Image-text content pushing method and image-text pushing model training method and device

Abstract

The invention provides an image-text content pushing method and an image-text pushing model training method and device, and relates to the technical field of artificial intelligence such as deep learning, and the method comprises the steps: obtaining an image-text type set corresponding to a user side, and a target pushing model trained under each image-text type in the image-text type set; obtaining an initial data set, and obtaining a candidate data set under each image-text type from the initial data set based on each target push model; obtaining a first image-text queue under each image-text type from the candidate data set to obtain a second image-text queue corresponding to the image-text type set; and obtaining a target push image-text content from the second image-text queue, and pushing the target push image-text content to the user side.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing, and in particular to the field of artificial intelligence technologies such as deep learning. Background Art

[0002] With the development of society, the use of text and image formats as a carrier for information dissemination is becoming more and more common in people's daily work and life. In this scenario, the degree of adaptation between the pushed text and image content and user needs is very important for the experience.

[0003] In the related technology, the click information of the user side can be obtained to push adapted content to the user side. In an immersive scenario, the user may not click on the graphic content, which may lead to poor adaptation between the pushed content and the user's personal preferences. Summary of the invention

[0004] The present invention discloses a method for pushing image and text content, and a method and device for training an image and text push model.

[0005] According to the first aspect of the present disclosure, a method for pushing graphic content is proposed, including: obtaining a graphic type set corresponding to a user terminal, and a target push model trained under each graphic type in the graphic type set; obtaining an initial data set, and obtaining a candidate data set under each graphic type from the initial data set based on each target push model; obtaining a first graphic queue under each graphic type from the candidate data set to obtain a second graphic queue corresponding to the graphic type set; obtaining target push graphic content from the second graphic queue, and pushing the target push graphic content to the user terminal.

[0006] According to the second aspect of the present disclosure, a training method for a picture and text push model is proposed, comprising: obtaining candidate push models to be trained under each sample picture and text type, and obtaining sample picture and text content under each sample picture and text type for a sample user terminal; collecting sample feedback information on the sample picture and text content under each sample picture and text type, and sample user preferences of the sample user terminal for each sample picture and text type; for any sample picture and text type, generating training samples of candidate push models corresponding to the sample picture and text type according to the sample picture and text content under the sample picture and text type, sample feedback information corresponding to the sample picture and text content, and sample user preferences corresponding to the sample picture and text type; for any sample picture and text type, training the candidate push model of the sample picture and text type according to the training samples of the sample picture and text type until the training is completed, thereby obtaining a trained target push model under the sample picture and text type, wherein the target push model is used to implement the picture and text content push method proposed in the first aspect of the above-mentioned claim.

[0007] According to a third aspect of the present disclosure, a system for pushing text and image content is proposed, comprising: a text and image content data layer, a graph aggregation layer, a preference distribution adjustment layer, a fusion layer and a distribution layer, wherein the text and image content data layer is used to store an initial data set of a server, and obtain a candidate data set under each text and image type from the initial data set through a pre-acquired target push model under each text and image type; the aggregation layer is used to roughly sort the candidate data sets under each text and image type to obtain a first text and image queue under each text and image type; the preference distribution adjustment layer is used to obtain a reference push type distribution corresponding to a user terminal, and obtain a second text and image queue corresponding to a text and image type set based on the first text and image queue of each text and image type according to the reference push type distribution; the fusion layer is used to finely sort the second text and image queue to obtain candidate push text and image content under each text and image type, and obtain the target push text and image content of the user terminal according to each candidate push text and image content; the distribution layer is used to push the target push text and image content to the user terminal.

[0008] According to a fourth aspect of the present disclosure, a device for pushing graphic content is proposed, comprising: a first acquisition module, used to acquire a graphic type set corresponding to a user terminal, and a target push model trained under each graphic type in the graphic type set; a second acquisition module, used to acquire an initial data set, and acquire a candidate data set under each graphic type from the initial data set based on each target push model; a third acquisition module, used to acquire a first graphic queue under each graphic type from the candidate data set to obtain a second graphic queue corresponding to the graphic type set; a push module, used to acquire target push graphic content from the second graphic queue, and push the target push graphic content to the user terminal.

[0009] According to the fifth aspect of the present disclosure, a training device for a picture and text push model is proposed, comprising: a fourth acquisition module, used to obtain candidate push models to be trained under each sample picture and text type, and to obtain sample picture and text content of a sample user terminal under each sample picture and text type; a collection module, used to collect sample feedback information on the sample picture and text content under each sample picture and text type, and sample user preferences of the sample user terminal for each sample picture and text type; a generation module, used to generate, for any sample picture and text type, training samples of candidate push models corresponding to the sample picture and text type according to the sample picture and text content under the sample picture and text type, sample feedback information corresponding to the sample picture and text content, and sample user preferences corresponding to the sample picture and text type; a training module, used to train, for any sample picture and text type, the candidate push model of the sample picture and text type according to the training samples of the sample picture and text type, until the training is completed, to obtain a trained target push model under the sample picture and text type, wherein the target push model is used to implement the picture and text content push device proposed in the fourth aspect.

[0010] According to the sixth aspect of the present disclosure, an electronic device is proposed, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for pushing image and text content proposed in the first aspect and / or the training method for the image and text push model proposed in the second aspect.

[0011] According to the seventh aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is proposed, wherein the computer instructions are used to enable the computer to execute the image and text content pushing method proposed in the first aspect and / or the image and text push model training method proposed in the second aspect.

[0012] According to an eighth aspect of the present disclosure, a computer program product is proposed, including a computer program, which, when executed by a processor, implements the method for pushing text and image content proposed in the first aspect and / or the method for training a text and image push model proposed in the second aspect.

