule is used for determining that the multimedia data, of which the similarity with the corresponding first multimedia data is greater than a preset threshold value, in the similar multimedia data set corresponding to each multimedia data is the second multimedia data matched with each first multimedia data.

In some embodiments, the apparatus further comprises:

the third acquisition module is used for acquiring the playing time length of each first multimedia data;

and the second calculating module is used for calculating the preset number of the second multimedia data matched with each first multimedia data according to the playing time length.

In some embodiments, the determining unit includes:

the cleaning subunit is used for cleaning the operation records corresponding to the identity data of each user based on the playing time length of each multimedia data;

and the extraction subunit is used for extracting the multimedia data actively clicked by the user from the operation records after the data cleaning to obtain the first multimedia data which are corresponding to the identity data of each user and are interesting to the user.

In some embodiments, the recommendation unit includes:

the acquiring subunit is used for acquiring the target user identity data of the target user and determining target attribute data based on the target user identity data;

the input subunit is used for inputting the target attribute data into the trained recommendation model to obtain output target multimedia data;

and the recommending subunit is used for recommending the target multimedia data to the target user.

The third aspect of the present application further provides a computer-readable storage medium, which stores a plurality of instructions adapted to be loaded by a processor to perform the steps of the multimedia data recommendation method provided in the first aspect of the present application.

A fourth aspect of the present application provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the multimedia data recommendation method provided in the first aspect of the present application when executing the computer program.

A fifth aspect of the present application provides a computer program product or computer program comprising computer instructions stored in a storage medium. The processor of the computer device reads the computer instructions from the storage medium, and the processor executes the computer instructions to make the computer device execute the steps of the multimedia data recommendation method provided by the first aspect.

According to the multimedia data recommendation method provided by the embodiment of the application, user information is obtained, and the user information comprises a plurality of user identity data and operation records of multimedia data corresponding to each user identity data; determining first multimedia data which are interesting to the user and correspond to each user identity data based on the operation records; acquiring second multimedia data matched with each first multimedia data in a multimedia database, and generating target multimedia data corresponding to each user identity data according to the first multimedia data and the second multimedia data; training a preset recommendation model by using training sample data to obtain a trained recommendation model, wherein the training sample data comprises attribute data corresponding to each user identity data and target multimedia data corresponding to each user identity data; and recommending the multimedia data to the target user by adopting the trained recommendation model. Therefore, the positive sample of the recommendation model is expanded by acquiring the multimedia data which are matched with the multimedia data interested by the user from the multimedia data set, so that the recommendation model is trained more fully, the recommendation effect of the recommendation model deployed in an industrial scene is improved, and the accuracy of multimedia data recommendation is improved.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the description below are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained based on these drawings without creative efforts.

FIG. 1 is a schematic diagram of a scenario of multimedia data recommendation in the present application;

FIG. 2 is a flow chart of a multimedia data recommendation method provided herein;

FIG. 3 is a graph showing a distribution of user average daily active behavior statistics for an information providing application;

FIG. 4 is another schematic flow chart of a multimedia data recommendation method provided in the present application;

FIG. 5 is a schematic flow chart of a multimedia data recommendation method provided in the present application;

FIG. 6 is a schematic structural diagram of a multimedia data recommendation device provided in the present application;

fig. 7 is a schematic structural diagram of a terminal provided in the present application;

fig. 8 is a schematic structural diagram of a server provided in the present application.

Detailed Description

The technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. It should be apparent that the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The embodiment of the invention provides a multimedia data recommendation method and device, a computer readable storage medium and computer equipment. The multimedia data recommendation method can be used in a multimedia data recommendation device. The multimedia data recommendation device can be integrated in computer equipment, and the computer equipment can be a terminal or a server. The terminal can be a mobile phone, a tablet Computer, a notebook Computer, an intelligent television, a wearable intelligent device, a Personal Computer (PC), and the like. The server may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing basic cloud computing services such as cloud service, a cloud database, cloud computing, a cloud function, cloud storage, network service, cloud communication, middleware service, domain name service, security service, network acceleration service (CDN), big data, an artificial intelligence platform, and the like.

Please refer to fig. 1, which is a schematic view of a scene recommended by multimedia data provided in the present application; as shown in the figure, when the computer device a obtains the user information, the user information includes a plurality of user identification data and an operation record of multimedia data corresponding to each user identification data. Determining first multimedia data which are interesting to the user and correspond to each user identity data based on the operation records; and acquiring second multimedia data matched with each first multimedia data from a multimedia database, and generating target multimedia data corresponding to each user identity data according to the first multimedia data and the second multimedia data. Training a preset recommendation model by using training sample data to obtain a trained recommendation model, wherein the training sample data comprises attribute data corresponding to each user identity data and target multimedia data corresponding to each user identity data; and recommending the multimedia data to the target user logging in the computer equipment A by adopting the trained recommendation model.

It should be noted that the scene diagram of multimedia data recommendation shown in fig. 1 is only an example, and the multimedia data recommendation scene described in the embodiment of the present application is for more clearly illustrating the technical solution of the present application, and does not constitute a limitation on the technical solution provided by the present application. As can be appreciated by those skilled in the art, with the evolution of multimedia data recommendations and the emergence of new service scenarios, the technical solution provided in the present application is also applicable to similar technical problems.

Based on the above-described implementation scenarios, detailed descriptions will be given below.

Embodiments of the present application will be described from the perspective of a multimedia data recommender, which may be integrated in a computer device. The Computer device may be a terminal, and the terminal may be a mobile phone, a tablet Computer, a notebook Computer, a smart television, a wearable smart device, a Personal Computer (PC), or the like. As shown in fig. 2, a schematic flow chart of a multimedia data recommendation method provided in the present application is shown, where the method includes:

step 101, obtaining user information.

Here, the user information may include user information of a plurality of users. Specifically, when the multimedia data recommendation method provided by the present application is used for performing multimedia data recommendation for a user in a specific application, the user information may include information of all users registered in a server corresponding to the application. The user information comprises user identity data, attribute data corresponding to the user identity data and operation records of multimedia data corresponding to the user identity data. The user Identity information may be a user Identity Document (ID) generated when the user registers in the application program. Attribute data corresponding to the user identity data may include, but is not limited to, the user's age, city of the user, check-in location information, academic history, and interests and hobbies of the user. The operation record of the multimedia data corresponding to the user identity data may include a historical operation record of the user corresponding to the user ID on the multimedia data in the application program. The historical operation records include but are not limited to operations of clicking, watching, browsing, like, evaluating, collecting and the like on the multimedia data; for viewing, browsing, etc., the history operation record may further include time length data of viewing, browsing, etc. The multimedia data may include plain text data, graphics data, video data, audio data, and the like.

