CN110110203B

CN110110203B - Resource information pushing method, server, resource information display method and terminal

Info

Publication number: CN110110203B
Application number: CN201810027205.7A
Authority: CN
Inventors: 刘志斌
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2018-01-11
Filing date: 2018-01-11
Publication date: 2023-04-28
Anticipated expiration: 2038-01-11
Also published as: CN110110203A

Abstract

The invention discloses a resource information pushing method, a server, a resource information display method and a terminal, and belongs to the technical field of networks. The method comprises the following steps: acquiring resource information; inputting the resource information and the use preference information of the target user into a multi-stage model, and outputting preference degree information of the target user for at least one use stage of the resource, wherein the multi-stage model is used for predicting preference degree information of the target user for different use stages of the resource according to content information of the different use stages of the resource; and pushing the resource information to the target user based on the at least one preference degree information. According to the method, the multi-stage model is introduced, the process of calculating the preference degree information by the multi-stage model is consistent with the process of using the resources by the user in a multi-stage manner, and the accuracy of the calculated preference degree information is high, so that the resource information is pushed based on the preference degree information of the multi-stage model, and the accuracy of pushing the resource information can be improved.

Description

Resource information pushing method, server, resource information display method and terminal

Technical Field

The present invention relates to the field of network technologies, and in particular, to a method and a server for pushing resource information, and a method and a terminal for displaying resource information.

Background

With the development of network technology, users can use various resource information through terminals, such as reading articles, playing games, watching movies, listening to music, etc. Along with the rapid increase of the quantity of the resource information in the network, in order to help the user find the resource information of interest in the massive resource information, the server can select the resource information of interest of the user and push the resource information of interest to the terminal so that the user can read the resource information of interest on the terminal.

In carrying out the present invention, the inventors have found that the related art has at least the following problems:

at present, the actual operation behavior of a user is not considered when pushing the resource information, and the pushing accuracy is poor.

Disclosure of Invention

The embodiment of the invention provides a resource information pushing method, a server, a resource information display method and a terminal, which can solve the problem of poor accuracy of pushing articles in the related technology. The technical scheme is as follows:

in a first aspect, a method for pushing resource information is provided, where the method includes:

acquiring resource information, wherein the resource information comprises content information of a plurality of different use stages of the resource;

inputting the resource information and the use preference information of the target user into a multi-stage model, and outputting preference degree information of the target user for at least one use stage of the resource, wherein the multi-stage model is used for predicting preference degree information of the target user for different use stages of the resource according to content information of the different use stages of the resource;

And pushing the resource information to the target user based on at least one preference degree information.

In a second aspect, a method for pushing resource information is provided, where the method includes:

acquiring the use preference information of a target user according to the operation behavior of the target user on the content information of the first use stage of the history resource, wherein the content information of the first use stage is used for providing previews of the history resource;

inputting content information of a first use stage of the resource and the use preference information into a first stage model, and outputting preference degree information of the target user on the resource;

pushing the resource information based on the preference degree information;

the first stage model is used for predicting preference degree information of the user on the resource according to content information of a first use stage of the resource.

In a third aspect, a resource information display method is provided, and the method includes:

when detecting the operation behavior of the target user on the content information of at least the first use stage of the historical resources, transmitting the operation behavior to a server;

receiving content information of a first use stage of the resource sent by the server;

Displaying content information of a first use stage of the resource;

wherein the resource is determined by the server based on content information of the first usage stage of the historical resource and the operational behavior of the first usage stage.

In a fourth aspect, there is provided a resource information pushing apparatus, the apparatus including: the device comprises an acquisition module, an input and output module and a pushing module.

The resource information comprises content information of a plurality of different use stages of the resource;

the input-output module is used for inputting the resource information and the use preference information of the target user into a multi-stage model and outputting preference degree information of the target user for at least one use stage of the resource, wherein the multi-stage model is used for predicting preference degree information of the target user for different use stages of the resource according to content information of the different use stages of the resource;

and the pushing module is used for pushing the resource information to the target user based on at least one piece of preference degree information.

In a fifth aspect, there is provided a resource information pushing apparatus, the apparatus including: the device comprises an acquisition module, an input and output module and a pushing module.

The acquisition module is used for acquiring the use preference information of the target user according to the operation behavior of the target user on the content information of the first use stage of the history resource, wherein the content information of the first use stage is used for providing the preview of the history resource;

the input-output module is used for inputting the content information of the first use stage of the resource and the use preference information into a first stage model and outputting preference degree information of the target user on the resource;

the pushing module is used for pushing the resource information based on the preference degree information;

In a sixth aspect, there is provided a resource information presentation apparatus, the apparatus comprising: the device comprises a sending module, a receiving module and a display module.

The sending module is used for sending the operation behavior to the server when detecting the operation behavior of the target user on the content information of at least the first use stage of the historical resources;

the receiving module is used for receiving the content information of the first use stage of the resource sent by the server;

The display module is used for displaying the content information of the first use stage of the resource;

In a seventh aspect, a server is provided, where the server includes a processor and a memory, where at least one instruction is stored in the memory, where the instruction is loaded and executed by the processor to implement the resource information pushing method in the first aspect and/or implement the resource information pushing method in the second aspect.

In an eighth aspect, there is provided a terminal, the terminal including a processor and a memory, the memory storing at least one instruction, the instruction being loaded and executed by the processor to implement the resource information presentation method in the third aspect.

In a ninth aspect, there is provided a computer readable storage medium having stored therein at least one instruction that is loaded and executed by a processor to implement the resource information pushing method of the first aspect and/or to implement the resource information pushing method of the second aspect.

In a tenth aspect, there is provided a computer-readable storage medium having stored therein at least one instruction that is loaded and executed by a processor to implement the resource information presentation method in the above third aspect.

The technical scheme provided by the embodiment of the invention has the beneficial effects that:

according to the method, the server and the terminal provided by the embodiment of the invention, the multi-stage model is introduced, the process of calculating the preference degree information by the multi-stage model is consistent with the process of using the resources by the user in multiple stages, and the accuracy of the calculated preference degree information is high, so that the resource information is pushed based on the preference degree information of the multi-stage model, and the accuracy of pushing the resource information can be improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic illustration of an implementation environment provided by an embodiment of the present invention;

Fig. 2 is a flowchart of a method for pushing resource information according to an embodiment of the present invention;

fig. 3 is an interface schematic diagram of a multi-stage article push provided in an embodiment of the present invention;

FIG. 4 is a schematic diagram of a multi-stage model according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of a multi-stage model according to an embodiment of the present invention;

FIG. 6 is a schematic diagram of staged computation of semantic information within a multi-stage model according to an embodiment of the present invention;

FIG. 7 is a flow chart of a model training method provided by an embodiment of the present invention;

fig. 8 is a flowchart of a method for pushing resource information according to an embodiment of the present invention;

FIG. 9 is a flow chart of a model training method provided by an embodiment of the present invention;

fig. 10 is a schematic structural diagram of a resource information pushing device according to an embodiment of the present invention;

fig. 11 is a schematic structural diagram of a resource information pushing device according to an embodiment of the present invention;

fig. 12 is a schematic structural diagram of a resource information display device according to an embodiment of the present invention;

fig. 13 is a schematic structural diagram of a server according to an embodiment of the present invention;

fig. 14 is a schematic structural diagram of a terminal 1400 according to an embodiment of the present invention.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

FIG. 1 is a schematic diagram of an implementation environment provided by an embodiment of the present invention. The implementation environment includes a plurality of terminals 101 and a plurality of servers 102. The terminals 101 are connected to a plurality of servers 102 through wireless or wired networks, where the terminals 101 may be computers, smartphones, tablet computers or other electronic devices, and each server 102 may be a server, a server cluster formed by a plurality of servers, or a cloud computing service center.

During model training, the server 102 may obtain a large amount of sample resource information, such as sample resource information sent by a provider of the resource information, which may be a website or platform, a third party server, a worker, a user, or the like. Server 102 may be trained based on a large amount of sample resource information to arrive at a multi-stage model or a first-stage model. In the process of pushing the resource information, the server 102 may acquire the resource information, push the resource information to the terminal 101 of the target user based on the resource information, the usage preference information of the target user, and the multi-stage model, or push the resource information to the terminal 101 of the target user based on the resource information, the usage preference information of the target user, and the first-stage model. Optionally, the server 102 may also have a resource database, which may be an article database, a video database, a music database, a merchandise database, or a game database, etc., for storing a large number of resources, and/or a user database. The user database is used to store usage preference information for a large number of users.

The method provided by the invention can be applied to scenes for providing various resources, wherein the resources can comprise articles, videos, music, commodities, games and the like, and in various scenes, as a multi-stage model or a first-stage model with high accuracy is used in the process of pushing the resource information, the resource information can be pushed more accurately, the accuracy of the processes of recommending the resource information, searching the resource information, sorting the resource information and the like is greatly improved, the user experience is obviously improved, the public praise of products and services is improved, and the business values such as larger flow change are brought.

Taking the resource as an example, the following describes the actual technical effects of the embodiment of the present invention in combination with three exemplary application scenarios:

(1) Can be applied in the context of recommending articles. In this scenario, the terminal may run a client that is only used to provide reading functionality, at which point the server may be the server associated with the reading application. Of course, the terminal may also run a client with multiple functions, for example, the client may be a social application client, which may provide social and reading functions.

When the server acquires an article, the registered user can be used as a target user, the article and the usage preference information of the target user are input into the multi-stage model, the preference degree information of the target user for at least the first reading stage of the article is output, that is, the preference degree information of at least the first reading stage is output, for example, the preference degree information of the target user for previewing articles such as titles, thumbnails and abstracts is output, and optionally, the preference degree information of other reading stages such as the preference degree information of the second reading stage is also output. Then, according to the preference degree information of at least the first reading stage, determining the order of recommending the articles, or determining at least one of the position, the order and the occupied page proportion of the articles in the recommended pages, or determining whether to push the articles.

(2) May be applied in the context of searching for articles. In this scenario, the terminal may run a client of the search engine, with the server being the server with which the search engine is associated. When the target user inputs a keyword in the search engine and clicks the confirmation option, the terminal can send an information pushing request to the server, after the server receives the information pushing request, the server searches a plurality of articles matched with the keyword, for each article in the plurality of articles, the article and the use preference information of the target user can be input into the multi-stage model, and the preference degree information of the target user for at least the first reading stage of the article is output. And then, determining a search result according to at least one piece of preference degree information of each article, for example, acquiring preference degree information of a first reading stage of a plurality of articles, sorting the articles according to the preference degree information of the first reading stage, and determining the sequence of the articles in the search result according to the sorting of the preference degree information. The higher the preference degree information of the first reading stage of the articles is, the earlier the articles are in the order of the search results, and the article with the highest preference degree information of the first reading stage can be used as the first article in the search results. The search results for the articles may then be sent to the terminal, pushing the articles to the terminal.

(3) May be applied in a scenario where articles are ranked. In this scenario, the terminal may run a client of the reading application, and the server is a server associated with the reading application. When the target user clicks an icon of the reading application on the terminal or clicks a certain option (for example, a 'one-touch option') in the reading application, the terminal may send an information push request to the server, after receiving the information push request, the server may acquire a plurality of articles, for each article in the plurality of articles, the article and the usage preference information of the target user may be input into the multi-stage model, and the output preference degree information of the target user for at least a first reading stage of the article may be output. And then, sorting the articles according to the preference degree information of the first reading stage according to at least one preference degree information of each article, and generating a push page carrying the articles according to the preference degree information of the articles. The higher the preference degree information of the first reading stage of the article is, the earlier the order of the article in the push page can be, or the article is located in the center position in the push page, or a larger proportion of the push page is occupied. The push page may then be sent to the terminal, pushing the article to the terminal.

The above description is given by taking the scenes such as recommending articles, searching articles, and sorting articles as an example, and in practice, the present invention is not limited to this, and the present invention can be applied to the scenes such as recommending videos, searching videos, and sorting videos, and further, can be applied to the scenes where any type of resource information is pushed.

At present, for applications with pushing functions, such as various reading/information applications, shopping applications, video applications and the like, an essential core module is used for modeling based on the use preference of a user so as to push resource information preferred by the user to the user, thereby achieving the effect of accurate pushing. However, current methods of modeling based on user usage preferences have a common assumption: the user carefully views the entire content presented to him (e.g., the entire text of an article, the entire content of a video), understands it rationally and comprehensively, evaluates the entire content rationally based on personal use preferences (excluding a number of irrational factors such as emotions) and decides on the next action (e.g., clicking, collecting, forwarding, commenting, etc.).