[0013] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0015] Figure 1 A schematic diagram of a flow chart of a method for pushing graphic content according to an embodiment of the present disclosure;

[0016] Figure 2 A schematic diagram of a flow chart of a method for pushing graphic content according to another embodiment of the present disclosure;

[0017] Figure 3 A schematic diagram of graphic content of an embodiment of the present disclosure;

[0018] Figure 4 A schematic diagram of a flow chart of a method for pushing graphic content according to another embodiment of the present disclosure;

[0019] Figure 5 A schematic diagram of a system for pushing graphic and text content according to an embodiment of the present disclosure;

[0020] Figure 6 This is a schematic diagram of the structure of a device for pushing graphic content according to an embodiment of the present disclosure;

[0021] Figure 7A schematic diagram of the structure of a training device for a picture and text push model according to an embodiment of the present disclosure;

[0022] Figure 8 A schematic block diagram of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0023] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0024] Data processing is a basic part of system engineering and automatic control. Data is a form of expression of facts, concepts or instructions, which can be processed by manual or automatic devices. After data is interpreted and given a certain meaning, it becomes information. Data processing is the collection, storage, retrieval, processing, transformation and transmission of data. The basic purpose of data processing is to extract and derive valuable and meaningful data for certain specific people from a large amount of data that may be disorganized and difficult to understand.

[0025] Deep learning (DL) refers to machine learning based on deep neural network models and methods. It is developed based on statistical machine learning, artificial neural network and other algorithm models, combined with the development of contemporary big data and large computing power. The most important technical feature of deep learning is the ability to automatically extract features. Neural network algorithms, computing power and data are the three elements for deep learning.

[0026] Artificial Intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. Artificial Intelligence is a branch of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing and expert systems. Since the birth of artificial intelligence, the theory and technology of artificial intelligence have become increasingly mature, and the application field has continued to expand. It can be imagined that the technological products brought by artificial intelligence in the future will be the "container" of human wisdom. Artificial Intelligence can simulate the information process of human consciousness and thinking.

[0027] Figure 1FIG. 1 is a flow chart of a method for pushing graphic content according to an embodiment of the present disclosure. Figure 1 As shown, the method includes:

[0028] S101, obtaining a set of image and text types corresponding to the user terminal, and a trained target push model for each image and text type in the set of image and text types.

[0029] In practice, more and more people choose video software equipped with immersive graphic scenes to read graphic content and obtain information in their daily work and life. In this scenario, users may use sliding and other methods to operate the graphic content, and will not click on the graphic content they are watching.

[0030] In this scenario, the user preference information of the user terminal can be collected based on the historical graphic content viewing records of the user terminal, and then the graphic content can be accurately pushed to the user terminal based on the collected user preference information.

[0031] In the disclosed embodiment, based on the acquired user preference information of the user terminal, all image and text types can be screened to obtain multiple image and text types preferred by the user terminal, and the set consisting of the multiple image and text types can be marked as the image and text type set corresponding to the user terminal.

[0032] In the disclosed embodiment, the push content screening conditions for each image and text type may be different. In this scenario, a trained model can be obtained for each image and text type, and the push content for the user end can be screened from the pre-acquired data set based on the trained model for each image and text type.

[0033] Among them, the trained model under each image and text type can be marked as the target push model under each image and text type.

[0034] S102, obtaining an initial data set, and obtaining candidate data sets of each image and text type from the initial data set based on each target push model.

[0035] In the embodiment of the present disclosure, a data set including graphic content of each graphic type in all graphic types set on the server can be marked as an initial data set, wherein the initial data set can be constructed based on historical uploaded graphic content received by the server.

[0036] As an example, for a certain dynamic video software, there are multiple users registered on the dynamic video software. In this example, an initial data set on the server of the dynamic video software can be constructed based on the graphic content works uploaded by each user within a historical time range.

[0037] In the disclosed embodiment, data sets under each image and text type can be filtered out from the initial data set according to each image and text type in the image and text type set corresponding to the user terminal. For any image and text type, the data set under the image and text type can be input into the target push model corresponding to the image and text type, and the image and text content of the data set can be filtered through the target push model, so as to filter out multiple data that meet the preset conditions, and the multiple data can be respectively marked as candidate data.

[0038] Furthermore, the set consisting of the plurality of candidate data is marked as a candidate data set under the graphic and text type.

[0039] S103: Acquire a first image and text queue under each image and text type from the candidate data set to obtain a second image and text queue corresponding to the image and text type set.

[0040] In the disclosed embodiment, for any image and text type, a preset image and text data processing strategy for the image and text type can be obtained, and a candidate data set of the image and text type can be processed according to the strategy to form each data in the candidate data set into a data queue in a preset format.

[0041] The data queue may be marked as the first image and text queue under the image and text type.

[0042] As a possible implementation method, an algorithm available for any image and text type can be obtained from the queue construction algorithm in the relevant technology, and the candidate data set under the image and text type can be processed based on the algorithm, and the first image and text queue under the image and text type can be obtained based on the result of the algorithm processing.

[0043] In the disclosed embodiment, the set of graphic content pushed to the user terminal includes multiple graphic types. In this scenario, it is also necessary to integrate the first graphic queues under each graphic type into one graphic queue, and further filter the pushed content from the integrated graphic queue, wherein the integrated queue can be marked as the second graphic queue corresponding to the graphic type set.

[0044] The second image and text queue includes the first image and text queues of all image and text types in the image and text type set.

[0045] S104, obtaining target push graphic content from the second graphic queue, and pushing the target push graphic content to the user end.

[0046] Optionally, user preference information of the user terminal can be obtained, and the compatibility between the graphic content and the user preference information can be screened from the second graphic queue based on the user preference information, and then the graphic content to be pushed to the user terminal can be screened from the second graphic queue and marked as the target pushed graphic content to the user terminal.

[0047] Among them, according to the adaptation algorithm in the relevant technology, each graphic content in the second graphic content queue and the user preference information of the user end can be algorithmically processed respectively, so as to obtain the adaptation degree between each graphic content in the second graphic content queue and the user preference information, and screening can be performed based on the adaptation degree to obtain the target pushed graphic content.