In recent years, the accuracy of product recommendation is greatly improved by using a recommendation system based on machine learning. Machine Learning (ML) is a multi-domain cross discipline, and relates to a plurality of disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and the like. The method specially studies how a computer simulates or realizes the learning behavior of human beings so as to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve the performance of the computer. Machine learning is the core of artificial intelligence, is the fundamental approach for computers to have intelligence, and is applied to all fields of artificial intelligence. Machine learning and deep learning generally include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and teaching learning.

When a recommendation system is used for recommending multimedia contents to a user, the inventor of the application finds that the sample data in the sample data for training the recommendation system has larger sparsity, so that the recommendation system cannot be trained sufficiently, and the recommendation effect of the model when the model is deployed in a service scene for recommendation is further influenced. Specifically, as shown in fig. 3, it is a distribution diagram of daily active behavior statistics of a user who provides an application program for a certain information by the inventor of the present application. As shown, more than 60% of users have only 1 to 5 active actions per day. In general, multimedia data of which a user has an active behavior is often multimedia data which is more interesting to the user and is also positive sample data more suitable for training a recommendation model. And the sparse positive sample data can lead to insufficient model training, thereby influencing the recommendation effect of the model after being deployed in an industrial scene.

In order to solve the technical problem that due to insufficient user initiative behavior, positive sample data of a recommendation model is sparse, and the deployment effect of the recommendation model is further influenced, the application provides a multimedia data recommendation method, and the method is described in detail below.

First, training data still needs to be acquired. Specifically, the user information may be obtained as described above, that is, the user identity data of all users registered in the server of the application, the attribute data corresponding to the user identity data, and the operation record of the user corresponding to the user identity data on the multimedia data provided by the application. Then, further processing is performed based on the user information.

And 102, determining the first multimedia data which are interesting to the user and correspond to each user identity data based on the operation records.

After the user information is acquired, determining the multimedia data which is corresponding to the user identity data and is interested by the user according to the operation record of the multimedia data corresponding to each user identity data contained in the user information. In the embodiment of the application, the multimedia data of which the user has an active behavior can be determined to be the multimedia data in which the user is interested; the active behavior may include click, like, favorite, comment, and other behaviors. The multimedia data of which the watching time length reaches the preset time length or the ratio of the watching time length to the total time length of the multimedia data reaches the preset ratio can also be determined as the multimedia data which is interested by the user. Alternatively, both of the above-described multimedia data may be determined as multimedia data of interest to the user.

In some embodiments, determining the first multimedia data of interest to the user corresponding to each user identity data based on the operation record comprises:

1. performing data cleaning on the operation record corresponding to each user identity data;

2. and extracting the multimedia data actively clicked by the user from the operation records after the data cleaning to obtain the first multimedia data which are corresponding to the identity data of each user and are interesting to the user.

In the embodiment of the present application, some user behaviors may be misoperation behaviors. Or, although a user clicks a certain multimedia data, the user finds that the multimedia data is not interesting after clicking, so that the browsing time of the multimedia data is very short, for example, less than 5s (seconds) or the browsing time accounts for less than 5% of the total duration of the multimedia data, and the multimedia data is not interesting to the user even if the user actively clicks. Therefore, in order to avoid the influence of such noise samples on the model training effect, data cleaning can be performed on the operation record data corresponding to each user identity data according to a preset cleaning rule. Specifically, all the viewing records of the multimedia data corresponding to the operation record of the multimedia data corresponding to any user identity data can be acquired, and the multimedia data of which the viewing duration is less than the preset duration and the ratio of the viewing duration to the total duration of the multimedia data is less than the preset ratio is removed according to the viewing duration to obtain the cleaned multimedia data.

Then, multimedia data actively clicked by the user are further extracted from the cleaned operation records, and the multimedia data are confirmed as first multimedia data interested by the user. In the embodiment of the application, a user actively clicks the multimedia data to be watched, and if the watching duration reaches the preset duration or the ratio of the duration to the total playing duration of the multimedia data reaches the preset ratio, it can be determined that the user has an active selection tendency and has certain stickiness, so that the multimedia data can be determined as the multimedia data which is interested by the user. Further, the method may be adopted to determine the first multimedia data, which is interested by the user, corresponding to each user identity data one by one.

It can be understood that the types of the multimedia data that the user is interested in may be multiple types, and therefore, there may be only one or multiple first multimedia data corresponding to each user identity data.

Step 103, obtaining second multimedia data matched with each first multimedia data from the multimedia database, and generating target multimedia data corresponding to each user identity data according to the first multimedia data and the second multimedia data.

In the embodiment of the present application, because as described above, insufficient number of positive samples corresponding to each user identity data may result in insufficient training of the recommendation model, the multimedia data recommendation method provided in the embodiment of the present application may enhance the positive sample data based on the current positive sample data of the user. Specifically, in the present application, the current positive sample data of the user is the first multimedia data corresponding to the user identity data, and the positive sample enhancement performed based on the first multimedia data may be that a certain amount of second multimedia data matched with the first multimedia data is acquired from a multimedia database. Wherein the second multimedia data similar to the first multimedia data may be multimedia data similar to the first multimedia data type or similar in content. Then, the second multimedia data is supplemented into the first multimedia data, and a set formed by the first multimedia data and the second multimedia data is used as a positive sample for model training. The set of first multimedia data and second multimedia data may be referred to herein as target multimedia data.

The target multimedia data is the multimedia data which is enhanced or expanded based on the first multimedia data. Therefore, the extension and enrichment of the positive sample data for training the recommendation model are realized, the training of the recommendation model can be more sufficient, and the trained recommendation model can be deployed in an industrial scene to achieve a better recommendation effect.

In some embodiments, obtaining second multimedia data in the multimedia database that matches each first multimedia data comprises:

1. calculating the similarity between each multimedia data in the multimedia database and each first multimedia data;

2. second multimedia data matching each first multimedia data is determined based on the similarity.