However, taking the resource as an article as an example, in reality, the user reading in the mobile internet era has several important features: (1) Users typically read in fragmented time, and because of the short time allotted for reading, they rarely read through, but rather quickly browse/pan, capturing content fragments of their own possible interest in a very short time. (2) The user generally does not adopt a single-stage reading mode of directly reading the full text of the article, but firstly looks at the title, the map matching and the abstract to quickly judge whether the user is interested, if the user is interested, the user clicks the full text to quickly read the article, and then reviews whether the actual full text content is consistent with the previously read title, map matching and abstract, and accumulates/refreshes the cognition about the reading application of whether the user is "text-to-text". (3) With the reduction of the threshold of reading, the audience of reading is wider and wider, the same article can be read by a plurality of different users, and the social level, the cultural level, the reading and understanding ability and the like of the users are different, so that the matching degree of the content of the article and the users is not only determined by the use preference of the users, but also influenced by the personalized attribute of the users.

The resource information pushing method provided by the invention can solve the problems by introducing a multi-stage model: (1) The multi-stage model is provided with a plurality of sub-models in use stages, semantic information output by each sub-model can determine whether the next sub-model can calculate semantic information, the actual reading process of a user is described through the process of sequential calculation of each sub-model, and the system architecture of the whole multi-stage model is more consistent with the cognition and decision process of the user. (2) The sub-models of each use stage of each user are personalized and different, and combine operation time (such as time for reading articles and time for watching videos), attribute information (such as academic) and the like, thereby integrating the consideration of different reading and understanding abilities of different users. (3) Based on the content information as input, the model of each stage of the user can comprehensively consider the operation time of the user (reflecting the reading time and the reading speed of the user), the attribute information of the user (such as learning, region, age, social level and the like), irrational information (such as weather information, user emotion information, traffic information and user health information), social information (such as who is to be together, friends sharing and chat topics on a social network, communities and circles where the user is located and the like) and impression information of the user on a provider of the resource information (such as whether the situation of text is frequently happened or not), so that the accuracy of recommending the resource information is further improved. (4) The modeling of the content information of all the using stages of the resources is not directly used any more, a hierarchical attention mechanism is constructed, and a selective and staged modeling mode is adopted, namely, the modeling of the content information of the current using stage of the resources is only used, so that the model is ensured to be more fit with the process of actually using the resources (such as reading articles) by a user.

Fig. 2 is a flowchart of a method for pushing resource information according to an embodiment of the present invention. The execution body of the embodiment of the invention is a server, referring to fig. 2, the method includes:

201. the server obtains the resource information.

The resource can be an article, a movie, a game, music or commodity, each piece of resource information comprises content information of a plurality of different use stages of the resource, each resource can be divided into a plurality of use stages, and the terminal can sequentially provide the content information of each use stage of the resource stage by stage in the process of providing the resource, so that a user is guided to use the resource step by step.

For the specific process of providing content information on a phase-by-phase basis, the presentation information of each use phase may be presented based on whether the presentation information of the last use phase is presented, i.e., the presentation information of each use phase may be presented based on whether a confirmation action of the presentation information of the last use phase is detected. Specifically, the terminal will display the content information of the first usage stage first, and display the content information of the second usage stage after detecting the confirmation of the content information of the first usage stage, and will not display the content information of the second usage stage when not detecting the confirmation of the content information of the first usage stage. Similarly, when a confirmation of the content information of the second usage stage is detected, the content information of the third usage stage is displayed, and so on.

Wherein, the confirmation behavior represents that the content information of the next use stage needs to be displayed is confirmed, and the confirmation behavior comprises at least one of clicking behavior, sliding behavior, long-press behavior and voice confirmation behavior. The clicking action may be an action of clicking a certain option or a certain link, the sliding action may be an action of sliding from top to bottom, an action of sliding from left to right, or an action of sliding in other directions, the long-press action may be an action of long-pressing a certain option or a certain picture, and the voice confirmation action may be an action of issuing a voice command.

For the specific content of the content information of each usage stage, optionally, the plurality of usage stages of the resource may include a first usage stage showing at least one of a title, a thumbnail, and a summary of the resource, and a second usage stage showing the full text of the resource. Of course, the first usage stage and the second usage stage of the resource may also display other content information, and the resource may further include a third usage stage or even more usage stages.

It should be noted that, in the process of providing the resource stage by stage, the first use stage of the resource may provide a preview of the resource, by pushing the content information of the first use stage of the resource, the user may preview the resource, and decide whether to use the resource in an advanced manner according to the preview of the resource, for example, determine whether the article is interested according to the title and the thumbnail of the article, browse the entire text of the article if the article is interested, determine whether the movie is interested according to the publicity section of the movie, view the entire content of the movie if the article is interested, and so on.

Each use stage of the resource is exemplarily described below with reference to an actual application scenario and a drawing.

(1) The resource is an article. The article can be designed in the following two design modes according to content information of a plurality of reading stages.

The content information of the first reading stage of the first design article comprises at least one of the title, the thumbnail and the abstract of the article, and the content information of the second reading stage of the article is the full text of the article. The text of the article may include text, image, video, music, animation, tag, hyperlink, etc., for example, the article may be an article published by public number or an article published by a resource platform.

Referring to fig. 3, an interface schematic diagram of a multi-stage article pushing provided by the embodiment of the present invention may be used to push at least one of a title, a thumbnail, and a abstract of an article first, and then push the entire text of the article when detecting a confirmation operation of a user on at least one of the title, the thumbnail, and the abstract.

The content information of the first reading stage of the second design and the article comprises at least one of the title, the thumbnail and the abstract of the article, the content information of the second reading stage of the article is a pilot-looking fragment of the article, and the content information of the third reading stage of the article is the full text of the article.

In the scene of pushing the article, at least one item of the title, the thumbnail and the abstract of the article can be pushed first, when the confirmation operation of the user on the at least one item of the title, the thumbnail and the abstract is detected, the test fragment of the article is pushed, and when the confirmation operation of the user on the test fragment is detected, the whole text of the article is pushed.

(2) The resource is video. The video can be designed in the following two design modes according to the content information of a plurality of watching stages.

The content information of the first viewing stage of the video comprises at least one of a poster, a first frame picture, a key frame picture, a title, a abstract, an actor and a key fragment of the video, and the content information of the second viewing stage of the video is the whole content of the video.

In the scene of pushing the video, at least one of the poster, the first frame picture, the key frame picture, the title, the abstract and the key fragment of the video can be pushed first, and when the confirmation operation of the user on at least one of the poster, the first frame picture, the key frame picture, the title, the abstract and the key fragment of the video is detected, the whole content of the video is pushed.

The content information of the first viewing stage of the second video comprises at least one of a poster, a first frame picture, a key frame picture, a title, an abstract, an actor and a key fragment of the video, the content information of the second viewing stage of the video is a pilot viewing fragment of the video, and the content information of the third viewing stage of the video is the whole content of the video.

In the scene of pushing the video, at least one of the poster, the first frame picture, the key frame picture, the title, the abstract and the key fragment of the video can be pushed first, when the confirmation operation of the user on at least one of the poster, the first frame picture, the key frame picture, the title, the abstract and the key fragment of the video is detected, the trial fragment of the video is pushed, and when the confirmation operation of the user on the trial fragment of the video is detected, the whole content of the video is pushed.

(3) The resource is music.

The content information of the first listening stage of the music may include, for example, at least one of a name of the music, a poster of the music, a picture of a singer, a name of the singer, a name of a song list to which the music belongs, a name of an album to which the music belongs, an album screenshot to which the music belongs, a name of a user who uploads the music, a category of the music, a number of plays of the music, and the content information of the second listening stage of the music includes at least one of the music, lyrics of the music, comments of the music, an album video to which the music belongs.

In the scene of pushing music, the content information of the first listening stage of the music can be pushed first, and when the confirmation operation of the user on the content information of the first listening stage is detected, the content information of the second listening stage of the music is pushed.

(3) The resource is a commodity, which may include content information for multiple purchase phases.

The content information of the first purchase stage of the commodity includes at least one of a picture of the commodity, a name of the commodity, an advertisement word of the commodity, a price of the commodity, a shipping cost of the commodity, sales volume of the commodity, a number of purchasers of the commodity, an attribute of the commodity, a high-frequency evaluation of the commodity, a location of the commodity, preference information of the commodity, a distance between the commodity and a target user, a name of a seller, an after-sales service of the commodity, a brand of the commodity, and a score of the commodity. The content information of the second purchase stage of the commodity includes at least one of details of the commodity, each evaluation of the commodity, question and answer information of the commodity, matching package information of the commodity, similar commodity information of the commodity, details of the seller, pictures of the commodity, names of the commodity, advertisement words of the commodity, prices of the commodity, freight of the commodity, sales volume of the commodity, number of purchases of the commodity, attribute of the commodity, high-frequency evaluation of the commodity, location of the commodity, preferential information of the commodity, distance between the commodity and a target user, names of the seller, after-sale service of the commodity, brands of the commodity, scores of the commodity, and the content information of the third purchase stage of the commodity includes at least one of names of the purchased commodity, number of the purchased commodity, order amount, remarks to the seller, and receiving address.

In the scene of pushing the commodity, the content information of the first purchasing stage of the commodity can be pushed first, when the confirmation operation of the user on the content information of the first purchasing stage is detected, the content information of the second purchasing stage of the commodity is pushed, and when the confirmation operation of the user on the content information of the second purchasing stage is detected, the content information of the third purchasing stage of the commodity is pushed.

(4) The resource is a game that may include content information for multiple phases of operation.

The content information of the first operation stage of the game includes at least one of a name of the game, an icon of the game, a size of the game, an advertisement word of the game, and a number of downloading persons of the game. The content information of the second operational stage of the game includes at least one of an introduction to the game, an animation of the game, a promotional video of the game, a size of the game, advertising words of the game, comments of the game, scoring of the game, a version of the game, a label of the game, a similar game of the game. The content information of the third operation stage of the game includes at least one of a combat screen, a competition screen, an administration screen, a adventure screen, a dialogue screen, a card drawing screen, a shooting screen, a role playing screen, a prop selection screen, a prop purchase screen, and a chess and card screen.

In the scene of pushing game, the content information of the first operation stage of the game can be pushed first, when the confirmation operation of the user on the content information of the first operation stage is detected, the content information of the second operation stage of the commodity is pushed, and when the confirmation operation of the user on the content information of the second operation stage is detected, the content information of the third operation stage of the commodity is pushed.

It should be noted that, when the method is applied to a scenario of actively pushing resource information, for example, in a recommended scenario, when a server obtains a plurality of resource information, it is possible to directly use each obtained resource information as resource information that can be pushed without obtaining information input by a target user, and then input each obtained resource information into a multi-stage model. When the method is applied to a scene of passive pushing of resource information, such as a search scene, a user can input keywords on a terminal, the terminal can send the keywords to a server, the server can search based on the keywords to obtain a plurality of resource information matched with the keywords, and then the resource information matched with the keywords is input into a multi-stage model.

202. The server obtains the use preference information of the target user according to the operation behavior of the target user on the content information of at least the first use stage of the historical resources.

The target user refers to a pushing object of the resource information, and the process of determining the target user can comprise the following two designs:

firstly, determining a target user according to an information push request of a terminal: the terminal may send an information push request to the server, where the information push request carries a user identifier, and the information push request is used to request the server to push resource information. When the server receives the information push request, the user identifier carried by the information push request can be obtained, and the user corresponding to the user identifier is taken as a target user. The information push request may be triggered by a start operation of an application such as a reading application or a shopping application, a click operation of a search option, or a start operation of a push page, which is not limited in this embodiment.

Design two, confirm the target user according to registered user: the server may use each registered user as a target user, or may select a user meeting a preset condition from each registered user, and use the user meeting the preset condition as the target user. The preset condition may be member recharging, payment triggering, push service subscription, location in a geographic area, having a certain attribute or belonging to a certain social group, etc., and may be determined according to actual requirements.

It should be noted that the above illustrates two ways of determining the target user, and the specific way of determining the target user is not limited in this embodiment, and in implementation, each user served by the server may be the target user. In addition, the number of the target users determined by the server may be one or more, and the number of the target users determined by the server is not limited in this embodiment.

Use preference information: the usage preference information is used to indicate usage preferences of the user for the resource information, such as a preference for reading articles, a preference for watching movies, a preference for listening to music, a preference for operating games, a preference for purchasing goods, and the like. The usage preference information may also be referred to as interest preference information, personal preference information, etc., and the usage preference information may be represented in a data structure of vectors.