[0048] Optionally, the push priority of each graphic content in the second graphic content queue can be obtained, and the graphic content in the second graphic content queue can be sorted according to each push priority, and then a set number of graphic content can be obtained from the sorted queue as the target pushed graphic content to the user end.

[0049] Furthermore, the selected target push graphic and text contents are pushed to the user terminal through the information interaction link between the server and the user terminal, and displayed on the user terminal.

[0050] The method for pushing graphic content proposed in the present disclosure obtains a set of graphic types corresponding to the user terminal, and a target push model under each graphic type, and then obtains a candidate data set under each graphic type from the initial data set according to each target push model. Optionally, a first graphic queue under each graphic type is obtained according to each candidate data set to form a second graphic queue corresponding to the graphic type set, and then the target push graphic content is screened out from the second graphic queue and pushed to the user terminal. In the present disclosure, the graphic content is screened according to each graphic type corresponding to the user preference information of the user terminal, so as to obtain the target push graphic content, improve the degree of adaptation between the target push graphic content and the personal preference of the user terminal, and improve the accurate acquisition of the target push graphic content in the immersive graphic scene. The target push graphic content is screened out through the second graphic queue composed of the first graphic queue under each graphic type, so that the target push graphic content can include multiple graphic types, which optimizes the diversity and richness of the target push graphic content, and thus optimizes the user experience.

[0051] In the above embodiment, the acquisition of the target push text and image content can also be combined with Figure 2 Further understanding, Figure 2 FIG. 1 is a flow chart of a method for pushing graphic content in another embodiment of the present disclosure. Figure 2 As shown, the method includes:

[0052] S201, obtaining an initial data set, and obtaining candidate data sets of each image and text type from the initial data set based on each target push model.

[0053] Optionally, user preference information of the user terminal is obtained, and the initial data set is screened based on the user preference information through the target push model under each image and text type to output candidate data sets of each image and text type.

[0054] In the embodiment of the present disclosure, the user terminal has corresponding personal preferences, and the relevant information of the personal preferences can be marked as the user preference information of the user terminal.

[0055] As an example, in an immersive picture and text scenario, a user can post multiple pictures and display the posted text at the bottom of the pictures.

[0056] Among them, feedback information such as the length of time the user stays on the graphic content, whether he swipes right to view subsequent pictures, the number of pictures viewed, whether he likes, whether he clicks to view the author's avatar, whether he follows the author, whether he views the comment area, and whether he makes comments in the comment area can be obtained, and user preference information for the graphic content can be obtained based on the collected feedback information.

[0057] It should be noted that, in this example, the feedback information from the user end collected by the server includes the feedback information mentioned above, and may also include feedback information generated by other operations performed by the user end on the graphic content, which is not specifically limited here.

[0058] As another example, the immersive graphic scene may also include Figure 3 The graphic content shown is as follows: Figure 3 As shown in the figure, users can publish graphic content with titles, pictures and article bodies, where the cover image can be displayed in the upper part of the graphic content, the title can be displayed in the middle part, and the summary information of the article can be displayed in the lower part, and users can be provided with the option of viewing the full text. Figure 3 In the scenario shown, the user can click the option to view the full text to enter the corresponding full-text page of the graphic content for viewing.

[0059] like Figure 3 As shown, in Figure 3 In the scenario shown, the user end can collect Figure 3 Feedback information such as the length of time the user stays on the graphic content, whether the user clicks to view the full text, the viewing time of the full text page, whether the user likes the content, whether the user clicks to view the author's avatar, whether the user follows the author, whether the user views the comment area, and whether the user posts a comment in the comment area are displayed, and then the user preference information of the user end for the graphic content is obtained based on the collected feedback information.

[0060] It should be noted that, in this example, the feedback information from the user end collected by the server includes the feedback information mentioned above, and may also include feedback information generated by other operations performed by the user end on the graphic content, which is not specifically limited here.

[0061] Optionally, for any image and text type, algorithmic processing can be performed on each image and text data and user preference information in the initial data based on a matching algorithm in related technologies, and the matching degree between each image and text data and the user preference information is obtained according to the result of the algorithmic processing, and candidate data sets under this image and text type are screened out from the initial data set based on the matching degree.

[0062] Optionally, for any image and text type, the initial data set can also be input into the target push model trained for the image and text type, and the feature information of each image and text data in the initial data set can be extracted respectively through the target push model, and compared with the user preference information of the user end based on the extracted feature information, and then some candidate data that matches the user preference information of the user end can be screened out from the initial data set to form a candidate data set for the image and text type.

[0063] Furthermore, candidate data sets of each image and text type are obtained.

[0064] S202: Acquire a first image and text queue under each image and text type from the candidate data set to obtain a second image and text queue corresponding to the image and text type set.

[0065] Optionally, for any image and text type, each first image and text in the candidate data set under the image and text type is roughly sorted to obtain a sorted first image and text queue.

[0066] In the disclosed embodiment, for any image and text type, each candidate data in the candidate data set of the image and text type has a different degree of matching with the user preference information. In this scenario, the image and text data in the candidate data set can be sorted based on a preset sorting strategy.

[0067] Optionally, based on the rough sorting algorithm in the relevant technology, each graphic data in the candidate data set can be roughly sorted by an algorithm based on the order of matching degree from high to low, and then the candidate data set can be roughly sorted, and the graphic content queue obtained after the rough sorting can be marked as the first graphic queue under the graphic type.

[0068] Optionally, a reference push type distribution of the user terminal is obtained according to the user preference information, and a third image and text queue under each image and text type is obtained from each first image and text queue according to the reference push type distribution.

[0069] In the disclosed embodiment, user preference information of the user terminal can be obtained, and based on the user preference information, each image and text type in the image and text type set is ranked in order of preference from high to low. Then, when pushing image and text content to the user terminal, the preference ranking is used to determine the proportion of each image and text type in the pushed content, and then the reference push type distribution of the user terminal is determined according to the proportion of each image and text type.