In the embodiment of the present application, the second multimedia data matching each first multimedia data is determined from the multimedia database, and the similarity between each first multimedia data and each multimedia data in the multimedia database may be calculated first. Specifically, the detailed description may be given taking any one of the target first multimedia data as an example. I.e. for example, there are 3 first multimedia data of interest to the user a, which are a, b and c, respectively, and for any target multimedia data a, the similarity between the multimedia data and each multimedia data in the multimedia database can be calculated. For example, the multimedia database has 1000 multimedia data, and the similarity between the multimedia data a and the 1000 multimedia data in the multimedia database can be calculated to obtain 1000 similarity values. Similarly, for the first multimedia data b and c, the similarity between the first multimedia data b and c and each multimedia data in the multimedia database can be calculated, and 1000 similarity values can be obtained. For the first multimedia data corresponding to other users, the similarity between each first multimedia data and each multimedia data in the multimedia database can be calculated one by the method.

After the similarity between each first multimedia data and each multimedia data in the multimedia database is obtained through calculation, the multimedia data with the similarity meeting the preset condition is determined to be second multimedia data matched with each first multimedia data according to the similarity.

The similarity between any target multimedia data and the target first multimedia data is calculated, the titles of the target multimedia data and the target first multimedia data can be respectively obtained, and then the text similarity between the titles of the target multimedia data and the target first multimedia data is calculated to obtain the similarity between the target multimedia data and the target first multimedia data.

In some embodiments, tags may be further added to the multimedia data, each multimedia data has several tag data for describing its content type, content main event, main character in the content, and the like, and the tag data of the target multimedia data and the target first multimedia data may be obtained first, and then the similarity between the target multimedia data and the target first multimedia data may be calculated according to the tag similarity of the target multimedia data and the target first multimedia data.

In some embodiments, calculating the similarity between each multimedia data in the multimedia database and each first multimedia data comprises:

1.1, acquiring a first vector corresponding to each multimedia data in a multimedia database;

1.2, acquiring a second vector corresponding to each first multimedia data;

and 1.3, calculating the cosine similarity of each first vector and each second vector to obtain the similarity of each multimedia data and each first multimedia data.

In the embodiment of the application, the similarity between each multimedia data in the multimedia database and each first multimedia data is calculated, and each multimedia data in the multimedia database can be mapped into a vector space according to a certain rule to obtain a first vector corresponding to each multimedia data in the multimedia database; then, similarly, each first multimedia data is mapped to a vector space according to a certain rule, and a second vector corresponding to each first multimedia data is obtained. In this way, a first vector corresponding to each multimedia data in the multimedia data and a second vector corresponding to each first multimedia data can be obtained. Then, for any target first multimedia data, the corresponding target second vector may be obtained. And then calculating the cosine similarity of each first vector and the target second vector one by one to obtain the similarity of each first vector and the target second vector, and further obtain the similarity of each multimedia data and each first multimedia data.

In some embodiments, obtaining a first vector corresponding to each multimedia data in the multimedia database comprises:

1.1.1, generating a first matrix according to the operation record, wherein the first matrix indicates the score corresponding to each user identity data and each multimedia data;

1.1.2, decomposing the first matrix into a second matrix corresponding to the user identity data and a third matrix corresponding to the multimedia data;

and 1.1.3, decomposing according to the third matrix to obtain a first vector corresponding to each multimedia data.

In the embodiment of the application, the score of each user for each multimedia data in the multimedia database can be generated according to the operation record of the multimedia data corresponding to each user identity data, and the score is used for evaluating the interest degree of the user for the multimedia data. Specifically, the score may be set to 0 to 10 points, with 0 points indicating no interest at all and 10 points indicating great interest. A scoring matrix may then be generated from each user's score on each multimedia data, or may be referred to as a first matrix. Specifically, for example, there are m users in total, and there are n pieces of multimedia data in the multimedia database in total, then an m × n matrix can be generated according to the rating of each user to each piece of multimedia data. Then, the first matrix can be split by adopting an alternating least square method to obtain a second matrix corresponding to the user identity data and a third matrix corresponding to the multimedia data. In particular, the first matrix may be decomposed into a second matrix of m x k and a third matrix of k x n, where k < m and k < n. Then, a k-dimensional vector corresponding to each multimedia data is further determined according to the third matrix of k × n.

In some embodiments, determining second multimedia data that matches each first multimedia data based on the similarity includes:

2.1, sequencing the multimedia data in the multimedia database according to the sequence of similarity between the multimedia data and each first multimedia data from high to low to obtain a sequencing sequence corresponding to each first multimedia data;

and 2.2, determining the first preset number of multimedia data in the sequencing sequence corresponding to each first multimedia data as second multimedia data matched with each multimedia data.

After the similarity between each multimedia data in the multimedia database and each first multimedia data is obtained through calculation, for any target first multimedia data, each multimedia data in the multimedia database and the similarity thereof can be obtained. Then, the multimedia data in the multimedia database can be further sorted based on the high-low order of similarity between each multimedia data and the multimedia data. Specifically, the first multimedia data may be sorted in an order from a high similarity to a low similarity, or may be sorted in an order from a low similarity to a high similarity, so as to obtain a sorting sequence corresponding to the target first multimedia data. In this application, the similarity between each multimedia data and the target first multimedia data is sorted from high to low. Wherein it is understood that each ordered sequence contains all multimedia data in the multimedia database. Then, further, an ordered sequence corresponding to each first multimedia data can be obtained.

After the sorting sequence corresponding to each first multimedia data is obtained, a preset number of multimedia data sorted in front (with high similarity) in each sorting sequence can be determined as second multimedia data matched with the corresponding first multimedia data. For example, the preset number may be set to 5, and the sorting sequence corresponding to the target first multimedia data is the first sequence, so that the first 5 multimedia data in the first sequence may be determined as the second multimedia data matching the target first multimedia data.

In some embodiments, determining that the first predetermined number of multimedia data in the sorted sequence corresponding to each first multimedia data is the second multimedia data matching each first multimedia data comprises:

2.2.1, generating a similar multimedia data set corresponding to each multimedia data according to the front preset number of multimedia data in the sequencing sequence corresponding to each first multimedia data;

and 2.2.2, determining that the multimedia data with the similarity larger than a preset threshold value with the corresponding first multimedia data in the similar multimedia data set corresponding to each multimedia data is the second multimedia data matched with each first multimedia data.