Specific procedure for constructing usage preference information: the server may obtain the usage preference information of the target user according to the operation behavior of the target user on the content information of at least the first usage stage of the history resource, that is, the server may obtain the usage preference information of the target user only according to the operation behavior of the target user on the content information of the first usage stage of the history resource, optionally, may also obtain the usage preference information of the target user in combination with the operation behavior of the target user on the content information of the first usage stage of the history resource and the operation behavior of the content information of other usage stages, for example, may obtain the usage preference information according to the operation behavior of the first usage stage and the operation behavior of the second usage stage, and may obtain the usage preference information according to the operation behaviors of all the usage stages, for example.

For the specific process of detecting the operation behavior of the target user on the content information of the at least first use stage of the history resource, the server can detect the operation behavior of the target user on the content information of the at least first use stage of the history resource by interacting with the client.

In the implementation, the server pushes the content information of the at least first use stage of the history resource to the client in advance, the client receives the content information of the at least first use stage of the history resource, displays the content information of the at least first use stage of the history resource, and when detecting the operation behavior of the target user on the content information of the at least first use stage of the history resource, sends the operation behavior to the server, the server can receive the operation behavior of the content information of the at least first use stage of the history resource sent by the client, so as to detect the operation behavior of the target user on the content information of the at least first use stage of the history resource.

In combination with the actual application scenario, the process of detecting the operation behavior of the target user on the content information of at least the first use stage of the historical resource by the server may specifically include the following design one to design six:

Designing one (aiming at articles) and acquiring duration time of browsing behavior when detecting the browsing behavior of the content information of any using stage of the historical resources by the target user.

In the process of browsing the content information of any use stage of the historical resources by the target user, the server can detect the browsing behavior of the target user and acquire the duration of the browsing behavior so as to acquire the use preference information of the target user according to the duration of the browsing behavior. For a specific process of acquiring the duration of the browsing behavior, the terminal may detect the duration of the browsing behavior of the target user on the resource, send the duration of the browsing behavior of the target user on the resource to the server, and the server may receive the duration of the browsing behavior, thereby acquiring the duration of the browsing behavior.

Regarding the process of detecting the duration of the browsing behavior of each usage stage by the terminal, the following two designs can be adopted:

designing one (aiming at the first using stage), in the process of pushing the content information of the first using stage of the historical resources, the terminal usually displays the content information of the first using stage of a plurality of historical resources on the same page, for example, displaying titles and thumbnails of a plurality of articles on a main interface of a reading application, displaying names and pictures of a plurality of commodities on a searching result page of the commodity, and the like, so that a user previews the plurality of historical resources in the page, and selects the interested resource from the plurality of historical resources to view the content information of the next using stage of the resource.

In order to determine the duration of browsing the content information of the first use stage of each historical resource by the target user, in the process that the terminal displays each content information in any page, the terminal may detect sliding operations triggered on the page, consider the time interval between the sliding operations as the duration of stay of the user's eyes on the page, and may consider the duration of stay on the page as the duration of browsing behavior of the content information of the first use stage of each resource in the page. In addition, the number of the content information in the page can be obtained, the ratio between the stay time length in the page and the number of the content information is calculated, and the ratio is used as the duration time length of the browsing action of each content information in the page.

For example, assuming that a screen page currently displayed by the terminal includes titles and thumbnails of 4 articles, when the terminal detects a sliding operation after 10s after starting to display the page, and scrolls to display the next screen page, the terminal may determine that the duration of stay in the screen page is 10s, and use 10s as the reading duration of the titles and thumbnails of each article of the 4 articles in the screen page. In addition, 10 s/4=2.5 s may be calculated, and 2.5s may be used as the reading time length of the title and the thumbnail of each of the 4 articles.

Design two, for the second or later use stage, for any use stage other than the first use stage of the history resource, the full page will typically be used to carry the content information of the history resource in that use stage, but no other history resource is involved. Therefore, in the process of displaying the page corresponding to the content information of any use stage except the first use stage of the historical resource, the terminal can record the current time point when the page is displayed, record the current time point again when the page is displayed (such as closing the page or switching to other pages) is displayed, calculate the time interval of the two recorded time points, and serve as the duration of the browsing behavior of the content information of the historical resource in the use stage.

Optionally, for content information in any use stage of any historical resource, if the duration of the browsing action of the content information is longer, for example, exceeds a preset duration threshold, which indicates that the target user is interested in the content information, the content information in the use stage may be obtained, and further, the usage preference information may be obtained according to the content information in the use stage, and if the duration of the browsing action of the content information is shorter, for example, does not exceed the preset duration threshold, which indicates that the target user is not interested in the content information, the usage preference information may not be obtained according to the content information in the use stage.

Wherein, for the specific process of acquiring the usage preference information according to the content information of the history resource, the server may employ a feature extraction algorithm to extract the features of the content information, and use the features of the content information as the usage preference information. The feature extraction algorithm may be any combination of DNN (Deep Neural Network, deep neural network algorithm), RNN (Recurrent neural Network ), TF-IDF (term frequency-reverse file frequency), word2vec, doc2 vec. Taking historical resources as an article as an example, the server may employ a feature extraction algorithm to extract keywords from content information of the article, and use the keywords as the usage preference information.

Design two (for audio class resources), when the listening behavior of the target user to the content information of any one of the usage phases of the history resources is detected, the duration of the listening behavior can be acquired.

In the process of the target user listening to the content information of any one of the use phases of the history resource, the server may acquire the duration of the listening behavior so as to acquire the use preference information of the target user according to the duration of the listening behavior, thereby predicting the use preference of the target user for the resource based on the use preference information. The terminal can acquire the duration of the playing of the historical resources as the duration of the listening behavior, and send the duration to the server. Specifically, when the terminal finishes playing the historical resources, the duration indicated by the progress bar of the player of the historical resources may be obtained, and the duration is taken as the duration of playing the historical resources, or the terminal may record the current time point when starting to play the historical resources, record the current time point again when finishing playing the historical resources, and record the time interval between the two time points as the duration of playing the historical resources.

Optionally, considering that the duration of the listening behavior of different historical resources is different, in order to ensure that the duration of the listening behavior of different historical resources is within a similar numerical range, the duration of the listening behavior may be normalized according to the duration of the historical resources, and a ratio of the duration of the listening behavior to the duration of the historical resources is used as the duration of the listening behavior finally obtained.

Design III (for video class resource), when the watching behavior of the content information of any using stage of the history resource by the target user is detected, the duration of the watching behavior is obtained.

The design is the same as the above design, and will not be described here again.

Designing a game behavior of content information of any use stage of the history resource by the target user, and acquiring at least one of duration, operation times and operation frequency of the game behavior when the game behavior is detected.

In the process that the target user triggers the game behavior on the game resource, the server can start timing when the target user starts the game resource, and finish timing when the target user exits the game resource, so as to obtain the recorded duration as the duration of the game behavior. In addition, the server can acquire the times of the operation behaviors such as the clicking times, the sliding times, the long pressing times and the like of the target user in the game process, and the operation times are obtained. In addition, the server may calculate a ratio of the number of operations to the duration as the operation frequency.

Design five (for arbitrary class resources), when the confirmation behavior of the target user on the content information of any use stage of the history resources is detected, record the confirmation behavior.

The confirmation behavior refers to a behavior of confirming entering a next use stage, and the confirmation behavior may include at least one of a click behavior, a slide behavior, a long press behavior, and a voice confirmation behavior. For example, for a text resource, the confirmation behavior may be a behavior of clicking on a title or a thumbnail of an article, for an audio resource or a video resource, the confirmation behavior may be a behavior of clicking on a start button of a player, for a game resource, the confirmation behavior may be a behavior of clicking on a play button, for a commodity resource, the confirmation behavior may be a behavior of clicking on a commodity picture, a behavior of clicking on a purchase option, and the like.

When the target user triggers a confirmation action on the content information, such as clicking on the title of an article, during the process of browsing the content information of any use stage of the history resource by the target user, the server can detect the confirmation action and record the confirmation action, which indicates that the target user wants to use the content information of the next use stage. The terminal can detect the confirmation of the target user on the content information, the confirmation of the target user on the content information is sent to the server, and the server receives the confirmation of the content information, so that the confirmation is recorded.

It should be noted that, the fifth design may be used alone as a process for detecting the operation behavior of the target user on the content information of at least the first usage stage of the history resource, or any one of the first design to the fourth design may be used in combination as a process for detecting the operation behavior of the target user on the content information of at least the first usage stage of the history resource, for example, when the fifth design and the first design are combined, the duration of the browsing behavior may be obtained during the process of the target user browsing the content information of any one of the usage stages of the history resource, and the confirmation behavior of the target user on the content information may be recorded.

Design six (for arbitrary class resource), when detecting the interactive behavior of the target user to the content information of any using stage of the historical resource, record the interactive behavior.

The interactive behavior refers to the behavior that a target user interacts with other users or providers of resource information. The other users may be friends of the target user or others who use the resource information, and the provider of the resource information may refer to an author of the resource, a user or manufacturer who issues the resource, and the like.

Illustratively, the interactive behavior may include at least one of a praise behavior, a comment behavior, a share behavior, a forward behavior, a collection behavior, a rewarding behavior, a shopping cart joining behavior, a privately believable behavior, a search behavior, a download behavior, a save behavior, a copy behavior, a K song behavior, a scoring behavior, a bullet screen sending behavior, a gift sending behavior, a scanning graphic code behavior, a purchase behavior, a voting behavior, a subscription behavior, a top-setting behavior, a stepping behavior, a reporting behavior, a personal information input behavior.

Wherein the order action can point to the operation of ordering options, the comment action can point to the action of inputting comments in a comment input box and a comment input page, the order action can point to the action of clicking the order options and paying a certain order amount, the shopping cart adding action can point to the action of adding resources to a virtual shopping cart of a target user, the private information action can point to the action of sending information to an author or a publisher of the resources, the searching action can point to the action of searching in the resources based on keywords, the downloading action can point to the action of downloading the resources to the local, the storing action can point to the action of storing the resources to a certain catalog or a certain account, the copying action can point to the action of copying fragments or all of the resources, the K song action can point to the action of clicking the K song options and singing songs, the scoring behavior may refer to the behavior of inputting a score to a resource, the transmitting barrage behavior may refer to the behavior of inputting a barrage to a resource, the posting behavior may refer to the behavior of transmitting a virtual gift to a host corresponding to the resource, the scanning graphic code behavior may refer to the behavior of scanning a graphic code carried by the resource, the purchasing behavior may refer to the behavior of purchasing the resource online, the voting behavior may refer to the behavior of clicking a plurality of voting options provided by the resource, the subscribing behavior may refer to the behavior of subscribing to a publisher of the resource (such as a blogger focusing on a microblog, subscribing a public number) so as to automatically receive the resource published by the publisher, the topping behavior may refer to the behavior of keeping a certain resource at the top of a page of an application program for display, the stepping behavior may refer to the behavior of clicking the stepping options, the reporting behavior may refer to the behavior of reporting that the resource contains bad information, the personal information input behavior may refer to a behavior of inputting personal information in input options provided by a resource (e.g., a questionnaire).

The first point is that the sixth design may be a process of detecting the operation behavior of the target user with respect to the content information in at least the first usage stage of the history resource alone, a process of detecting the operation behavior of the target user with respect to the content information in at least the first usage stage of the history resource in combination with any one of the first to fourth designs, or a process of detecting the operation behavior of the target user with respect to the content information in at least the first usage stage of the history resource in combination with the fifth design. For example, when the design six is combined with the design one, in the process of browsing the content information of any use stage of the history resource by the target user, the duration of the browsing behavior can be obtained, and the interaction behavior of the target user on the content information can be recorded. For another example, when the design six and the design three are combined, in the process of viewing the content information of any use stage of the history resource by the target user, the duration of the viewing behavior can be obtained, and the interaction behavior of the target user on the content information can be recorded.

The second point to be described is that, for the specific process of recording the interactive behavior, the server can record whether the interactive behavior occurs or not, and also can record the interactive content carried by the interactive behavior, such as the comment content, the amount of appreciation, and the like. The following describes an exemplary procedure of recording interactive contents in interactive behavior by (1) to (7).

(1) When the comment behavior of the target user on the content information of any use stage of the history resource is detected, the content of the comment is recorded.