[0070] As an example, the type of graphics and text that is most preferred by the user side may be determined to have the highest proportion in the number of pushed contents, followed by decreasing proportions in sequence, thereby obtaining a reference push type distribution in this example.

[0071] In this scenario, the image and text contents can be screened from the first image and text queue of each image and text type according to the reference push type distribution of the user end obtained, which can be understood in combination with the following contents:

[0072] Optionally, the number of first images and texts corresponding to the third image and text queue is obtained, and the proportion of the number of images and texts of each image and text type is determined according to the reference push type distribution.

[0073] In the embodiment of the present disclosure, in order to improve the richness of the types of graphics and text included in the target pushed graphics and text content, at least part of the graphics and text content can be filtered out from each first graphics and text queue based on the reference push type distribution, and the queue consisting of at least part of the filtered graphics and text content can be marked as the third graphics and text queue under each graphics and text type.

[0074] In this scenario, the graphic content included in the third graphic queue is limited by a preset quantity, and this quantity can be marked as the first graphic quantity of the third graphic queue.

[0075] In the disclosed embodiment, the reference push type distribution of the user terminal carries the proportion of each graphic type in the graphic type set corresponding to the user terminal in the target push graphic content, and the proportion can be marked as the graphic quantity proportion of each graphic type.

[0076] Optionally, the second image and text quantity of each image and text type is obtained based on the image and text quantity ratio and the first image and text quantity. For any image and text type, due to the order from front to back, the third image and text queue of the second image and text quantity is obtained from the first image and text queue of the image and text type.

[0077] In the disclosed embodiment, for any image and text type, the number of image and text contents occupied by the image and text type in the third image and text queue can be obtained based on the proportion of the image and text quantity corresponding to the image and text type and the total first image and text quantity included in the third image and text queue, and the quantity can be marked as the second image and text quantity of the image and text type, wherein the proportion of the image and text quantity corresponding to the image and text type can be multiplied by the first image and text quantity, and the product value of the two can be determined as the second image and text quantity of the image and text type.

[0078] In this scenario, for any image and text type, the image and text content can be screened from the first image and text queue of the image and text type based on the second image and text quantity corresponding to the image and text type in order from front to back, and the second image and text content is obtained, and the queue formed by them is marked as the third image and text queue of the image and text type.

[0079] Optionally, the third image and text queues under each image and text type are integrated to obtain a second image and text queue corresponding to the image and text type set.

[0080] As a possible implementation manner, the third image and text queues of each image and text type may be sequentially spliced ​​based on a preset order, and the spliced ​​image and text queue obtained is marked as the second image and text queue after the third image and text queues are integrated.

[0081] As another possible implementation method, the third image and text queues under each image and text type can be broken up separately, and the broken up image and text contents can be queued based on a random order, and then the constructed image and text queue can be marked as the second image and text queue obtained by integrating the third image and text queues.

[0082] S203: Obtain target pushed graphic content from the second graphic queue.

[0083] Optionally, a preset first push quantity of the user terminal is obtained, and a second push quantity of each image and text type is determined according to the first push quantity and a proportion of the image and text quantity of each image and text type.

[0084] In the scenario of pushing graphic content to the user terminal, there is a set number of graphic content pushes on the user terminal. This number can be used as the number of graphic content included in the target pushed graphic content pushed to the user terminal in each push round, and this number can be marked as the first push number of the user terminal.

[0085] In this scenario, for any type of image and text, the first push quantity can be multiplied by the proportion of the image and text corresponding to the image and text to obtain the product value of the two, wherein the product value is the number of image and text contents under the image and text type in the target pushed image and text content pushed to the user end, wherein this number can be marked as the second push quantity of the image and text type.

[0086] Optionally, the second image and text queue is finely sorted to obtain a sorted fourth image and text queue.

[0087] In the disclosed embodiment, based on the precise sorting algorithm in the related art and based on the degree of match between each graphic content in the second graphic queue and the user preference information, the second graphic queue can be processed by a precise sorting algorithm, and the queue obtained after the precise sorting can be marked as the fourth graphic queue.

[0088] Optionally, for any image and text type, according to the second push quantity of the image and text type, a second push quantity of candidate pushed image and text contents are obtained from the fourth image and text queue.

[0089] In the disclosed embodiment, for any image and text type, the second pushed quantity of image and text content of the image and text type can be screened out from the fourth image and text queue based on the order from front to back as the candidate pushed image and text content corresponding to the image and text type.

[0090] Optionally, target pushed graphic and text content of the user terminal is obtained according to the candidate pushed graphic and text content under each graphic and text type.

[0091] In the disclosed embodiment, after obtaining the candidate push graphic content under each graphic type, the candidate push graphic content can be sorted and integrated according to the degree of match between each candidate push graphic content and the user preference information, and the integrated graphic content can be determined as the target push graphic content to be pushed to the user end.

[0092] The method for pushing graphic and text content proposed in the present invention screens out graphic and text content queues under various graphic and text types according to user preference information, thereby improving the degree of adaptation between the target pushed graphic and text content and the personal preferences of the user end, as well as the efficiency of obtaining the target pushed graphic and text content. The graphic and text content is screened according to the reference push type distribution, so that the target pushed graphic and text content includes multiple graphic and text types corresponding to the user preference information, thereby improving the richness and type diversity of the target pushed content.

[0093] Regarding the target push model proposed in the above embodiment, it can be combined with Figure 4 Further understanding, Figure 4 FIG. 1 is a flow chart of a training method for a push model according to an embodiment of the present disclosure. Figure 4 As shown, the method includes:

[0094] S401, obtaining candidate push models to be trained under each sample image and text type, and obtaining sample image and text content of a sample user terminal under each sample image and text type.

[0095] In an embodiment of the present disclosure, the image and text types available on the server can be marked as sample image and text types in a model training scenario. In order to implement the screening of image and text content under each sample image and text type, a corresponding push content screening model can be set for each sample image and text type as a candidate push model to be trained under each sample image and text type.