In this embodiment of the application, after the sorting sequence corresponding to each first multimedia data is obtained, a similar multimedia data set corresponding to each first multimedia data may be generated according to a preset number of multimedia data sorted in the top in each sorting sequence. Then, the similarity value between the multimedia data in the similar multimedia data set corresponding to each first multimedia data and the corresponding first multimedia data is determined one by one, and the multimedia data with the similarity value higher than the preset threshold value with the corresponding first multimedia data is determined as the second multimedia data matched with the first multimedia data. Specifically, a similar multimedia data set corresponding to each first multimedia data may be generated according to the first 5 multimedia data in the sorted sequence corresponding to each first multimedia data, and then multimedia data with a similarity greater than 90% to the corresponding first multimedia data in the similar multimedia data set corresponding to each first multimedia data may be further determined as second multimedia data matching each first multimedia data.

In this way, after the predetermined number of multimedia data with higher similarity to the first multimedia data is determined, the similarity between the predetermined number of multimedia data and the first multimedia data is further compared with the predetermined similarity threshold, and only the multimedia data with the similarity higher than the similarity threshold to the first multimedia data is determined as the second multimedia data matching the first multimedia data. Therefore, the second multimedia data matched with the first multimedia data can be screened in the aspects of similarity sequencing and a similarity threshold, and the condition that the training process of the recommendation model is interfered due to the fact that noise samples are expanded is avoided.

In some embodiments, determining that the first predetermined number of multimedia data in the sorted sequence corresponding to each first multimedia data is before the second multimedia data matching each first multimedia data further includes:

A. acquiring the playing time length of each first multimedia data;

B. and calculating the preset number of the second multimedia data matched with each first multimedia data according to the playing time length.

In this embodiment, the second amount of the second multimedia data matching with the first multimedia data may be determined according to the playing duration of each first multimedia data. Specifically, the playing time length of the first multimedia data in the operation record of the first multimedia data that is interested by the user may be obtained first. And then, calculating the preset number of second multimedia data matched with the first multimedia data according to the playing time length of the first multimedia data. Therefore, the longer the playing time of the first multimedia data is, the more the second multimedia data matched with the first multimedia data is, the more positive samples corresponding to the first multimedia data can be obtained, and thus, the samples for training the recommendation model can include more positive samples corresponding to the first multimedia data. For the first multimedia data which is interested by the user, the playing time length of the first multimedia data corresponds to the browsing time length of the user, and also corresponds to the interest degree of the user. If the longer the playing time is, the higher the interest level of the user is, in the present application, more positive sample data is expanded for the user to enhance the ratio of the user in the positive sample data, thereby improving the recommendation possibility. Therefore, the accuracy of the recommendation system for the multimedia data can be further improved.

And 104, training a preset recommendation model by using the training sample data to obtain the trained recommendation model.

After second multimedia data matched with first multimedia data interested by the user is determined, and target multimedia data corresponding to each user identity data is generated based on the first multimedia data and the corresponding second multimedia data. The attribute data corresponding to each user identity data and the target multimedia data corresponding to each user identity data may be used as training sample data to train a preset recommendation Model, where the preset recommendation Model may be a double tower recall Model, that is, a Deep Structured Semantic Model (DSSM).

Specifically, attribute data corresponding to the user identity data, for example, data such as the age, the city to which the user belongs, check-in location information, a scholarship, and interests and hobbies of the user, may be used as input of the model, and the recommendation model may be trained using target multimedia data corresponding to the user identity data, that is, the expanded positive sample data, as the tag data of the training sample, to obtain the trained recommendation model.

And 105, recommending the multimedia data to the target user by adopting the trained recommendation model.

After the recommendation model is trained to obtain the trained recommendation model, the recommendation model can be deployed in an application scene. Then, in use, the recommendation model may obtain user identity data corresponding to the user client and make multimedia data recommendations for it based on the user identity data.

In some embodiments, the performing multimedia data recommendation on the target user by using the trained recommendation model includes:

1. acquiring target user identity data of a target user, and determining target attribute data based on the target user identity data;

2. inputting the target attribute data into the trained recommendation model to obtain output target multimedia data;

3. and recommending the target multimedia data to the target user.

In the embodiment of the application, after the target user needing to perform multimedia data recommendation is determined, the target tree data can be determined based on the target user identity data obtained from the target user. After the target attribute data of the user is determined, the target attribute data is input to the trained recommendation model to obtain output target multimedia data, and it can be understood that one or more pieces of target multimedia data may be provided. And then pushing the target multimedia data to a client corresponding to the target user for display.

According to the foregoing description, in the multimedia data recommendation method provided in the embodiment of the present application, user information is obtained, where the user information includes a plurality of user identity data and an operation record of multimedia data corresponding to each user identity data; determining first multimedia data which are interesting to the user and correspond to each user identity data based on the operation records; acquiring second multimedia data matched with each first multimedia data in a multimedia database, and generating target multimedia data corresponding to each user identity data according to the first multimedia data and the second multimedia data; training a preset recommendation model by using training sample data to obtain a trained recommendation model, wherein the training sample data comprises attribute data corresponding to each user identity data and target multimedia data corresponding to each user identity data; and recommending the multimedia data to the target user by adopting the trained recommendation model. Therefore, the positive sample of the recommendation model is expanded by acquiring the multimedia data which are matched with the multimedia data interested by the user from the multimedia data set, so that the recommendation model is trained more sufficiently, the recommendation effect of the recommendation model deployed in an industrial scene is improved, and the accuracy of multimedia data recommendation is improved.

Accordingly, the embodiment of the present application will further describe in detail the multimedia data recommendation method provided by the present application from the perspective of a computer device, where the computer device may be a terminal or a server. As shown in fig. 4, another schematic flow chart of a multimedia data recommendation method provided in the present application is shown, where the method includes:

step 201, the computer device obtains user information of all users in the target application program.

The target application program provides an application program for multimedia data needing to be recommended by the multimedia data recommendation method provided by the application. Specifically, the application may be an information providing application, such as Tencent News; it may also be a video playing application, such as Tencent video; it may also be a short video application, such as tremble; it may also be an audio playing application such as himalayan. The multimedia data described in the present application may include any one or more of the above-mentioned multimedia data such as news information, long video, short video, audio data, and the like.

The user information of all users in the target application program may be account information of all users registered in a server corresponding to the target application program, user attribute information associated with the account information, a viewing record, a clicking record, a commenting record, a collecting record, a forwarding record, and the like of multimedia data provided by the target application program in the target application program by a user corresponding to the account information. Specifically, the viewing record for the multimedia data further includes viewing duration data for the multimedia data.

Before the computer recommends multimedia data for the user of the target application program, the computer may first obtain the user information of all users of the target application program, and accordingly train the recommendation model of the recommendation system, so as to perform accurate multimedia data recommendation by using the trained recommendation model.