The target user can input comment content aiming at content information of any use stage of the historical resources, the comment content can comprise characters, expressions, pictures and the like, and the server can record the comment content so as to acquire the use preference information of the target user according to the comment content. The terminal can detect comment operation of a target user, acquire comment content input by the user, send the comment content to the server, and the server can receive the comment content so as to obtain the comment content.

Optionally, for the process of obtaining the usage preference information according to the content of the comment, the server may perform emotion analysis on the content of the comment, when the content of the comment expresses positive emotion, it indicates that the target user is interested in the content information, may obtain the content information of the usage stage, further obtain the usage preference information according to the content information of the usage stage, and when the content of the comment expresses negative emotion, it indicates that the target user is not interested in the content information, and may not obtain the content information of the usage stage.

For example, when the target user reviews the content information of a certain use stage of an article, and supposing that the content of the review is "praise and call" for the author, the server may analyze that the content of the review expresses forward emotion, and confirm that the target user is interested in the content information of the use stage, then acquire the use preference information according to the content information of the use stage. If the content of the comment is' small plaited and deceived, and if the comment is broken, the server can analyze that the content of the comment expresses negative emotion, and confirm that the target user is not interested in the content information of the using stage, the server does not acquire the use preference information according to the content information of the using stage.

(2) And when the sharing behavior of the target user on the content information of any use stage of the historical resources is detected, the sharing times or the number of the shared objects is recorded.

In the process of using the content information of any use stage of the historical resources, the target user can share the content information of any use stage of the historical resources to a certain object, the object can be a certain friend, a certain social group or a certain application, and the server can record the sharing times, namely the times of the target user to share the content information of the historical resources, so that the use preference information of the target user can be obtained according to the sharing times. Similarly, the server may record the number of objects to be shared, where the number of objects to be shared may be equal to the number of times of sharing when the content information is shared to one object at a time, and the number of objects to be shared may be accumulated each time when the content information is shared to a plurality of objects at a time, to obtain the total number of objects to be shared. The terminal can notify the server of the sharing object when detecting the sharing operation of the target user on the content information, and the server can detect the sharing action of the target user on the content information.

Optionally, for the process of acquiring the usage preference information according to the sharing behavior, when the number of times that the target user shares a certain content information of a certain historical resource is greater, for example, the number of sharing times is greater than a certain threshold, the server may learn that the target user is interested in the content information, and may acquire the usage preference information according to the content information.

(3) When a scoring action of the target user on content information of any one of the use stages of the history resource is detected, a score of the scoring action is recorded.

The target user may score content information of any use stage of the history resource, and the server may record the scored score, so as to obtain the use preference information of the target user according to the scored score. The terminal can detect scoring operation of the target user, acquire the score input by the user, send the score to the server, and the server can receive the score, so that the score of scoring behavior is obtained.

Alternatively, for the process of acquiring the usage preference information according to the scoring behavior, when the target user scores a certain content information of a certain historical resource higher, for example, the score is greater than a certain score threshold, the server may learn that the target user is interested in the content information, and may acquire the usage preference information according to the content information.

(4) When a reward act of the target user on content information of any one of the use stages of the history resource is detected, an amount of the reward act is recorded.

The target user may be rewarded for content information of any one of the usage stages of the history resource, and the server may record the amount of the rewarding so as to acquire the usage preference information of the target user according to the amount of the rewarding. The terminal can detect the rewarding operation of the target user, acquire the rewarding amount of the user, send the rewarding amount to the server, and the server can receive the rewarding amount, so that the rewarding amount is obtained.

Alternatively, for the process of acquiring the usage preference information according to the reward behavior, when the target user rewards a certain content information of a certain historical resource, the server may learn that the target user is interested in the content information, and may acquire the usage preference information according to the content information.

It should be noted that, considering different economic conditions of different users, the amount of the appreciation can be normalized according to the personal economic conditions of the users, so as to ensure that the usage preference information of different users is in a similar numerical range.

(5) When the gift sending behavior of the content information of any use stage of the history resource by the target user is detected, at least one of the number, the category and the amount of the transmitted virtual gift is recorded.

The target user can give gifts according to the content information of any use stage of the historical resources, and the server can record the number, the type and the amount of the virtual gift sent by the target user so as to acquire the use preference information of the target user according to the number, the type and the amount of the virtual gift. The terminal can detect gift sending operation of a target user, acquire the number, the type and the amount of the virtual gift input by the user, send the virtual gift to the server, and receive the number, the type and the amount of the virtual gift by the server, so that the number, the type and the amount of the virtual gift are obtained.

(6) When the purchasing behavior of the content information of any one of the use stages of the history resource by the target user is detected, at least one of the number and the amount of the purchasing behavior is recorded.

When the historical resources are commodities, the target user can perform purchasing operation aiming at content information of any using stage of the historical resources, and the server can record at least one of the quantity and the amount of the commodities purchased by the target user so as to acquire the use preference information of the target user according to the quantity and the amount of the commodities purchased. The terminal can detect the purchase operation of the target user, acquire at least one of the number and the amount of the commodity purchased by the user, send the at least one of the number and the amount of the commodity purchased to the server, and receive the at least one of the number and the amount of the commodity purchased by the server, so that the at least one of the number and the amount of the commodity purchased is obtained.

(7) When the bullet screen transmitting behavior of the content information of any use stage of the history resource by the target user is detected, the content of the transmitted bullet screen is recorded.

The target user can send the barrage aiming at the content information of any use stage of the historical resources, and the server can record the content of the barrage so as to acquire the use preference information of the target user according to the content of the barrage. The terminal can detect the barrage sending operation of the target user, acquire the barrage content input by the user, send the barrage content to the server, and the server can receive the barrage content, so that the barrage content is obtained.

Optionally, for the process of acquiring the usage preference information according to the content of the barrage, the server may perform emotion analysis on the content of the barrage, when the content of the barrage expresses positive emotion, it indicates that the target user is interested in the content information, may acquire the content information of the usage stage, further acquire the usage preference information, and when the content of the comment expresses negative emotion, it indicates that the target user is not interested in the content information, and may not acquire the content information of the usage stage.

The first point to be noted is that (1) to (7) above may form a process of designing six or recording interactive behaviors by any combination, for example, when (2) and (4) are combined, if the target user triggers both the sharing behavior and the viewing behavior on the content information, the sharing behavior and the viewing behavior can be detected, and the number of times of sharing and the amount of viewing can be recorded for the content information at any use stage of the history resource.

The second point to be described is that, when the above-described designs one to six are combined with each other to form a process of detecting the operation behavior of the target user on the content information of at least the first use stage of the history resource, when the user's use preference information is constructed, the weighted sum may be performed for one resource based on at least one operation behavior recorded by the resource, the result of the weighted sum may be regarded as the user's use preference information for the resource, and the user's use preference information may be generated for one user based on the user's use preference information for a plurality of resources and the content information of the resource.

For example, when the design I and the design II are combined, the duration of the browsing behavior can be obtained, the interaction behavior can be recorded, and the duration of the browsing behavior and the interaction behavior are weighted and summed to obtain the use preference information of the resource. Different weights can be determined in advance for different operation behaviors, and the weight of each operation behavior is determined according to actual service requirements, for example, different weights can be determined for duration of browsing behaviors and interaction behaviors.

The third point to be described is that the above description is given by taking the use preference information obtained according to the operation behavior of at least the first use stage as an example, and alternatively, the use preference information of the target user may be obtained according to the operation behavior of at least the first use stage, the social information, and the attribute information of the target user. The attribute information of the target user may include at least one of an academic, a region, an age, a social hierarchy, a gender, a work type/property, and a financial/liability status of the target user, and the social information includes at least one of a friend identifier of the target user, a friend sharing topic and a chat topic of the target user, a social group in which the user is located, and a social circle.

The fourth point to be described is that the server may construct the usage preference information in advance or may construct the usage preference information in real time. Specifically, for a scenario in which the usage preference information is pre-built, the server may pre-build the usage preference information of each user, store the usage preference information of each user in the user database, and query the user database to obtain the pre-built usage preference information after determining the target user. Further, in consideration of timeliness of interest preference, the usage preference information of each user may be updated periodically according to the operation behavior of each user in the last period of time. For example, the server may update the user's usage preference information according to the user's operational behavior of content information for at least the first usage stage of each historical resource for the present month every month. For the scene of constructing the usage preference information in real time, the server can store the historical resources and the operation behaviors of the target user for each usage stage of the historical resources in the user database, and after the target user is determined, the historical resources used by the target user and the operation behaviors of the target user for each usage stage of the historical resources are read from the user database, and the usage preference information is constructed according to the historical usage resource information of the target user and the operation behaviors of each usage stage.

203. The server inputs the resource information and the usage preference information of the target user into the multi-stage model, and outputs preference degree information of the target user for at least one usage stage of the resource.

In this embodiment, in combination with the multi-stage usage of resource information, a multi-stage model is designed accordingly, and the multi-stage model includes a plurality of sub-models of usage stages, and each sub-model can be understood as a classifier. The multi-stage model is illustrated from the two angles (1) and (2), respectively:

(1) Function of each sub-model: each sub-model is used for outputting semantic information corresponding to the use stage according to the content information corresponding to the use stage. That is, the first sub-model is used to output semantic information of the first usage stage based on the content information of the first usage stage, the second sub-model is used to output semantic information of the second usage stage based on the content information of the first usage stage, and so on. Wherein the semantic information may be represented in a data structure of vectors, the semantic information and the usage preference information may be in the same potential semantic space (latent semantic space).

(2) Relationship between multiple sub-models: when the semantic information output by the sub-model of each use stage is matched with the use preference information, the sub-model of the next use stage is triggered to calculate the semantic information, and when the semantic information output by the sub-model of each use stage is matched with the non-use preference information, the sub-model of the next use stage cannot calculate the semantic information. That is, when the semantic information output by the first sub-model is matched with the usage preference information, the sub-model of the second usage stage is triggered to calculate the semantic information, and when the semantic information output by the first sub-model is not matched with the usage preference information, the sub-model of the second usage stage cannot calculate the semantic information. Similarly, when the semantic information output by the second sub-model matches the usage preference information, the sub-model in the third usage stage is triggered to calculate the semantic information, and so on.

It should be noted that, the multiple sub-models may be parallel to each other, that is, the semantic information output by the previous sub-model may not be input into the next sub-model, but only be used as a trigger condition of the next sub-model, that is, it is determined whether the next sub-model will calculate the semantic information. Illustratively, taking a resource as an article, a multi-stage model including two sub-models of the usage stage as an example, the structure of the multi-stage model may be as shown in fig. 4.

Through the architecture and the calculation flow of the multi-stage model, the process of using resource information by the target user stage by stage can be described through the process of sequentially calculating semantic information by using sub-models of each using stage, so that the multi-stage model is ensured to coincide with multi-stage operation behaviors, and further the accuracy can be greatly improved when information recommendation is carried out according to the multi-stage model.

In the multi-stage operation, the user is interested in the content information of the current use stage of the resource, the confirmation operation is triggered on the content information of the current use stage, and the terminal displays the content information of the next use stage so that the user can use the content information of the next use stage. Taking an article as an example, in the first reading stage, the user reads the title and the thumbnail of the article, and is interested in the title and the thumbnail of the article, and clicks the title and the thumbnail of the article to enter the second reading stage to read the whole text of the article.

Correspondingly, in the multi-stage model, after the sub-model of the current use stage outputs semantic information, when the preference degree information of the current use stage accords with a preset condition according to the semantic information and the use preference information, the sub-model of the next use stage is triggered to calculate the semantic information. Taking an article as an example, in a first reading stage, the first sub-model outputs semantic information according to content information, and when determining that preference degree information of the first reading stage accords with a preset condition according to the semantic information and reading preference information, the sub-model of a second reading stage is triggered to calculate the semantic information. The preset condition can be determined according to actual requirements, when the preference degree information is similarity, the preset condition can be that the similarity reaches the preset similarity, and when the pushing precision is required to be higher, the preset similarity can be set to be higher.

For the process of calculating the preference degree information in the multi-stage model, the server may input the content information of the first use stage into the sub-model of the first use stage, output the semantic information of the first use stage, calculate and output the preference degree information of the first use stage according to the semantic information and the preference information, and when the preference degree information meets the preset condition, input the content information of the second use stage into the sub-model of the second use stage so as to calculate and output the preference degree information of the second use stage, and so on. That is, whenever the preference degree information of the current use stage meets the preset condition, the content information of the next use stage is input into the next sub-model. In particular, for the sub-model of the current usage stage, the process of calculating the semantic information may specifically include the following steps one to three.