[0096] Optionally, relevant sample data may be collected to train each candidate push model, wherein the user terminal used to collect the sample data may be marked as a sample user terminal.

[0097] In this scenario, the graphic content of each sample graphic type displayed on the sample user terminal may be marked as sample graphic content of each sample graphic type.

[0098] S402 , collecting sample feedback information on sample graphic content under each sample graphic type, and sample user preferences of sample users for each sample graphic type.

[0099] In the embodiment of the present disclosure, for any sample image and text type, feedback information of the sample user terminal under the sample image and text type may be collected, and the collected information may be marked as sample feedback information of the sample user terminal under the sample image and text type.

[0100] As an example, for Figure 3 The sample graphic content under the sample graphic type shown can collect sample feedback information of the sample user terminal on the sample graphic content, wherein the sample feedback information may include the sample user terminal Figure 3 The displayed information includes the length of time the user stays on the graphic and text content, whether the user clicks to view the full text, the length of time the full text page is viewed, whether the user likes the content, whether the user clicks to view the author's avatar, whether the user follows the author, whether the user views the comment section, and whether the user posts a comment in the comment section.

[0101] In the disclosed embodiment, the sample user terminal has a corresponding preference degree for each sample image and text type, wherein the sample user terminal can annotate the preference degree for the sample image and text content under each sample image and text type according to a preset annotation method, and then obtain the sample user preference of the sample user terminal for each sample image and text type based on the annotation results.

[0102] S403, for any sample image and text type, generating a training sample of a candidate push model corresponding to the sample image and text type according to the sample image and text content under the sample image and text type, sample feedback information corresponding to the sample image and text content, and sample user preference corresponding to the sample image and text type.

[0103] In the disclosed embodiment, for any sample image and text type, the sample user preference may be used as label information of the sample image and text content and sample feedback information to obtain a training sample of a candidate push model corresponding to the sample image and text type.

[0104] Optionally, for any sample image and text type, an algorithm can be constructed based on samples in related technologies to perform algorithmic processing on the sample image and text content of the sample user terminal under the sample image and text type, the sample feedback information corresponding to the sample image and text content, and the sample user preference of the sample user terminal for the sample image and text type, and then obtain the training sample of the candidate push model corresponding to the sample image and text type based on the result of the algorithmic processing.

[0105] S404: For any sample image and text type, a candidate push model of the sample image and text type is trained according to the training samples of the sample image and text type until the training is completed to obtain a trained target push model of the sample image and text type.

[0106] Among them, the target push model is used to achieve the above Figures 1 to 3 The method for pushing graphic and text content proposed in the embodiment.

[0107] In the disclosed embodiment, for any sample image and text type, a training sample corresponding to the sample image and text type can be obtained, and the training sample can be input into a candidate push model corresponding to the sample image and text type for model training to obtain the training loss of the candidate push model, and the parameters of the candidate push model can be adjusted based on the training loss.

[0108] Furthermore, the next training sample is returned to continue training the candidate push model after reference adjustment until the training end condition is met, and the training of the candidate push model is ended, and the model obtained after the training is determined as the trained target push model under the sample image and text type.

[0109] Among them, the corresponding training end conditions can be set according to the training rounds. For the current round of model training, the total training rounds of the candidate push model after the end of this round of training can be compared with the preset training end conditions. If the current total training rounds meet the training end conditions, the training of the model can be ended, and the model obtained after the end of the current round of training is determined as the trained target push model.

[0110] Accordingly, corresponding training end conditions can also be set according to the model output results. For the current round of model training, the model output results of this round can be compared with the preset training end conditions. When the output results meet the training end conditions, the training of the model can be ended, and the model obtained after the current round of training is determined as the trained target push model.

[0111] The training method of the push model proposed in the present disclosure obtains sample feedback information of the sample user terminal under each image and text type, generates training samples under each image and text type, wherein, for any image and text type, the candidate push model under the image and text type is trained based on the training samples under the image and text type until the training is completed, and a trained target push model under the image and text type is obtained. In the present disclosure, each candidate push model is trained according to the sample image and text content, sample feedback information and sample user preferences under each sample image and text type, so that any trained target push model can learn the correlation between the sample image and text content, sample feedback information and sample user preferences under the sample image and text type to which it belongs, and in the scenario of filtering push content based on the target push model, the filtering accuracy and filtering efficiency of the image and text push content are improved, thereby improving the image and text content push efficiency and push accuracy of the user terminal, and optimizing the user experience.

[0112] The present disclosure also proposes a graphic content push system that can be combined with Figure 5 understand, Figure 5 FIG. 1 is a schematic diagram of a system for pushing graphic content according to an embodiment of the present disclosure. Figure 5 As shown, the graphic content push system 500 includes a graphic content data layer 51, a convergence layer 52, a preference distribution adjustment layer 53, a fusion layer 54 and a distribution layer 55, wherein:

[0113] The graphic content data layer 51 is used to store the initial data set of the server and obtain the candidate data set of each graphic type from the initial data set through the pre-acquired target push model of each graphic type.

[0114] like Figure 5 As shown, the initial data set can be stored in Figure 5 The graphic content data layer 51 shown in the figure is input into the target push model under each pre-acquired graphic type, wherein the initial data set includes the graphic content of each of the multiple graphic types, which may include Figure 5 The short video types, small video types, dynamic graphic types, and static graphic types shown are not specifically limited here.

[0115] For any image and text type, the target push model under the image and text type can be used to filter the image and text content of the initial data set, and the set of filtered image and text data is marked as a candidate data set under the image and text type.

[0116] It should be noted that the candidate data sets may be screened according to the degree of matching between each data in the initial data set and the user preference information of the user terminal, or according to other preset methods, which are not specifically limited here.

[0117] The aggregation layer 52 is used to roughly sort the candidate data sets under each image and text type to obtain the first image and text queue under each image and text type.