Step 202, the computer device performs data cleaning on the user information, and determines first multimedia data which are interesting to each user according to the user information after the data cleaning.

After the user information of each user is obtained, in order to ensure that the recommendation model is effectively trained, noise samples in the user information need to be cleaned first. Specifically, the number of browsing records of each user for the multimedia data may be obtained first, and the user data whose browsing records are less than a certain number may be removed. For example, if the total number of multimedia data browsed by some users since registration is less than 10, or the number of multimedia browsing records in a recent preset time period is less than 10, it may be determined that the randomness of multimedia data browsing by the users is relatively high, and the multimedia data browsing by the users is used as a sample for training a recommendation model, so that the model training is biased, and thus, the user data is eliminated. And moreover, the multimedia data browsing records of which the browsing time length in each user browsing record is less than the preset time length or the browsing time length accounts for the total multimedia data time length can be cleaned. For example, if the browsing time of a user for a certain multimedia data is less than 15s, or the playing time of the multimedia data occupies less than 55% of the total time, it may be determined that the multimedia data is not interesting to the user, and in order to avoid that the data interferes with the training of the recommendation model, the data needs to be removed.

Then, in the cleaned user information, the multimedia data which is interested by the user can be determined according to the click condition of the user on the multimedia data. For example, multimedia data actively clicked and viewed by the user may be determined as first multimedia data of interest to the user, and multimedia data having approval, collection, and forwarding behaviors of the user may be determined as first multimedia data of interest to the user. The first multimedia data is also positive sample data that can be used for training the recommendation model.

In step 203, the computer device calculates the similarity between each multimedia data in the multimedia database and the first multimedia data.

In the embodiment of the application, after the first multimedia data which is interesting to the user is determined, the similarity between each multimedia data in the multimedia database of the target application program and the first multimedia data is further calculated.

Specifically, in the embodiment of the present application, an Alternating Least Squares (ALS) method may be used to calculate the similarity between each multimedia data in the multimedia database and the first multimedia data.

Specifically, the similarity between each multimedia data in the multimedia database and the first multimedia data is calculated by using an alternating least square method, and the vector representation of each multimedia data in the multimedia database can be determined first, and then the cosine similarity between the vector of each multimedia data and the vector corresponding to the first multimedia data is calculated.

In the case where the user has a score for each multimedia data, a scoring matrix corresponding to the multimedia data can be generated. Based on the assumption of ALS that the scoring matrix is low rank, the scoring matrix can be approximated by decomposing it into the product of the two matrices. For example, if the number of users is m and the number of multimedia data in the multimedia database is n, then a scoring matrix of m × n may be obtained. The scoring matrix may be decomposed to obtain an m x k user matrix and a k x n multimedia data matrix, and the scoring matrix may be approximated as a product of the two matrices. Wherein k is less than m and k is less than n. A vector corresponding to each multimedia data may be determined based on the matrix of multimedia data.

However, in more cases, the user's scoring of the multimedia data is unclear. In this case, then, an implicit feedback model needs to be established to optimize the computation of the implicit vector corresponding to each multimedia data, and a loss function can be defined for the optimization:

wherein, X and Y are respectively the user matrix corresponding to the minimum loss function and the matrix corresponding to the multimedia data. u is the number of users, i is the number of multimedia data, and γ is a hyper-parameter. p is a radical of _ui Indicates whether the user u is interested in the item i, and

x _u representing a user vector, y _i Representing a vector of multimedia data. c. C _ui Represents confidence, and:

c _ui ＝1+αr _ui

wherein, alpha is a hyperparameter, r _ui The following expression is satisfied:

then, the loss function is solved by an alternate least square method to obtain a matrix Y corresponding to the final multimedia data, and the matrix is decomposed to obtain a vector Y corresponding to each multimedia data _i 。

After the vector corresponding to each multimedia data is obtained, the similarity between each multimedia data and the first multimedia data can be calculated according to the cosine similarity between the vectors.

In step 204, the computer device calculates the amount of the second multimedia data matching the first multimedia data, and determines the second multimedia data matching the first multimedia data based thereon.

In the embodiment of the present application, data enhancement is performed on first multimedia data that is interested in a user, or when a positive sample is expanded, the number of expansions may be determined according to a play duration of the first multimedia data. Specifically, it can be determined according to the following formula:

second multimedia data amount = min (play duration/14, β)

The unit of the playing time length is second, and beta is a hyper-parameter, so that the number of the second multimedia data is not more than beta.

After the target number of the second multimedia data is determined, the target number of multimedia data having a higher similarity to the first multimedia data may be determined as the second multimedia data matching the first multimedia data according to the number.

In step 205, the computer device generates a training sample according to the first multimedia data and the second multimedia data.

After the second multimedia data matched with the first multimedia data is determined, the positive sample data corresponding to the user can be determined according to the set formed by the first multimedia data and the second multimedia data. Then, the computer device may obtain the positive sample data corresponding to each user, where it can be understood that the positive sample data at this time is the extended positive sample data.

After positive sample data corresponding to each user is obtained, determining a training sample according to the user attribute information of each user and the positive sample data corresponding to the user. The user attribute information is input during model training, and positive sample data corresponding to a user is a label output by a training model.

And step 206, training the double-tower regression model by the computer equipment by adopting the training sample to obtain the trained double-tower regression model.

After the training sample for training the recommended model is determined, the training sample may be used to train a preset two-tower regression model. The training process can refer to the related art, and is not described in detail herein. After training, a trained two-tower regression model can be obtained.

And step 207, the computer equipment carries out multimedia data recommendation on the target user by adopting the trained double-tower regression model.

After the trained double-tower regression model is obtained, the trained double-tower regression model is deployed in an application scene. When multimedia data is recommended, the computer equipment firstly acquires account information of a user logged in an application program client, and acquires target attribute information of the user according to the account information. And then inputting the target attribute information into the trained double-tower regression model to obtain target multimedia data output by the model. The computer device displays the target multimedia data in the display area of the application program client.