And step one, inputting the content information of the current use stage into the sub-model of the current use stage in the multi-stage model, and outputting the semantic information of the current use stage.

Aiming at the concrete process of calculating the semantic information in the sub-model, the sub-model can be an algorithm of various deep neural network types such as a recurrent neural network, a circulating neural network and a convolution neural network, and the sub-model can analyze the semantics of the display information corresponding to the use stage to obtain the semantic information. Taking a submodel corresponding to the whole text of the article as an example, when the submodel is a recurrent neural network, the semantics of each sentence can be gradually synthesized from each word in the article, and then the semantics of the whole sentence and the whole text can be obtained. When the submodel is a cyclic neural network, each word in the article can be input one by one in a cyclic mode, a hidden layer is maintained, all the above information is reserved in the hidden layer, and the hidden layer of the last word represents the semantic of the whole text by circularly calculating from the first word to the last word of the article. When the sub-model is a convolutional neural network, the local information of each part of the article can be modeled, and full-text semantics can be integrated from each local information through a pooling layer of the convolutional neural network.

Calculating preference degree information of the current use stage according to the semantic information and the use preference information of the current use stage, and outputting the preference degree information of the current use stage.

After the semantic information output by the submodel of the current use stage is obtained, the preference degree information of the current use stage can be calculated and output according to the semantic information and the use preference information, wherein the preference degree information is used for indicating the preference degree of the target user on the resource information, and the preference degree information is determined by the display information and the use preference information of the target user. Wherein the semantic information and the usage preference information may be both represented as vectors of potential semantic space (latent semantic space), and the server may calculate a similarity between the vector of semantic information and the vector of usage preference information, such as cosine similarity, inner product similarity, euclidean distance, etc., and use the similarity between the two vectors as preference degree information between the two information.

And thirdly, when the preference degree information of the current use stage accords with a preset condition, inputting the content information of the next use stage into the submodel of the next use stage.

The server can judge whether the preference degree information of the current use stage accords with the preset condition, when the preference degree information accords with the preset condition, the target user is interested in the content information of the current use stage, the content information of the current use stage is clicked, the terminal can display the content information of the next use stage, so that the server can input the content information of the next use stage into the submodel of the next use stage, the semantic information of the next use stage is output by the next submodel, and the like.

When the preference degree information of the current use stage does not accord with the preset condition, the target user is not interested in the content information of the current use stage, the content information of the current use stage is not clicked, but the content information of the resource information is directly ignored, the content information of the next resource information is continuously checked, and therefore the server does not input the content information of the next use stage into the submodel of the next use stage, but ends the decision flow, and then judges the content information of each use stage of the next resource information.

And combining the calculation flow, and triggering the sub-model calculation semantic information of the next use stage when the preference degree information of the current use stage meets the preset condition. When the preference degree information of the current use stage does not accord with the preset condition, the sub-model calculation semantic information of the next use stage is not triggered, and each sub-model calculation semantic information after the next sub-model is not further triggered. For example, assuming that the multi-stage model includes sub-models of four use stages, in which preference degree information of a first use stage does not meet a preset condition, none of the sub-models of the second to fourth use stages calculates semantic information.

Thus, for any one resource information, the number of semantic information output by the multi-stage model may be less than or equal to the number of usage stages. When part of the sub-models are not calculated and output, the number of semantic information is smaller than the number of using stages, and the content information of the using stages is not interested when the target user is in a certain using stage. When each sub-model is computed and output, the number of semantic information will be equal to the number of usage phases, indicating that the content information of each usage phase is of interest to the target user until the last usage phase.

For example, for an article, assume that the article includes two reading phases, and when the first sub-model performs computation and outputs semantic information, and the second sub-model does not perform computation and output semantic information, the number of semantic information is 1, which is smaller than the number of using phases (i.e., 2), indicating that the target user is not interested in the content information of the first reading phase. And when the first sub-model and the second sub-model are both calculated and semantic information is output, the content information of the first reading stage is indicated to be interested by the target user.

Optionally, considering that social levels, cultural levels, reading understanding ability, consumption ability and other aspects of different users are different, in the process of obtaining the semantic information, not only the use preference of the target user can be considered, but also the browsing time of the target user, the attribute/feature of the target user, weather, emotion, traffic condition, physical health degree, who the target user is to stay with, chat topics participated by the target user on a social network, communities and circles where the target user is located, impressions/cognition of the target user on a provider of a resource service and the like can be comprehensively considered, so that each sub-model can be used for calculating the semantic information of the corresponding use stage according to the content information and the use additional information of the corresponding use stage when the preference degree information of the last use stage accords with preset conditions. The usage additional information comprises at least one of operation duration corresponding to the usage stage, attribute information of the target user, irrational information, social information and impression information.

The operation duration of the corresponding use phase: i.e. the duration of the operational behaviour of the target user on the content information of the corresponding usage phase of the resource. Taking the resource as an article as an example, the operation duration of each reading stage of the article is the duration of the browsing behavior of the target user on the content information of the corresponding reading stage of the article. The process of acquiring the operation duration of each use stage of the resource is detailed in the process of acquiring the use preference information, which is not described herein.

For example, the reading duration of the first reading phase of the article may be a duration that the target user continues when reading the title and the thumbnail, and the reading duration of the second reading phase may be a duration that the target user continues when reading the full text. Taking the resource as a video as an example, the viewing duration of the first viewing stage of the video may be the duration of the target user when viewing the advertisement segment of the video, the viewing duration of the second viewing stage of the video may be the duration of the target user when viewing the trial segment of the video, and the viewing duration of the third viewing stage of the video may be the duration of the target user when viewing the entire content of the video. Taking the resource as music as an example, the listening duration of the first listening stage of the music may be the duration of the target user when browsing the name of the music and the poster of the music, and the listening duration of the second listening stage of the music may be the duration of the target user when listening to the music piece. Taking the resource as an example of the commodity, the purchase duration of the first purchase stage of the commodity can be the duration of the target user when browsing the picture of the commodity, the name of the commodity and the price of the commodity, and the purchase duration of the second purchase stage of the commodity can be the duration of the target user when browsing the details of the commodity and each evaluation of the commodity. Taking the resource as an example of the game, the operation duration of the first operation stage of the game can be the duration of the target browsing the name of the game and the advertisement word of the game, the operation duration of the second operation stage of the game can be the duration of the target user browsing the introduction of the game and the comment of the game, and the operation duration of the third operation stage of the game can be the duration of the target user when operating the game.

Attribute information of the target user: including at least one of the user's academic, regional, age, social hierarchy, gender, job type/nature, financial/liability status, the attribute information of the target user may have an association and influence on the user's reading comprehensiveness and reading speed.

Irrational information: the information processing system comprises at least one of user emotion information, user health information, weather information, traffic information, date event information and labor state information, wherein the labor state information comprises urgent state and/or leisure state, and the irrational information can influence the operation behavior of a user.

Social information: the social information can cause social influence (or external influence or environmental influence) except personal willingness of the target user on the operation behavior of the target user, wherein the social information comprises at least one of friend identification of the target user, friend sharing topics and chat topics participated by the target user, social groups of the user and social circles.

Impression information: an impression indicating a degree of matching between the content information of the corresponding usage stage of the provider of the resource information and the content information of the next usage stage. Taking the resource as an article, the corresponding use stage as a first reading stage, the next use stage as a second reading stage as an example, the impression information can be whether the reading application frequently generates a new and innovative title to cheat the user about clicking or not. The impression information affects the behavior of the target user to decide whether to click on the content information in the current use stage, so in this embodiment, the impression information is also used as an input parameter of the reading sub-model in the current use stage, and affects the process of calculating the semantic information by the reading sub-model according to the content information.

In this embodiment, referring to fig. 5, which shows a schematic structural diagram of a multi-stage model provided by the embodiment of the present invention, a sub-model may be designed by using a wide & deep model (wide depth model) architecture in the multi-stage model. The linear model of the wide & deep model is usually used to fuse some low-dimensional/sparse features, and may also be artificially designed and constructed features already accumulated in the service, and the deep neural network model is usually used to extract features from unstructured information such as text, images, videos, music, and the like. Accordingly, the usage additional information may be calculated by the linear model pair in the sub-model, and the content information may be calculated by the deep neural network model in the sub-model.

In particular, in any sub-model of the multi-stage model using stages, a linear model may be applied to the output of the deep neural network model for transforming the data output by the deep neural network model into semantic information. Referring to fig. 6, which shows a schematic diagram of calculating semantic information stepwise in a multi-stage model, during calculation inside a sub-model, content information of a current usage stage may be input into a deep neural network model of the sub-model of the current usage stage, and a vector corresponding to the content information is output, where the vector of the content information may be understood as a vector representation of the content information in a latent semantic space. And then, inputting the vector corresponding to the content information and the additional information of the current use stage into a linear model of the sub-model of the current use stage, and outputting semantic information of the current use stage. The deep neural network model can be obtained through training according to content information corresponding to the using stage, and the linear model can be obtained through training according to additional information corresponding to the using stage.

In combination with the concept of the use stage and the content information in this embodiment, this calculation mode may be understood as a hierarchical attention mechanism (attention model) in the deep learning field, and the linear model may be understood as a layer of attention (attention layer) in the hierarchical attention mechanism. The hierarchical attention mechanism refers to that in the process of coding an image or a language, the deep neural network model does not need to input all contexts each time, only needs to selectively input the currently focused context, but in the embodiment, the deep neural network does not need to input all content information of resources each time of calculation, and only needs to input the content information of the current use stage of the resources.

In practice, the linear model may specifically include the following three designs:

the first design and the linear model can be preset functions, and input parameters of the preset functions comprise content information of the current use stage and use additional information. That is, the function expression of the linear model is fixed, and the linear model uses at least one of the content information of the current use stage, the operation duration of the use stage, the attribute information of the target user, the irrational information, the social information and the impression information as the input parameters, so that different semantic information can be obtained under different input parameters. The specific design of the preset function can be determined according to actual requirements, for example, a linear function, a nonlinear function, max-pooling (maximum pooling) or the like with unchanged dimensions such as weighting.

The second, linear model is designed as a preset function with adjustable parameters that are determined based on the content information of the current stage of use and using additional information. That is, the functional expression of the linear model is fixed and has adjustable parameters that are changed by at least one of content information of a current usage stage, an operation duration of the usage stage, attribute information of a target user, irrational information, social information, impression information.

And designing a third linear model which comprises a plurality of candidate functions, wherein when the linear model calculates the content information of the current use stage, the adopted candidate functions are determined based on the content information of the current use stage and the additional information. That is, the function expression of the linear model is not fixed, and the candidate function employed in the calculation is changed by the influence of at least one of the content information of the current use stage, the operation time length of the use stage, the attribute information of the target user, the irrational information, the social information, and the impression information. Alternatively, each candidate function may also have adjustable parameters that vary depending on the content information of the current stage of use and the use of additional information.

The first point to be noted is that when the multi-stage model calculates the semantic information in combination with the impression information, the impression information may be updated according to the output semantic information in the process of calculating the semantic information by the multi-stage model. After outputting the semantic information of the current use stage of the resource, the impression information of the last use stage can be updated according to the preference degree information corresponding to the semantic information of the current use stage, so as to improve the accuracy of the impression information of the last use stage.

For example, when the preference degree information of the current use stage accords with the preset condition, the content information of the last use stage accurately reflects the semantics of the content information of the current use stage, and after the target user sees the content information of the current use stage, the impression is better, so that the impression information of the last use stage can be increased to characterize the process of impression improvement. When the preference degree information of the current use stage does not accord with the preset bar, the condition that the content information of the last use stage fails to accurately reflect the semantics of the content information of the current use stage appears, and the user is deceived by a new and new title, so that the impression is worse after the target user sees the content information of the current use stage, and therefore the server can decrement the impression information of the last use stage to characterize the impression degradation process.

In the second aspect, the present embodiment is described by taking the case where the use additional information is input to the linear model as an example, and in the implementation, the use additional information may be input to the deep neural network model, and the deep neural network model may output a vector corresponding to the content information based on the content information of the current use stage and the use additional information.

204. The server performs resource information pushing based on at least one preference degree information.

After obtaining at least one preference degree information of the target user on the resource, the server can push the resource information based on the at least one preference degree information. For example, the content information of the first use stage of the resource is sent to the client logged in by the target user, and the client can receive the content information of the first use stage of the resource sent by the server and display the content information of the first use stage of the resource, so that the target user browses the pushed resource information on the client. The pushed resource information accords with the use preference of the target user, so that the personal interest requirement of the target user can be met.