[0118] In the embodiment of the present disclosure, Figure 5 The aggregation layer shown performs rough sorting on the candidate data sets under each image and text type, wherein the candidate data sets under any image and text type can be roughly sorted based on the rough sorting algorithm in the relevant technology, and the queue composed of each image and text content obtained after the rough sorting is marked as the first image and text queue of the image and text type.

[0119] The preference distribution adjustment layer 53 is used to obtain the reference push type distribution corresponding to the user terminal, and obtain the second image and text queue corresponding to the image and text type set based on the first image and text queue of each image and text type according to the reference push type distribution.

[0120] In the disclosed embodiment, the user terminal has its own preference for each type of image and text. In this scenario, the push proportion of each type of image and text in the pushed content can be determined based on the user preference information of the user terminal, thereby obtaining a reference push type distribution corresponding to the user terminal.

[0121] In this scenario, you can Figure 5 The preference distribution adjustment layer 53 shown in the figure screens the first image and text queue under each image and text type, wherein the preference distribution adjustment layer 53 can be Figure 5 The preference distribution adjustment layer 53 shown determines the proportion of each graphic type in the total pushed content, and determines the quantity of graphic content of each graphic type according to the proportion and the total pushed content quantity.

[0122] Optionally, for any graphic type, Figure 5 The preference distribution adjustment layer 53 shown screens the graphic content from the first graphic queue under the graphic type according to the number of graphic content of the graphic type, and obtains the third graphic queue under the graphic type.

[0123] Further, through Figure 5 The preference distribution adjustment layer 53 shown integrates the third image and text queues under each image and text type to obtain a total image and text queue composed of the third image and text queues, and marks it as the second image and text queue corresponding to the image and text type set.

[0124] The fusion layer 54 is used to precisely sort the second image and text queue to obtain candidate push image and text content under each image and text type, and obtain target push image and text content of the user terminal according to each candidate push image and text content.

[0125] In the embodiment of the present disclosure, Figure 5 The fusion layer shown performs precise sorting on the second image and text queue, and selects image and text contents to be pushed under each image and text type from the precisely sorted image and text queue, and marks them as candidate pushed image and text contents under each image and text type.

[0126] Further, through Figure 5 The fusion layer 54 shown integrates the candidate pushed graphic content under each graphic type to obtain the target pushed graphic content to be pushed to the user end.

[0127] The distribution layer 55 is used to push the target push graphic content to the user end.

[0128] Optionally, by Figure 5 The distribution layer 55 shown pushes the target push graphic content acquired by the fusion layer 54 to the user end based on the data interaction link between the server and the user end.

[0129] The graphic content push system proposed in the present invention realizes the screening of graphic content based on the user preferences of the user end, and performs coarse sorting and fine sorting on the screened graphic content, thereby improving the degree of adaptation between the final target pushed graphic content and the user preferences, improving the type diversity and content richness of the target pushed graphic content, and optimizing the user experience.

[0130] An embodiment of the present disclosure further proposes a graphic content pushing device. Since the graphic content pushing device proposed in the embodiment of the present disclosure corresponds to the graphic content pushing methods proposed in the above-mentioned embodiments, the implementation methods of the above-mentioned graphic content pushing methods are also applicable to the graphic content pushing device proposed in the embodiment of the present disclosure, and will not be described in detail in the following embodiments.

[0131] Figure 6 FIG. 1 is a schematic diagram of a structure of a device for pushing graphic content according to an embodiment of the present disclosure. Figure 6 As shown, the device 600 for pushing graphic content includes a first acquisition module 61, a second acquisition module 62, a third acquisition module 63 and a push module 64, wherein:

[0132] The first acquisition module 61 is used to acquire a set of image and text types corresponding to the user terminal, and a trained target push model for each image and text type in the set of image and text types;

[0133] The second acquisition module 62 is used to acquire an initial data set, and acquire candidate data sets of each image and text type from the initial data set based on each target push model;

[0134] The third acquisition module 63 is used to acquire the first image and text queue under each image and text type from the candidate data set to obtain the second image and text queue corresponding to the image and text type set;

[0135] The push module 64 is used to obtain target push graphic content from the second graphic queue, and push the target push graphic content to the user end.

[0136] In the disclosed embodiment, the second acquisition module 62 is further used to: obtain user preference information of the user terminal; and filter the initial data set based on the user preference information through the target push model under each image and text type to output candidate data sets of each image and text type.

[0137] In the disclosed embodiment, the third acquisition module 63 is also used to: for any image and text type, roughly sort each first image and text in the candidate data set under the image and text type to obtain a sorted first image and text queue; obtain a reference push type distribution of the user terminal according to user preference information, and obtain a third image and text queue under each image and text type from each first image and text queue according to the reference push type distribution; integrate the third image and text queue under each image and text type to obtain a second image and text queue corresponding to the image and text type set.

[0138] In the disclosed embodiment, the third acquisition module 63 is also used to: obtain the first number of images and texts corresponding to the third image and text queue; determine the proportion of the number of images and texts of each image and text type based on the reference push type distribution; obtain the second number of images and texts of each image and text type based on the proportion of the number of images and texts and the first number of images and texts; for any image and text type, due to the order from front to back, obtain the third image and text queue of the second number of images and texts from the first image and text queue of the image and text type.

[0139] In the disclosed embodiment, the push module 64 is also used to: obtain a preset first push quantity of the user terminal, and determine a second push quantity of each image and text type based on the first push quantity and the proportion of the image and text quantity of each image and text type; finely sort the second image and text queue to obtain a sorted fourth image and text queue; for any image and text type, according to the second push quantity of the image and text type, obtain the second push quantity of candidate push image and text content from the fourth image and text queue; according to the candidate push image and text content under each image and text type, obtain the target push image and text content of the user terminal.