Fig. 5 is a schematic flowchart of a multimedia data recommendation method provided in the present application. As shown in the figure, after the user information 10 is acquired, data cleaning is performed on the user information 10 to obtain cleaned data 20, then a vector of each multimedia data is obtained by adopting an ALS matrix decomposition method based on the cleaned data 20, and the similarity between each multimedia data is further calculated to obtain similarity data 30, where the similarity data 30 includes a similarity value between each multimedia data and any other multimedia data. Then, the original positive sample 40 of the user is further subjected to data enhancement according to the similarity data 30, or referred to as expanding the original positive sample 40, so as to obtain an enhanced sample 50. And finally, training the recommendation model by using the enhancement sample 50 to obtain a trained recommendation model 60. After the recommendation model obtained by training by the method is deployed in an industrial scene, the index of Area Under the Receiver Operating Characteristic Curve (ROC) Curve (AUC) of the recommendation model is obviously increased Under an off-line condition, and the time for which the user uses the application program by all people in the on-line scene is also obviously increased. The Click Through Rate (CTR) of video has also increased significantly.

In order to better implement the method, the embodiment of the invention also provides a multimedia data recommendation device, and the multimedia data recommendation device can be integrated in a terminal.

For example, as shown in fig. 6, for a schematic structural diagram of a multimedia data recommendation device provided in an embodiment of the present application, the multimedia data recommendation device may include an obtaining unit 301, a determining unit 302, a generating unit 303, a training unit 304, and a recommending unit 305, as follows:

an obtaining unit 301, configured to obtain user information, where the user information includes a plurality of user identity data and an operation record of multimedia data corresponding to each user identity data;

a determining unit 302, configured to determine, based on the operation record, first multimedia data that is of interest to the user and corresponds to each user identity data;

a generating unit 303, configured to obtain second multimedia data matched with each first multimedia data in a multimedia database, and generate target multimedia data corresponding to each user identity data according to the first multimedia data and the second multimedia data;

a training unit 304, configured to train a preset recommendation model by using training sample data to obtain a trained recommendation model, where the training sample data includes attribute data corresponding to each user identity data and target multimedia data corresponding to each user identity data;

and the recommending unit 305 is configured to recommend multimedia data to the target user by using the trained recommendation model.

In some embodiments, a generation unit comprises:

and the determining subunit is used for determining the second multimedia data matched with each first multimedia data based on the similarity.

In some embodiments, a compute subunit includes:

In some embodiments, the first obtaining module comprises:

In some embodiments, determining the subunit includes:

In some embodiments, the determining module comprises:

In some embodiments, the multimedia data recommendation apparatus provided by the present application further includes:

and the second calculating module is used for calculating the preset quantity of the second multimedia data matched with each first multimedia data according to the playing time length.

In some embodiments, the determining unit comprises:

In some embodiments, a recommendation unit comprises:

the acquisition subunit is used for acquiring target user identity data of a target user and determining target attribute data based on the target user identity data;

In specific implementation, the above units may be implemented as independent entities, or may be combined arbitrarily, and implemented as the same or several entities, and specific implementations of the above units may refer to the foregoing method embodiment, which is not described herein again.

As can be seen from the above description, in the multimedia data recommendation method provided in this embodiment of the present application, the obtaining unit 301 obtains user information, where the user information includes a plurality of user identity data and an operation record of multimedia data corresponding to each user identity data; the determining unit 302 determines, based on the operation record, first multimedia data that is of interest to the user and corresponds to each user identity data; the generating unit 303 obtains second multimedia data matched with each first multimedia data from a multimedia database, and generates target multimedia data corresponding to each user identity data according to the first multimedia data and the second multimedia data; the training unit 304 trains a preset recommendation model by using training sample data to obtain a trained recommendation model, wherein the training sample data comprises attribute data corresponding to each user identity data and target multimedia data corresponding to each user identity data; the recommending unit 305 recommends multimedia data for the target user by using the trained recommendation model. Therefore, the positive sample of the recommendation model is expanded by acquiring the multimedia data which are matched with the multimedia data interested by the user from the multimedia data set, so that the recommendation model is trained more fully, the recommendation effect of the recommendation model deployed in an industrial scene is improved, and the accuracy of multimedia data recommendation is improved.

An embodiment of the present application also provides a computer device, which may be a terminal, as shown in fig. 7, where the terminal may include a Radio Frequency (RF) circuit 401, a memory 402 including one or more computer-readable storage media, an input unit 403, a display unit 404, a sensor 405, an audio circuit 406, a Wireless Fidelity (WiFi) module 407, a processor 408 including one or more processing cores, and a power supply 409. Those skilled in the art will appreciate that the terminal structure shown in fig. 7 is not intended to be limiting and may include more or fewer components than those shown, or some components may be combined, or a different arrangement of components. Wherein:

the RF circuit 401 may be used for receiving and transmitting signals during a message transmission or communication process, and in particular, for receiving downlink information of a base station and then sending the received downlink information to the one or more processors 408 for processing; in addition, data relating to uplink is transmitted to the base station. In general, RF circuitry 401 includes, but is not limited to, an antenna, at least one Amplifier, a tuner, one or more oscillators, a Subscriber Identity Module (SIM) card, a transceiver, a coupler, a Low Noise Amplifier (LNA), a duplexer, and the like. In addition, RF circuitry 401 may also communicate with networks and other devices via wireless communications. The wireless communication may use any communication standard or protocol, including but not limited to Global System for Mobile communications (GSM), general Packet Radio Service (GPRS), code Division Multiple Access (CDMA), wideband Code Division Multiple Access (WCDMA), long Term Evolution (LTE), email, short Message Service (SMS), and the like.

The memory 402 may be used to store software programs and modules, and the processor 408 executes various functional applications and information interactions by operating the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data (such as audio data, a phonebook, etc.) created according to the use of the terminal, etc. Further, the memory 402 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 408 and the input unit 403 access to the memory 402.

The input unit 403 may be used to receive input numeric or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control. In particular, in a particular embodiment, the input unit 403 may include a touch-sensitive surface as well as other input devices. The touch-sensitive surface, also referred to as a touch display screen or a touch pad, may collect touch operations by a user (e.g., operations by a user on or near the touch-sensitive surface using a finger, a stylus, or any other suitable object or attachment) thereon or nearby, and drive the corresponding connection device according to a predetermined program. Alternatively, the touch sensitive surface may comprise two parts, a touch detection means and a touch controller. The touch detection device detects the touch direction of a user, detects a signal brought by touch operation and transmits the signal to the touch controller; the touch controller receives touch information from the touch sensing device, converts it to touch point coordinates, and sends the touch point coordinates to the processor 408, and can receive and execute commands from the processor 408. In addition, the touch sensitive surface can be implemented in various types, such as resistive, capacitive, infrared, and surface acoustic wave. The input unit 403 may include other input devices in addition to the touch-sensitive surface. In particular, other input devices may include, but are not limited to, one or more of a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, a joystick, and the like.