The following is an exemplary description of the first to third designs regarding preference degree information used when pushing resource information.

And firstly, pushing the resource information based on preference degree information of the first use stage.

Considering that the preference degree information of the first use stage reflects the first impression of the target user on the resource, after the target user previews the resource, the decision on whether the resource fits the own interests is made, whether the content information of the second or even later use stages is displayed or not is directly determined, and the importance is extremely strong. For example, assuming that the content information of the first use stage of the movie is actors and titles, if the target user is not interested in the actors and titles, the probability of watching the movie is small, and if the target user is a fan of the actors of the movie, the probability of watching the movie is large. As another example, if the target user is not interested in the title and thumbnail of the article, the probability of browsing the entire text of the article may be small. Therefore, the server can push the resource information only based on the preference degree information of the first use stage, so that previews of the resources preferred by the target user are pushed preferentially, for example, titles and thumbnails of articles preferred by the target user are pushed preferentially, the titles and thumbnails of the articles displayed by the client are guaranteed to be the titles and thumbnails preferred by the target user, the target user is attracted to view the whole text of the articles, and the effect of accurate recommendation is achieved.

The specific manner of pushing the resource based on the preference degree information of the first use stage is exemplarily described below by (1.1) to (1.3).

(1.1) pushing content information of the first usage stage of the resource when the preference degree information of the first usage stage is greater than a first threshold.

The server may determine whether the preference degree information of the first usage stage is greater than a first threshold, and push content information of the first usage stage of the resource when the preference degree information of the first usage stage is greater than the first threshold. When the preference degree information of the first use stage is not greater than the first threshold value, the content information of the first use stage of the resource is not pushed, namely, the content information of each use stage of the resource is not pushed. The first threshold is used for indicating a minimum value of preference degree information corresponding to content information of a first use stage of the pushable resource, and the first threshold can be determined according to actual service requirements and can be preset by a developer.

Through (1.1), the effect of accurate recommendation can be realized: when the preference degree information of the first use stage is larger than the first threshold value, the preference degree information of the first use stage indicates that the user is interested in the content information of the first use stage to a strong enough degree, and the content information of the second use stage is likely to be checked in a progressive manner, so that the content information of the first use stage of the resource is pushed, the user is ensured to check the content information of the first use stage of the interested resource, the probability of the user for using the resource in a progressive manner is ensured to be as large as possible, and the conversion rate of pushing the content information of the first use stage is improved. When the preference degree information of the first use stage of the resource is not greater than the first threshold value, the preference degree information of the first use stage indicates that the user is interested in the content information of the first use stage to a weaker degree, and the content information of the second use stage cannot be further checked, so that the content information of the first use stage of the resource is obviously pushed to be in nonsensical behavior, and therefore the content information of the first use stage of the resource cannot be pushed so as not to disturb the user.

For example, for a scene of pushing an article, when preference degree information of the title and the thumbnail of the article is larger than a first threshold value, the preference degree information indicates that the user is interested in the title and the thumbnail is strong enough, and the user is likely to click to view the text, and the title and the thumbnail of the article are pushed at the moment, so that the title and the thumbnail of the pushed article are the title and the thumbnail of the user interested in the text, and the user is attracted to view the text of the article.

(1.2) pushing content information of the first usage stage of the resource when the preference degree information of the first usage stage is greater than the first threshold value and not greater than the second threshold value.

In distinction from the above (1.1), when the server determines whether the preference degree information of the first usage stage is greater than the first threshold, it may further determine whether the preference degree information of the first usage stage is greater than the second threshold, and when the preference degree information of the first usage stage is greater than the first threshold and not greater than the second threshold, the content information of the first usage stage of the resource is pushed, wherein the second threshold is greater than the first threshold, the second threshold is used for indicating a minimum value of the preference degree information corresponding to the content information of the first usage stage of the resource that can push the multi-stage content information, and the second threshold may be determined according to actual service requirements and may be preset by a developer. The technical effects of (1.2) are described below with reference to (1.3).

(1.3) pushing content information of at least one usage stage of the resource when the preference degree information of the first usage stage is greater than the second threshold.

Through (1.2) and (1.3), on the basis of realizing the effect of accurate recommendation, the speed of displaying the content information of the resource can be increased: when the preference degree information of the first use stage is larger than the first threshold value and not larger than the second threshold value, the content information of the first use stage of the resource is pushed only if the target user is interested in the content information of the first use stage, and then the content information of the second use stage of the resource is pushed again when the confirmation operation of the content information of the first use stage is detected. When the preference degree information of the first use stage is larger than the second threshold value, the target user is indicated to be interested in the content information of the first use stage, the target user is likely to advance the content information of each use stage of the resource, and the confirmation operation of the target user is not required to be waited, and the content information of other use stages of the resource is also pushed together when the content information of the first use stage is pushed, so that the terminal pre-loads the content information of each use stage of the resource. After that, when the target user triggers a confirmation operation on the content information of the first use stage, the terminal can directly display the content information of the second use stage without interaction with the server because the terminal has acquired the content information of the second use stage, thereby improving the speed of displaying the content information of the resource.

And secondly, pushing the resource information based on at least one statistic value of preference degree information, wherein the statistic value is a weighted sum value, an average value or a maximum value.

The server can calculate the statistic value of at least one preference degree information, and push the resource information according to the statistic value of the at least one preference degree information, so that the process of pushing the resource information is guaranteed to integrate interest preference of a target user on the content information of at least one use stage of the resource. The statistical value may be a weighted sum, an average or a maximum value.

Specifically, the server may set a corresponding weight coefficient for each use stage in advance, and perform weighted summation on each preference degree information based on the weight coefficient of each use stage to obtain a weighted sum value of at least one preference degree information, and perform resource information pushing according to the weighted sum value. The weight coefficient of each use stage can be determined according to actual requirements, for example, a larger weight can be set for the critical use stage such as the first use stage, the purchase stage and the like. In addition, the server may calculate an average value of at least one preference degree information, and perform resource information pushing according to the average value of the at least one preference degree information, or the server may select a maximum value from the at least one preference degree information, and perform resource information pushing according to the maximum value of the at least one preference degree information.

The specific manner of pushing resources based on the statistics of the at least one preference degree information is exemplarily described below by (2.1) to (2.3).

(2.1) pushing content information of the first usage stage of the resource when the statistics are greater than a third threshold.

The design is the same as the above (1.1), and the description thereof will be omitted.

(2.2) pushing content information of the first usage stage of the resource when the statistics are greater than the third threshold and not greater than the fourth threshold.

The design is the same as the above (1.2), and the description thereof will be omitted.

(2.3) pushing content information of at least one usage stage of the resource when the statistics are greater than a fourth threshold, the fourth threshold being greater than the third threshold.

The design is the same as the above (1.3), and the description thereof will be omitted.

And thirdly, pushing the resource information based on preference degree information of the appointed using stage.

In combination with the actual service requirement, the provider of the resource information may set a certain usage stage of at least one usage stage as a designated usage stage in advance, for example, set a usage stage that has a key influence on the service as a designated usage stage, for example, for resources such as commodities, the key is that the funds flow, the usage stage that triggers the purchase operation may be set as a designated usage stage, for articles, the key is that the traffic is generated and the number of people browsed, and the usage stage that triggers the forwarding operation may be set as a designated usage stage. Then, the server can determine the appointed use stage according to the setting operation, and push the resource information based on the preference degree information of the appointed use stage, so that the process of pushing the resource information is ensured to meet the actual service requirement, namely, in the process of providing the content information of the appointed use stage of the resource, the server attracts a user to use the resource in a further order, improves the conversion rate of the resource in the appointed use stage, and brings great commercial value.

The specific manner of pushing resources based on the statistics of the at least one preference degree information is exemplarily described below through (3.1) to (3.3).

(3.1) pushing content information of the designated usage stage of the resource when the preference degree information of the designated usage stage is greater than a fifth threshold.

(3.2) pushing content information of the specified usage stage of the resource when the preference degree information of the specified usage stage is greater than a fifth threshold and not greater than a sixth threshold, the sixth threshold being greater than the fifth threshold.

(3.3) pushing content information of at least one usage stage of the resource when the preference degree information of the designated usage stage is greater than a sixth threshold.

In summary, the preference degree information adopted in pushing the resource information is described above through the first to third designs, and the specific manner of pushing the resource information is described below through the first to third designs.

And (A) determining the order of pushing the resource information based on at least one preference degree information.

The server may have a plurality of resource information available for pushing, and when at least one preference degree information of any one resource information is obtained, an order of pushing the resource information may be determined based on the at least one preference degree information, so that each resource information is pushed in turn in a certain order. The order of pushing the resource information is positively correlated with the at least one preference degree information, namely, the higher the at least one preference degree information of the resource information is, the earlier the order of pushing the resource information is, so as to ensure that the target user prefers to use the resource preferentially to be pushed.

It should be noted that, the design a may be combined with any one of the above designs one to three:

design a combines the above designs one: the order in which the resource information is pushed may be determined based on preference degree information of the first use stage of the resource. Wherein the plurality of resources may be ordered according to the preference degree information of the first use stage, and the order of pushing the content information of the first use stage of each resource is determined according to the order of the preference degree information of the first use stage, for example, if the preference degree information of the first use stage of a certain resource is highest, the content information of the first use stage of the resource is pushed first.

Design a combines the above designs two: the order of pushing the resource information may be determined based on the statistics of the at least one preference degree information, wherein the plurality of resources may be ordered according to the statistics of the at least one preference degree information, and the order of pushing the content information of the first usage stage of each resource is determined according to the order of the statistics of the at least one preference degree information, e.g. if the statistics of the at least one preference degree information of a certain resource is highest, the content information of the first usage stage of the resource is pushed first.

Design a combines the above designs three: the order in which the resource information is pushed may be determined based on preference degree information specifying the use stage. The plurality of resources may be ordered according to the preference degree information of the designated use stage, and the order of pushing the content information of the first use stage of each resource is determined according to the order of the preference degree information of the designated use stage, for example, if the preference degree information of the designated use stage of a certain resource is highest, the content information of the first use stage of the resource is pushed first.

And B, determining at least one of the position, the sequence and the occupied page proportion of the resource information in the push page based on the at least one preference degree information.

When the server obtains at least one piece of preference degree information of any piece of resource information, at least one of the position, the sequence and the occupied page proportion of the resource information in the push page can be determined based on the at least one piece of preference degree information, a push page carrying a plurality of pieces of resource information is generated according to at least one of the position, the sequence and the occupied page proportion of each piece of resource information in the push page, and the push page is sent to the terminal. The higher the at least one preference degree information of the resource information is, the closer the position of the resource information in the pushed page is to the center, the earlier the sequence is, and the larger the proportion of occupied pages is, so that the display sequence/display style of the pushed page is optimized, and the pushed page is ensured to meet the interest demands of target users.

It should be noted that the design B may be combined with any one of the above designs one to three.

Design B combines the above designs one: at least one of a location, an order, and an occupied page proportion of the resource information in the push page may be determined based on preference degree information of the first use stage of the resource.

For the process of determining the order of the resource information in the push page, the plurality of resources may be ordered according to the preference degree information of the first use stage, and the order of each resource information in the push page may be determined according to the order of the preference degree information of the first use stage of each resource. For example, the push page may be in a list format, where the order of each resource information in the list is equal to the order of the corresponding preference information, e.g. if the preference information of the first use stage of a certain resource is highest, then it is located first in the list.

For the process of determining the position of the resource information in the push page, for a plurality of resource information to be pushed, the resource information with highest preference degree information in the first use stage can be selected, and the resource information is arranged at the center/top of the push page so as to ensure that the resource information is highlighted.

For the process of determining the occupation proportion of the resource information in the push page, a mapping relation between the preference degree information and the occupation proportion in the first use stage can be pre-established, and the occupation proportion and the preference degree information in the mapping relation are positively correlated. After determining the preference degree information of the first use stage of a certain resource, the occupation proportion corresponding to the preference degree information of the first use stage can be queried based on the mapping relation, and the corresponding occupation proportion is used as the occupation proportion of the resource information in the push page. And then, generating a push page according to the occupied page proportion of each piece of resource information.

In addition, the design B combines the design B, and at least one of the position, the sequence and the occupied page proportion of the resource information in the push page can be determined based on the statistical value of the at least one preference degree information of the resource. Design B in combination with the design III, at least one of the position, the sequence and the occupied page proportion of the resource information in the pushed page can be determined based on the preference degree information of the specified use stage of the resource. The process of combining the design B with the design two or the design three is the same as the process of combining the design B with the design one, and the description thereof will be omitted.