[0140] The device for pushing graphic content proposed in the present disclosure obtains a set of graphic types corresponding to the user terminal, and a target push model under each graphic type, and then obtains a candidate data set under each graphic type from the initial data set according to each target push model. Optionally, a first graphic queue under each graphic type is obtained according to each candidate data set to form a second graphic queue corresponding to the graphic type set, and then the target push graphic content is screened out from the second graphic queue and pushed to the user terminal. In the present disclosure, the graphic content is screened according to each graphic type corresponding to the user preference information of the user terminal, so as to obtain the target push graphic content, thereby improving the degree of adaptation between the target push graphic content and the personal preference of the user terminal, and improving the accurate acquisition of the target push graphic content in the immersive graphic scene. The target push graphic content is screened out through the second graphic queue composed of the first graphic queue under each graphic type, so that the target push graphic content can include multiple graphic types, which optimizes the diversity and richness of the target push graphic content, and thus optimizes the user experience.

[0141] An embodiment of the present disclosure further proposes a training device for a picture and text push model. Since the training device for the picture and text push model proposed in the embodiment of the present disclosure corresponds to the training methods for the picture and text push model proposed in the above-mentioned embodiments, the implementation methods of the above-mentioned training methods for the picture and text push model are also applicable to the training device for the picture and text push model proposed in the embodiment of the present disclosure, and will not be described in detail in the following embodiments.

[0142] Figure 7 FIG. 1 is a structural diagram of a training device for a picture and text push model according to an embodiment of the present disclosure. Figure 7 As shown, the training device 700 of the picture and text push model includes a fourth acquisition module 71, a collection module 72, a generation module 73 and a training module 74, wherein:

[0143] The fourth acquisition module 71 is used to acquire candidate push models to be trained under each sample image and text type, and acquire sample image and text content of the sample user terminal under each sample image and text type;

[0144] A collection module 72, used to collect sample feedback information on sample graphic content under each sample graphic type, and sample user preferences of sample users for each sample graphic type;

[0145] A generating module 73, for generating, for any sample image and text type, a training sample of a candidate push model corresponding to the sample image and text type according to the sample image and text content under the sample image and text type, the sample feedback information corresponding to the sample image and text content, and the sample user preference corresponding to the sample image and text type;

[0146] The training module 74 is used to train the candidate push model of the sample image and text type according to the training sample of the sample image and text type for any sample image and text type until the training is completed, and obtain the trained target push model under the sample image and text type, wherein the target push model is used to implement the above Figure 6 The embodiment provides a device for pushing graphic and text content.

[0147] The training device of the push model proposed in the present disclosure obtains sample feedback information of the sample user terminal under each image and text type, and generates training samples under each image and text type, wherein, for any image and text type, the candidate push model under the image and text type is trained based on the training samples under the image and text type until the training is completed, and a trained target push model under the image and text type is obtained. In the present disclosure, each candidate push model is trained according to the sample image and text content, sample feedback information and sample user preferences under each sample image and text type, so that any trained target push model can learn the correlation between the sample image and text content, sample feedback information and sample user preferences under the sample image and text type to which it belongs, and in the scenario of pushing content screening based on the target push model, the screening accuracy and screening efficiency of the image and text push content are improved, thereby improving the image and text content push efficiency and push accuracy of the user terminal, and optimizing the user experience.

[0148] According to an embodiment of the present disclosure, the present disclosure also proposes an electronic device, a readable storage medium, and a computer program product.

[0149] Figure 8 A schematic block diagram of an example electronic device 800 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0150] like Figure 8 As shown, the device 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 809 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0151] A number of components in the device 800 are connected to the I / O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 806, such as various types of displays, speakers, etc.; a storage unit 809, such as a disk, an optical disk, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the device 800 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0152] The computing unit 801 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 801 performs the various methods and processes described above, such as a method for pushing text content and / or a training method for a text content push model. For example, in some embodiments, the method for pushing text content and / or the training method for a text content push model may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 809. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the method for pushing text content and / or the training method for a text content push model described above may be executed. Alternatively, in other embodiments, the computing unit 801 may be configured to execute the image and text content pushing method and / or the image and text pushing model training method in any other appropriate manner (for example, by means of firmware).

[0153] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0154] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be presented to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, implements the functions / operations specified in the flow chart and / or block diagram. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0155] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0156] To propose interactions with a user account, the systems and techniques described herein may be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user account; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user account can propose input to the computer. Other types of devices may also be used to propose interactions with the user account; for example, feedback proposed to the user account may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user account may be received in any form (including acoustic input, voice input, or tactile input).

[0157] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user account computer with a graphical user account interface or a web browser through which a user account can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0158] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0159] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0160] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for pushing graphic content, wherein: The method comprises: Obtaining a set of image and text types corresponding to the user end, and a trained target push model for each image and text type in the set of image and text types; Acquire an initial data set, and acquire candidate data sets of each image and text type from the initial data set based on each target push model; Acquire a first image and text queue under each image and text type from the candidate data set to obtain a second image and text queue corresponding to the image and text type set; The target pushed graphic content is acquired from the second graphic content queue, and the target pushed graphic content is pushed to the user terminal.

2. The method according to claim 1, wherein: The obtaining of the initial data set, and obtaining candidate data sets of each image and text type from the initial data set based on each target push model, includes: Obtain user preference information from the user end; The initial data set is screened based on the user preference information through the target push model under each image and text type to output a candidate data set for each image and text type.

3. The method according to claim 1, wherein: The acquiring the first image and text queue under each image and text type from the candidate data set to obtain the second image and text queue corresponding to the image and text type set includes: For any image and text type, roughly sorting each first image and text in the candidate data set under the image and text type to obtain a sorted first image and text queue; Acquire a reference push type distribution of the user terminal according to the user preference information, and acquire a third image and text queue under each image and text type from each first image and text queue according to the reference push type distribution; The third image and text queues under each image and text type are integrated to obtain the second image and text queue corresponding to the image and text type set.