The display unit 404 may be used to display information input by or provided to the user and various graphical user interfaces of the terminal, which may be made up of graphics, text, icons, video, and any combination thereof. The Display unit 404 may include a Display panel, and optionally, the Display panel may be configured in the form of a Liquid Crystal Display (LCD), an Organic Light-Emitting Diode (OLED), or the like. Further, the touch-sensitive surface may overlay the display panel, and when a touch operation is detected on or near the touch-sensitive surface, the touch operation is transmitted to the processor 408 to determine the type of touch event, and then the processor 408 provides a corresponding visual output on the display panel according to the type of touch event. Although in FIG. 7 the touch sensitive surface and the display panel are implemented as two separate components for input and output functions, in some embodiments the touch sensitive surface may be integrated with the display panel for input and output functions.

The terminal may also include at least one sensor 405, such as a light sensor, motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor that may adjust the brightness of the display panel according to the brightness of ambient light, and a proximity sensor that may turn off the display panel and/or the backlight when the terminal is moved to the ear. As one of the motion sensors, the gravity acceleration sensor can detect the magnitude of acceleration in each direction (generally, three axes), can detect the magnitude and direction of gravity when the mobile phone is stationary, and can be used for applications of recognizing the posture of the mobile phone (such as horizontal and vertical screen switching, related games, magnetometer posture calibration), vibration recognition related functions (such as pedometer and tapping), and the like; as for other sensors such as a gyroscope, a barometer, a hygrometer, a thermometer, and an infrared sensor, which can be configured in the terminal, detailed description is omitted here.

Audio circuitry 406, a speaker, and a microphone may provide an audio interface between the user and the terminal. The audio circuit 406 may transmit the electrical signal converted from the received audio data to a speaker, and convert the electrical signal into a sound signal for output; on the other hand, the microphone converts the collected sound signal into an electric signal, which is received by the audio circuit 406 and converted into audio data, which is then processed by the audio data output processor 408, and then sent to another terminal, for example, via the RF circuit 401, or the audio data is output to the memory 402 for further processing. The audio circuitry 406 may also include an earbud jack to provide peripheral headset communication with the terminal.

WiFi belongs to a short-distance wireless transmission technology, and the terminal can help a user to receive and send emails, browse webpages, access streaming media and the like through the WiFi module 407, and provides wireless broadband internet access for the user. Although fig. 7 shows the WiFi module 407, it is understood that it does not belong to the essential constitution of the terminal, and may be omitted entirely as needed within the scope not changing the essence of the invention.

The processor 408 is a control center of the terminal, connects various parts of the entire handset using various interfaces and lines, and performs various functions of the terminal and processes data by operating or executing software programs and/or modules stored in the memory 402 and calling data stored in the memory 402, thereby integrally monitoring the handset. Alternatively, processor 408 may include one or more processing cores; preferably, the processor 408 may integrate an application processor, which handles primarily the operating system, user interface, applications, etc., and a modem processor, which handles primarily the wireless communications. It will be appreciated that the modem processor described above may not be integrated into processor 408.

The terminal also includes a power source 409 (e.g., a battery) for powering the various components, which may preferably be logically coupled to the processor 408 via a power management system that may be used to manage charging, discharging, and power consumption. The power source 409 may also include any component such as one or more dc or ac power sources, recharging systems, power failure detection circuitry, power converters or inverters, power status indicators, and the like.

Although not shown, the terminal may further include a camera, a bluetooth module, and the like, which will not be described herein. Specifically, in this embodiment, the processor 408 in the terminal loads the executable file corresponding to the process of one or more application programs into the memory 402 according to the following instructions, and the processor 408 runs the application programs stored in the memory 402, thereby implementing various functions:

acquiring user information, wherein the user information comprises a plurality of user identity data and an operation record of multimedia data corresponding to each user identity data; determining first multimedia data which are interesting to the user and correspond to each user identity data based on the operation records; acquiring second multimedia data matched with each first multimedia data in a multimedia database, and generating target multimedia data corresponding to each user identity data according to the first multimedia data and the second multimedia data; training a preset recommendation model by using training sample data to obtain a trained recommendation model, wherein the training sample data comprises attribute data corresponding to each user identity data and target multimedia data corresponding to each user identity data; and recommending the multimedia data to the target user by adopting the trained recommendation model.

It should be noted that the computer device provided in the embodiment of the present application and the method in the foregoing embodiment belong to the same concept, and specific implementation of the above operations may refer to the foregoing embodiment, which is not described herein again.

An embodiment of the present application further provides a computer device, where the computer device may be a server, and as shown in fig. 8, is a schematic structural diagram of the computer device provided in the present application. Specifically, the method comprises the following steps:

the computer device may include components such as a processing unit 501 of one or more processing cores, a storage unit 502 of one or more storage media, a power module 503, and an input module 504. Those skilled in the art will appreciate that the computer device configuration illustrated in FIG. 8 does not constitute a limitation of computer devices, and may include more or fewer components than those illustrated, or some components may be combined, or a different arrangement of components. Wherein:

the processing unit 501 is a control center of the computer device, connects various parts of the whole computer device by using various interfaces and lines, executes various functions of the computer device and processes data by running or executing software programs and/or modules stored in the storage unit 502 and calling data stored in the storage unit 502, thereby performing overall monitoring of the computer device. Optionally, the processing unit 501 may include one or more processing cores; preferably, the processing unit 501 may integrate an application processor and a modem processor, wherein the application processor mainly handles operating systems, user interfaces, application programs, and the like, and the modem processor mainly handles wireless communications. It is to be understood that the above-described modem processor may not be integrated into the processing unit 501.

The storage unit 502 may be used to store software programs and modules, and the processing unit 501 executes various functional applications and data processing by running the software programs and modules stored in the storage unit 502. The storage unit 502 may mainly include a storage program area and a storage data area, where the storage program area may store an operating system, an application program required by at least one function (such as a sound playing function, an image playing function, a web access, and the like), and the like; the storage data area may store data created according to use of the computer device, and the like. Further, the storage unit 502 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory unit 502 may also include a memory controller to provide the processing unit 501 access to the memory unit 502.

The computer device further comprises a power module 503 for supplying power to each component, and preferably, the power module 503 may be logically connected to the processing unit 501 through a power management system, so as to implement functions of managing charging, discharging, power consumption management, and the like through the power management system. The power module 503 may also include any component of one or more dc or ac power sources, recharging systems, power failure detection circuitry, power converters or inverters, power status indicators, and the like.