And C, determining whether to push the resource information or not based on at least one preference degree information.

The server may determine whether to push the resource information or not to push the resource information based on the at least one preference degree information. In addition, when it is determined that the resource information is not pushed, the next resource information may be continuously acquired, and the above steps are repeatedly performed to determine whether to push the next resource information again. The specific process of the design C is the same as the designs (1.1) to (1.3), the designs (2.1) to (2.3), and the designs (3.1) to (3.3), and in addition, the design C can be combined with any one of the designs one to three. And will not be described in detail herein.

According to the method provided by the embodiment of the invention, the multi-stage model is introduced, the process of calculating the preference degree information by the multi-stage model is consistent with the process of using the resources by the user in a multi-stage manner, and the accuracy of the calculated preference degree information is high, so that the resource information is pushed based on the preference degree information of the multi-stage model, and the accuracy of pushing the resource information can be improved.

The training process of the multi-stage model is explained below by means of the embodiment of fig. 7.

Fig. 7 is a flowchart of a model training method according to an embodiment of the present invention. Referring to fig. 7, the method includes the steps of:

701. The server obtains a plurality of sample resource information.

This step is similar to step 201 described above, except that the sample resource information includes not only content information of a plurality of usage phases, but also sample tags of the content information of each usage phase, so that the server trains the model according to the sample tags.

The sample label is used for indicating whether a sample user triggers clicking operation on content information of a corresponding use stage of the sample resource information, and the sample label can be manually marked in the sample resource information by a developer. The sample tag may include a first sample tag indicating that the sample user has triggered a confirmation operation for content information of a corresponding usage stage of the sample resource information, and a second sample tag indicating that the sample user has not triggered a confirmation operation for content information of a corresponding usage stage of the sample resource information.

In implementation, the first sample tag and the second sample tag may be represented by numbers, where the first sample tag is 1, the second sample tag is 0, or the first sample tag and the second sample tag may also be represented by other forms, and the embodiment of the present invention does not limit the form of representation of the sample tag.

702. The server inputs each sample resource information and the usage preference information of the sample user into the initial multi-stage model, and outputs preference degree information of the sample user for at least one usage stage of each sample resource.

The initial multi-stage model may include a plurality of initial sub-models of the usage stage, each initial sub-model is provided with initial parameters, semantic information of the corresponding usage stage may be predicted and output according to content information of the corresponding usage stage of the sample resource, preference degree information is calculated according to the semantic information of the corresponding usage stage and the usage preference information, and when the preference degree information meets a preset condition, the next initial sub-model is triggered to calculate the semantic information. In addition, the initial parameters can be adjusted according to semantic information so as to achieve the purpose of training the initial sub-model. The process of calculating the preference degree information of the sample resource information by using the initial multi-stage model is the same as that of step 203, and will not be described herein.

Optionally, in conjunction with the design using the additional information, in step 702, the server may also input the additional information into the multi-stage model on the basis of inputting each sample resource information and the usage preference information of the sample user into the initial multi-stage model, and the initial sub-model of each usage stage in the initial multi-stage model may output the semantic information of the corresponding usage stage according to the content information and the additional information of the corresponding usage stage.

703. And the server adjusts parameters of the initial sub-model of each using stage according to the deviation between the preference degree information of each using stage of each sample resource information and the sample label of the corresponding using stage until the deviation corresponding to the preference degree information output by each initial sub-model is smaller than a preset threshold value.

After the server obtains the preference degree information of any use stage, the deviation between the preference degree information of the use stage and the sample label of the use stage can be calculated, the deviation can reflect the accuracy degree of the semantic information output by the initial sub-model, and the smaller the deviation is, the more the semantic information output by the initial sub-model can reflect the semantic of the content information. And then, the server can judge whether the deviation is smaller than a preset threshold value, when the deviation is larger than the preset threshold value, the parameters of the initial sub-model in the using stage are adjusted, preference degree information is calculated according to the initial sub-model again, the parameters are adjusted again until the deviation between the preference degree information calculated according to the initial sub-model and the sample label is smaller than the preset threshold value, and then the training of the initial sub-model is completed. The preset threshold may be determined according to actual requirements, for example, when the accuracy of the submodel is required to be higher, the higher the preset threshold is set.

In summary, the embodiments of fig. 2 and fig. 7 above illustrate a multi-stage model and a scheme for pushing resource information based on preference information of at least one usage stage output by the multi-stage model. In the implementation, only the first stage model may be established, and the resource information may be pushed based on the preference degree information output by the first stage model. This scheme is illustrated below by the embodiment of fig. 8:

fig. 8 is a flowchart of a method for pushing resource information according to an embodiment of the present invention. The execution body of the embodiment of the invention is a server, referring to fig. 8, the method includes:

801. the server obtains the resource information.

This step is similar to step 201 described above, and will not be described here.

802. The server obtains the use preference information of the target user according to the operation behavior of the target user on the content information of the first use stage of the history resource.

The content information of the first usage stage is used to provide previews of historical resources, such as titles and thumbnails of articles are used to provide previews of articles, posters and profiles of movies are used to provide previews of movies, etc.

This step is similar to step 202 described above, except that the server may simply acquire the operation behavior of the target user on the content information of the first usage stage of the history resource, and acquire the usage preference information according to the operation behavior of the target user on the content information of the first usage stage of the history resource, without acquiring and considering the operation behavior of the target user on the content information of other usage stages of the history resource.

By this way of obtaining the usage preference information, the following technical effects can be achieved:

first, accuracy of prediction preference is improved: taking an article as an example, most users often seldom carefully browse the full text of the article, and the full text is often subjected to glance, skip and other modes, so that the operation behavior of the full text is likely to not accurately reflect the preference of the users. The titles and thumbnails of the articles are usually short in content and highlighted in display, are essence of the articles, the attention of the user is usually focused on the titles and thumbnails of the articles, the browsing behaviors of the titles and thumbnails of the articles accurately reflect the interested degree of the users on the articles, and the method is also a core for determining whether the users can read in steps. In this embodiment, the usage preference information is obtained through the operation behavior of the content information in the first usage stage of the historical resources, so that the usage preference information is ensured to accurately describe the interested degree of the user when previewing the resources, and the accuracy is higher.

Second, the amount of calculation of the prediction preference is reduced. The operation behaviors except the operation behaviors of the first use stage are not required to be acquired, the use preference information of the target user is not required to be acquired according to the operation behaviors except the operation behaviors of the first use stage, and the calculation amount is small.

803. The server inputs the content information and the usage preference information of the first usage stage of the resource into the first stage model, and outputs the preference degree information of the target user on the resource.

The step is similar to step 203, and the first stage model is similar to the sub-model of the first use stage in the multi-stage model, and is used for predicting preference degree information of the user on the resource according to the content information of the first use stage of the resource. The difference is that the first stage model is a complete and independent model, and only the preference degree information output by the first stage model, namely the preference degree information of the first use stage of the resource, is required to be obtained, and other models are not required to be triggered to calculate. Then, since the preference degree information only needs to be calculated from the content information of the first use stage, the calculation amount can be greatly reduced.

In addition, in combination with the design using the additional information, the content information, the usage preference information and the usage additional information of the first usage stage of the resource may be input into the first stage model, and the first stage model may output the preference degree information of the target user for the resource according to the content information, the usage preference information and the usage additional information of the first usage stage of the resource.

804. And the server pushes the resource information based on the preference degree information.

The same procedure as the above step 204 is different in that the server only needs to push the resource information based on the preference degree information of the first use stage, and does not need to consider the preference degree information of other use stages.

The following is an exemplary description of the specific manner of pushing the resource information based on the preference degree information from design one to design three.

And firstly, pushing content information of a first use stage of the resource when the preference degree information is larger than a first threshold value.

The design is the same as the above step 204 (1.1), and will not be described here.

And secondly, pushing content information of the first use stage of the resource when the preference degree information is larger than the first threshold value and not larger than the second threshold value.

The design is the same as the above step 204 (1.2), and will not be described here.

And thirdly, pushing content information of at least one use stage of the resource when the preference degree information is larger than a second threshold value, wherein the second threshold value is larger than the first threshold value.

The design is the same as the above step 204 (1.3), and will not be described here.

In addition, when pushing the resource information based on the preference degree information, the order of pushing the resource information may be determined based on the preference degree information, and the specific process is the same as the design a of the step 204, and will not be described herein. Or, when pushing the resource information based on the preference degree information, at least one of the position, the sequence and the occupied page proportion of the resource information in the pushed page may be determined based on the preference degree information, and the specific process is the same as the design B in the step 204, and will not be described herein. Or, when pushing the resource information based on the preference degree information, whether to push the resource information may be determined based on the preference degree information, and the specific process is the same as the design C in the step 204, which is not described herein.

According to the method provided by the embodiment, the first stage model is introduced, the process of calculating the preference degree information by the first stage model is matched with the process of analyzing the interest degree of the resource according to the previewing of the resource by the user, and when the resource information is pushed based on the preference degree information of the first stage model, the previewing of the resource can be ensured to meet the interest preference of the user, and the accuracy of pushing the resource information can be improved.

The training process of the first stage model is explained below by way of the embodiment of fig. 9.

Fig. 9 is a flowchart of a model training method according to an embodiment of the present invention. Referring to fig. 9, the method includes the steps of:

901. the server obtains a plurality of sample resource information.

This step is similar to step 701 described above, except that each sample resource information may include only the content information of the first usage stage and the sample tag of the content information of the first usage stage.

902. The server inputs each sample resource information and the use preference information of the sample user into the initial first-stage model, and outputs preference degree information of the sample user for each sample resource.

The initial first-stage model may be provided with initial parameters, and semantic information of the first use stage may be output according to content information of the first use stage of the sample resource, and preference degree information may be calculated according to the semantic information of the first use stage and the preference information. In addition, the initial parameters can be adjusted according to semantic information so as to achieve the purpose of training an initial first-stage model. The process of calculating the preference degree information of the sample resource information by using the initial first-stage model is the same as that of step 703, and will not be described here again.

Optionally, in combination with the design using the additional information, the initial first-stage model may further use the additional information as an input variable, and the server may input content information, usage preference information, and usage additional information of a first usage stage of the sample resource into the initial first-stage model, so that the initial first-stage model outputs preference degree information of the sample user for the sample resource according to the content information, the usage preference information, and the usage additional information of the first usage stage of the resource.

903. And the server adjusts parameters of the initial first-stage model according to the deviation between the preference degree information of the first use stage of each sample resource information and the sample label of the first use stage until the deviation corresponding to the preference degree information output by the initial first-stage model is smaller than a preset threshold value.

After the server obtains the preference degree information of the first use stage, the deviation between the preference degree information of the first use stage and the sample label of the first use stage can be calculated, the deviation can reflect the accuracy degree of the semantic information output by the initial first stage model, and the smaller the deviation is, the more the semantic information output by the initial first stage model can reflect the semantic of the content information. And then, the server can judge whether the deviation is smaller than a preset threshold value, and when the deviation is larger than the preset threshold value, the initial first-stage model is adjusted, preference degree information is calculated according to the initial first-stage model again, parameters are adjusted again until the deviation between the preference degree information calculated according to the initial first-stage model and the sample label is smaller than the preset threshold value, and then the training of the initial first-stage model is completed. The preset threshold may be determined according to actual requirements, for example, when the accuracy of the initial first-stage model is required to be higher, the higher the preset threshold is set.

Fig. 10 is a schematic structural diagram of a resource information pushing device according to an embodiment of the present invention. Referring to fig. 10, the apparatus includes: an acquisition module 1001, an input-output module 1002 and a push module 1003.

An obtaining module 1001, configured to obtain resource information, where the resource information includes content information of a plurality of different usage phases of a resource;

an input-output module 1002, configured to input the resource information and usage preference information of the target user into a multi-stage model, and output preference degree information of the target user for at least one usage stage of the resource, where the multi-stage model is configured to predict preference degree information of the target user for different usage stages of the resource according to content information of the different usage stages of the resource;

and a pushing module 1003, configured to perform resource information pushing on the target user based on at least one preference degree information.

In one possible design, the multi-stage model includes a plurality of sub-models of use stages, each sub-model of use stage is used for outputting semantic information of the corresponding use stage according to content information of the corresponding use stage, and when the semantic information output by the sub-model of each use stage matches with the use preference information, the sub-model of the next use stage is triggered to calculate the semantic information.