4. The method according to claim 3, wherein: The obtaining of the reference push type distribution of the user terminal according to the user preference information, and obtaining the third image and text queue of each image and text type from each first image and text queue according to the reference push type distribution, includes: Obtain the number of first images and texts corresponding to the third image and text queue; Determine the proportion of images and texts of each image and text type according to the reference push type distribution; According to the proportion of the number of images and texts and the first number of images and texts, obtaining the second number of images and texts of each type of image and text; For any image and text type, due to the order from front to back, the third image and text queue of the second number of images and texts is obtained from the first image and text queue of the image and text type.

5. The method according to claim 1, wherein: The step of obtaining target pushed graphic content from the second graphic queue includes: Acquire a preset first push quantity of the user terminal, and determine a second push quantity of each type of image and text according to the first push quantity and a proportion of the number of images and texts of each type of image and text; Finely sorting the second image and text queue to obtain a sorted fourth image and text queue; For any image and text type, according to the second push quantity of the image and text type, obtaining the second push quantity of candidate pushed image and text contents from the fourth image and text queue; The target pushed graphic content of the user terminal is obtained according to the candidate pushed graphic content under each graphic type.

6. A method for training a push model, wherein: The method comprises: Obtain candidate push models to be trained under each sample image and text type, and obtain sample image and text content of the sample user terminal under each sample image and text type; Collecting sample feedback information on sample graphic content under each sample graphic type, and sample user preferences of the sample user terminal for each sample graphic type; For any sample image and text type, generating a training sample of a candidate push model corresponding to the sample image and text type according to the sample image and text content under the sample image and text type, sample feedback information corresponding to the sample image and text content, and sample user preferences corresponding to the sample image and text type; For any sample image and text type, the candidate push model of the sample image and text type is trained according to the training samples of the sample image and text type until the training is completed, so as to obtain a trained target push model under the sample image and text type, wherein the target push model is used to implement the image and text content push method described in any one of claims 1 to 5 above.

7. A system for pushing graphic content, wherein: The system includes a graphic content data layer, a graphic aggregation layer, a preference distribution adjustment layer, a fusion layer, and a distribution layer, wherein: The graphic content data layer is used to store the initial data set of the server, and obtain the candidate data set of each graphic type from the initial data set through the pre-acquired target push model of each graphic type; The convergence layer is used to roughly sort the candidate data sets under each image and text type to obtain the first image and text queue under each image and text type; The preference distribution adjustment layer is used to obtain a reference push type distribution corresponding to the user end, and obtain a second image and text queue corresponding to the image and text type set based on the first image and text queue of each image and text type according to the reference push type distribution; The fusion layer is used to precisely sort the second image and text queue to obtain candidate push image and text content under each image and text type, and obtain the target push image and text content of the user terminal according to each candidate push image and text content; The distribution layer is used to push the target push graphic content to the user terminal.

8. A device for pushing graphic content, wherein: The device comprises: A first acquisition module is used to acquire a set of image and text types corresponding to the user terminal, and a target push model trained for each image and text type in the set of image and text types; A second acquisition module is used to acquire an initial data set, and acquire candidate data sets of each image and text type from the initial data set based on each target push model; A third acquisition module is used to acquire a first image and text queue under each image and text type from the candidate data set to obtain a second image and text queue corresponding to the image and text type set; The push module is used to obtain target push graphic content from the second graphic queue, and push the target push graphic content to the user terminal.

9. The device according to claim 8, wherein: The second acquisition module is further used for: Obtain user preference information from the user end; The initial data set is screened based on the user preference information through the target push model under each image and text type to output a candidate data set for each image and text type.

10. The device according to claim 8, wherein: The third acquisition module is further used for: For any image and text type, roughly sorting each first image and text in the candidate data set under the image and text type to obtain a sorted first image and text queue; Acquire a reference push type distribution of the user terminal according to the user preference information, and acquire a third image and text queue under each image and text type from each first image and text queue according to the reference push type distribution; The third image and text queues under each image and text type are integrated to obtain the second image and text queue corresponding to the image and text type set.

11. The device according to claim 10, wherein: The third acquisition module is further used for: Obtain the number of first images and texts corresponding to the third image and text queue; Determine the proportion of images and texts of each image and text type according to the reference push type distribution; According to the proportion of the number of images and texts and the first number of images and texts, obtaining the second number of images and texts of each type of image and text; For any image and text type, due to the order from front to back, the third image and text queue of the second number of images and texts is obtained from the first image and text queue of the image and text type.

12. The device according to claim 8, wherein: The push module is also used for: Acquire a preset first push quantity of the user terminal, and determine a second push quantity of each type of image and text according to the first push quantity and a proportion of the number of images and texts of each type of image and text; Finely sorting the second image and text queue to obtain a sorted fourth image and text queue; For any image and text type, according to the second push quantity of the image and text type, obtaining the second push quantity of candidate pushed image and text contents from the fourth image and text queue; The target pushed graphic content of the user terminal is obtained according to the candidate pushed graphic content under each graphic type.

13. A training device for a push model, wherein: The device comprises: The fourth acquisition module is used to acquire the candidate push models to be trained under each sample image and text type, and acquire the sample image and text content of the sample user terminal under each sample image and text type; A collection module, used to collect sample feedback information on sample graphic content under each sample graphic type, and sample user preferences of the sample user terminal for each sample graphic type; A generation module, for generating, for any sample image and text type, a training sample of a candidate push model corresponding to the sample image and text type according to the sample image and text content under the sample image and text type, sample feedback information corresponding to the sample image and text content, and sample user preferences corresponding to the sample image and text type; A training module is used to train a candidate push model of any sample image and text type according to the training samples of the sample image and text type until the training is completed, thereby obtaining a trained target push model for the sample image and text type, wherein the target push model is used to implement the image and text content push device described in any one of claims 8 to 12.

14. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 5 and / or claim 6.

15. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 5 and / or claim 6.

16. A computer program product, comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 5 and / or claim 6.

Images