The computer device may also include an input module 504, the input module 504 operable to receive input numeric or character information and to generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

Although not shown, the computer device may further include a display unit and the like, which are not described in detail herein. Specifically, in this embodiment, the processing unit 501 in the computer device loads the executable file corresponding to the process of one or more application programs into the storage unit 502 according to the following instructions, and the processing unit 501 runs the application programs stored in the storage unit 502, so as to implement various functions as follows:

acquiring user information, wherein the user information comprises a plurality of user identity data and an operation record of multimedia data corresponding to each user identity data; determining first multimedia data which are interesting to the user and correspond to each user identity data based on the operation records; second multimedia data matched with each first multimedia data are obtained from a multimedia database, and target multimedia data corresponding to each user identity data are generated according to the first multimedia data and the second multimedia data; training a preset recommendation model by using training sample data to obtain a trained recommendation model, wherein the training sample data comprises attribute data corresponding to each user identity data and target multimedia data corresponding to each user identity data; and recommending the multimedia data to the target user by adopting the trained recommendation model.

It will be understood by those skilled in the art that all or part of the steps of the methods of the above embodiments may be performed by instructions, or by instructions controlling associated hardware, which may be stored in a computer-readable storage medium and loaded and executed by a processor.

To this end, embodiments of the present invention provide a computer-readable storage medium having stored therein a plurality of instructions, which can be loaded by a processor to perform the steps of any of the methods provided by the embodiments of the present invention. For example, the instructions may perform the steps of:

The above operations can be implemented in the foregoing embodiments, and are not described in detail herein.

Wherein the computer-readable storage medium may include: read Only Memory (ROM), random Access Memory (RAM), magnetic or optical disks, and the like.

Since the instructions stored in the computer-readable storage medium can execute the steps in any method provided by the embodiment of the present invention, the beneficial effects that can be achieved by any method provided by the embodiment of the present invention can be achieved, for details, see the foregoing embodiments, and are not described herein again.

According to an aspect of the application, there is provided, among other things, a computer program product or computer program comprising computer instructions stored in a storage medium. The computer instructions are read from the storage medium by a processor of the computer device, and the processor executes the computer instructions to cause the computer device to perform the method provided in the various alternative implementations of fig. 2 or fig. 4 described above.

The foregoing describes in detail a multimedia data recommendation method, apparatus, computer-readable storage medium, and computer device provided in the embodiments of the present invention, and specific examples are applied herein to explain the principles and embodiments of the present invention, and the descriptions of the foregoing embodiments are only used to help understand the method and core ideas of the present invention; meanwhile, for those skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.

Claims

1. A method for recommending multimedia data, the method comprising:

2. The method of claim 1, wherein obtaining second multimedia data in the multimedia database matching each first multimedia data comprises:

calculating the similarity between each multimedia data in the multimedia database and each first multimedia data;

second multimedia data matching each first multimedia data is determined based on the similarity.

3. The method of claim 2, wherein calculating the similarity between each multimedia data in the multimedia database and each first multimedia data comprises:

acquiring a first vector corresponding to each multimedia data in a multimedia database;

acquiring a second vector corresponding to each first multimedia data;

and calculating the cosine similarity of each first vector and each second vector to obtain the similarity of each multimedia data and each first multimedia data.

4. The method of claim 3, wherein the obtaining the first vector corresponding to each multimedia data in the multimedia database comprises:

generating a first matrix according to the operation record, wherein the first matrix indicates the score corresponding to each user identity data and each multimedia data;

decomposing the first matrix into a second matrix corresponding to the user identity data and a third matrix corresponding to the multimedia data;

and decomposing according to the third matrix to obtain a first vector corresponding to each multimedia data.

5. The method of claim 2, wherein the determining second multimedia data matching each first multimedia data based on the similarity comprises:

sequencing the multimedia data in the multimedia database according to the sequence of similarity between the multimedia data and each first multimedia data from high to low to obtain a sequencing sequence corresponding to each first multimedia data;

and determining the first preset number of multimedia data in the sequencing sequence corresponding to each first multimedia data as second multimedia data matched with each multimedia data.

6. The method of claim 5, wherein the determining that the first predetermined number of multimedia data in the ordered sequence corresponding to each first multimedia data is the second multimedia data matching each first multimedia data comprises:

generating a similar multimedia data set corresponding to each multimedia data according to a preset number of multimedia data in the sequencing sequence corresponding to each first multimedia data;

and determining the multimedia data with the similarity greater than a preset threshold value with the corresponding first multimedia data in the similar multimedia data set corresponding to each multimedia data as second multimedia data matched with each first multimedia data.

7. The method of claim 5, wherein the determining that the first predetermined number of multimedia data in the ordered sequence corresponding to each first multimedia data is before the second multimedia data matching with each first multimedia data, further comprises:

acquiring the playing time length of each first multimedia data;

and calculating the preset number of the second multimedia data matched with each first multimedia data according to the playing time length.

8. The method of claim 1, wherein the determining first multimedia data of interest to the user corresponding to each user identity data based on the operation record comprises:

performing data cleaning on the operation record corresponding to each user identity data based on the playing duration of each multimedia data;

and extracting the multimedia data actively clicked by the user from the operation records after the data cleaning to obtain the first multimedia data which are corresponding to the identity data of each user and are interesting to the user.

9. The method of claim 1, wherein the performing multimedia data recommendation on the target user by using the trained recommendation model comprises:

acquiring target user identity data of a target user, and determining target attribute data based on the target user identity data;

inputting the target attribute data into the trained recommendation model to obtain output target multimedia data;

and recommending the target multimedia data to the target user.

10. An apparatus for recommending multimedia data, said apparatus comprising:

11. The apparatus of claim 10, wherein the generating unit comprises:

12. The apparatus of claim 11, wherein the computing subunit comprises:

and the calculating module is used for calculating the cosine similarity of each first vector and each second vector to obtain the similarity of each multimedia data and each first multimedia data.

13. A computer-readable storage medium storing instructions adapted to be loaded by a processor to perform the steps of the method for recommending multimedia data according to any of claims 1 to 9.

14. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the multimedia data recommendation method of any of claims 1 to 9 when executing the computer program.

15. A computer program, characterized in that it comprises computer instructions stored in a storage medium, which computer instructions are read from the storage medium by a processor of a computer device, the processor executing the computer instructions causing the computer device to perform the steps of the multimedia data recommendation method of any of claims 1 to 9.