In one possible design, the input-output module 1002 includes:

the input/output sub-module is used for inputting the content information of the current use stage into the sub-model of the current use stage in the multi-stage model and outputting the semantic information of the current use stage;

the calculating sub-module is used for calculating preference degree information of the current use stage according to semantic information and the use preference information of the current use stage;

the input/output sub-module is also used for outputting preference degree information of the current use stage;

the input/output sub-module is further configured to input content information of a next use stage into a sub-model of the next use stage when the preference degree information of the current use stage meets a preset condition.

In one possible design, the input-output sub-module is further configured to: inputting the content information of the current use stage into a deep neural network model of a sub-model of the current use stage, outputting a vector corresponding to the content information, and training the deep neural network model according to the content information of the corresponding use stage; and inputting the vector corresponding to the content information and the additional information of the current use stage into a linear model of a sub-model of the current use stage, outputting semantic information of the current use stage, and training the linear model according to the additional information of the corresponding use stage.

In one possible design, the linear model is a preset function, and the input parameters of the preset function include content information of the current use stage and use additional information; or alternatively, the first and second heat exchangers may be,

the linear model is a preset function with adjustable parameters, and the adjustable parameters are determined based on the content information of the current use stage and the use additional information; or alternatively, the first and second heat exchangers may be,

the linear model includes a plurality of candidate functions, and the candidate functions used in calculating the content information of the current use stage are determined based on the content information of the current use stage and the use additional information.

In one possible design, the training process of the multi-stage model includes:

acquiring a plurality of sample resource information, wherein each sample resource information comprises content information of a plurality of using stages of the sample resource and a sample label of the content information of each using stage, and the sample label is used for indicating whether the corresponding content information is displayed or not;

inputting the information of each sample resource and the use preference information of the sample user into an initial multi-stage model, and outputting preference degree information of the sample user for at least one use stage of each sample resource;

And adjusting parameters of the initial sub-model of each using stage according to the deviation between the preference degree information of each using stage of each sample resource and the sample label of the corresponding using stage until the deviation of the preference degree information output by each initial sub-model is smaller than a preset threshold value.

In one possible design, the plurality of usage phases includes a first usage phase that is a phase that exposes at least one of a title, a thumbnail, a summary of the resource, and a second usage phase that is a phase that exposes the full text of the resource.

In one possible design, the multi-stage model includes a plurality of sub-models of usage stages, each sub-model of usage stage further configured to output semantic information corresponding to the usage stage based on content information and usage additional information corresponding to the usage stage.

In one possible design, the usage additional information includes at least one of an operation duration of a usage period, attribute information of the target user, irrational information, social information, impression information;

wherein the irrational information includes at least one of user emotion information, user health information, weather information, traffic information, date event information, labor status information, and the impression information is used for indicating an impression of the target user on a degree of matching between content information of a corresponding use stage and content information of a next use stage of the provider of the resource information.

In one possible design, the process of obtaining the usage preference information includes:

and acquiring the use preference information of the target user according to the operation behavior of the target user on the content information of at least the first use stage of the historical resources.

In one possible design, the obtaining module 1001 is further configured to: when the browsing behavior of the target user on the content information of any use stage of the historical resources is detected, acquiring the duration of the browsing behavior; and/or when detecting the listening behavior of the target user to the content information in any use stage of the historical resources, acquiring the duration of the listening behavior; and/or when the watching behavior of the content information of any using stage of the historical resources by the target user is detected, acquiring the duration of the watching behavior; and/or when detecting the game behavior of the content information of any use stage of the historical resources by the target user, acquiring at least one of duration, operation times and operation frequency of the game behavior; and/or when detecting the confirmation behavior of the target user to the content information of any use stage of the historical resources, recording the confirmation behavior, wherein the confirmation behavior refers to the behavior of confirming to enter the next use stage; and/or when detecting the interaction behavior of the target user on the content information of any use stage of the historical resources, recording the interaction behavior, wherein the interaction behavior refers to the behavior that the target user interacts with other users or the provider of the resource information.

In one possible design, the obtaining module 1001 is further configured to: recording the comment content when detecting the comment behavior of the target user on the content information of any use stage of the historical resources; and/or when the sharing behavior of the target user on the content information of any use stage of the historical resources is detected, recording the sharing times or the number of the shared objects; and/or recording a score of the scoring action when the scoring action of the target user on the content information of any using stage of the historical resources is detected; and/or, when detecting a reward behavior of the target user for content information at any one of the use stages of the history resource, recording an amount of the reward behavior; and/or recording at least one of the number, the category, and the amount of the virtual gift sent when the gift sending behavior of the content information of any one of the use stages of the history resource by the target user is detected; and/or, when detecting the purchasing behavior of the content information of any using stage of the historical resources by the target user, recording at least one of the quantity and the amount of the purchasing behavior; and/or recording the content of the transmitted barrage when the content information transmitting barrage behavior of any use stage of the historical resources by the target user is detected.

In one possible design, the confirmation behavior includes at least one of a click behavior, a slide behavior, a long press behavior, a voice confirmation behavior.

In one possible design, the resource is an article;

the content information of the first reading stage of the article comprises at least one item of title, thumbnail and abstract of the article, and the content information of the second reading stage of the article is the whole text of the article; or alternatively, the first and second heat exchangers may be,

the content information of the first reading stage of the article comprises at least one of a title, a thumbnail and a abstract of the article, the content information of the second reading stage of the article is a pilot-viewing fragment of the article, and the content information of the second reading stage of the article is the full text of the article.

In one possible design, the asset is video;

the content information of the first viewing stage of the video comprises at least one of a poster, a first frame picture, a key frame picture, a title, an abstract, an actor and a key fragment of the video, and the content information of the second viewing stage of the video is the whole content of the video; or alternatively, the first and second heat exchangers may be,

the content information of the first viewing stage of the video comprises at least one of a poster, a first frame picture, a key frame picture, a title, an abstract, an actor and a key fragment of the video, the content information of the second viewing stage of the video is a pilot viewing fragment of the video, and the content information of the third viewing stage of the video is the whole content of the video.

In one possible design, the resource is music;

the content information of the first listening stage of the music includes at least one of a name of the music, a poster of the music, a picture of a singer, a name of the singer, a title of a song list to which the music belongs, a name of an album to which the music belongs, an album screenshot to which the music belongs, a user name of the uploaded music, a category of the music, and a number of times of playing the music;

the content information of the second listening phase of music includes at least one of music, lyrics of music, comments of music, album video to which music belongs.

In one possible design, the resource is a commodity;

the content information of the first purchase stage of the commodity includes at least one of a picture of the commodity, a name of the commodity, an advertisement word of the commodity, a price of the commodity, a freight rate of the commodity, sales volume of the commodity, a number of purchasers of the commodity, an attribute of the commodity, a high-frequency evaluation of the commodity, a location of the commodity, preference information of the commodity, a distance between the commodity and a target user, a name of a seller, an after-sales service of the commodity, a brand of the commodity, and a score of the commodity;

the content information of the second purchase stage of the commodity includes at least one of details of the commodity, each evaluation of the commodity, question and answer information of the commodity, matching package information of the commodity, similar commodity information of the commodity, details of a seller, pictures of the commodity, names of the commodity, advertisement words of the commodity, prices of the commodity, freight of the commodity, sales volume of the commodity, number of purchases of the commodity, attribute of the commodity, high-frequency evaluation of the commodity, location of the commodity, preferential information of the commodity, distance between the commodity and a target user, names of the seller, after-sale service of the commodity, brands of the commodity, and scores of the commodity;

The content information of the third purchase stage of the commodity includes at least one of a name of the commodity purchased, a number of the commodity purchased, an order amount, a remark to the seller, and a receiving address.

In one possible design, the resource is a game;

the content information of the first operation stage of the game includes at least one of a name of the game, an icon of the game, a size of the game, an advertisement word of the game, and a number of downloading persons of the game;

the content information of the second operational stage of the game includes at least one of an introduction to the game, an animation of the game, a promotional video of the game, a size of the game, advertising words of the game, comments of the game, scoring of the game, versions of the game, labels of the game, and similar games of the game;

the content information of the third operation stage of the game includes at least one of a combat screen, a competition screen, an administration screen, a adventure screen, a dialogue screen, a card drawing screen, a shooting screen, a role playing screen, a prop selection screen, a prop purchase screen, and a chess and card screen.

In one possible design, the push module 1003 is further configured to: determining an order in which the resource information is pushed based on at least one preference degree information; and/or determining at least one of a position, an order and an occupied page proportion of the resource information in the push page based on the at least one preference degree information; and/or determining whether to push the resource information based on the at least one preference degree information.

In one possible design, the push module 1003 is further configured to: pushing resource information based on preference degree information of the first use stage; and/or pushing resource information based on at least one statistic value of preference degree information, wherein the statistic value is a weighted sum value, an average value or a maximum value; and/or pushing the resource information based on preference degree information of the designated using stage.

In one possible design, the push module 1003 is further configured to: the pushing of the resource information based on the preference degree information of the first use stage comprises the following steps:

pushing content information of the first use stage of the resource when the preference degree information of the first use stage is larger than a first threshold value; or pushing the content information of the first use stage of the resource when the preference degree information of the first use stage is larger than the first threshold value and not larger than the second threshold value; or pushing content information of at least one use stage of the resource when the preference degree information of the first use stage is larger than a second threshold value, wherein the second threshold value is larger than the first threshold value; and/or the number of the groups of groups,

pushing content information of the first use stage of the resource when the statistic value is larger than a third threshold value; or pushing the content information of the first use stage of the resource when the statistical value is larger than the third threshold value and not larger than the fourth threshold value; or pushing content information of at least one using stage of the resource when the statistic value is larger than a fourth threshold value; and/or the number of the groups of groups,

Pushing content information of the appointed use stage of the resource when the preference degree information of the appointed use stage is larger than a fifth threshold value; or pushing content information of the specified use stage of the resource when the preference degree information of the specified use stage is larger than the fifth threshold value and not larger than the sixth threshold value; or pushing content information of at least one usage stage of the resource when the preference degree information of the designated usage stage is greater than a sixth threshold.

Fig. 11 is a schematic structural diagram of a resource information pushing device according to an embodiment of the present invention. Referring to fig. 11, the apparatus includes: an acquisition module 1101, an input-output module 1102, and a push module 1103.

An obtaining module 1101, configured to obtain usage preference information of a target user according to an operation behavior of the target user on content information of a first usage stage of a historical resource, where the content information of the first usage stage is used to provide a preview of the historical resource;

an input/output module 1102, configured to input content information of a first use stage of a resource and the use preference information into a first stage model, and output preference degree information of the target user for the resource;

a pushing module 1103, configured to perform resource information pushing based on the preference degree information;

In one possible design, the pushing module 1103 is configured to: pushing content information of a first use stage of the resource when the preference degree information is larger than a first threshold value; or pushing the content information of the first use stage of the resource when the preference degree information is larger than the first threshold value and not larger than the second threshold value; or pushing content information of at least one use stage of the resource when the preference degree information is larger than a second threshold value, wherein the second threshold value is larger than the first threshold value.

In one possible design, the operation behavior of the content information of the first usage stage refers to a browsing behavior of the content information of the first usage stage of the history resource, where the content information of the first usage stage of the history resource includes at least one of a title, a thumbnail, and a summary of the history resource.

In one possible design, the apparatus further comprises:

and the receiving module is used for receiving the duration of the browsing behavior of the target user on the content information of the first use stage of the historical resources, which is sent by the terminal.

It should be noted that: the resource information pushing device provided in the above embodiment only illustrates the division of the above functional modules when pushing the resource information, and in practical application, the above functional allocation may be completed by different functional modules according to needs, that is, the internal structure of the server is divided into different functional modules, so as to complete all or part of the functions described above. In addition, the resource information pushing device and the resource information pushing method provided in the foregoing embodiments belong to the same concept, and detailed implementation processes of the resource information pushing device and the resource information pushing method are detailed in the method embodiments and are not repeated here.

Fig. 12 is a schematic structural diagram of a resource information display device according to an embodiment of the present invention, referring to fig. 12, the device includes: a transmitting module 1201, a receiving module 1202 and a presentation module 1203.

A transmitting module 1201, configured to, when detecting an operation behavior of the target user on the content information of at least the first usage stage of the history resource, transmit the operation behavior to the server;

a receiving module 1202, configured to receive content information of a first usage stage of a resource sent by the server;

a display module 1203, configured to display content information of the first usage stage of the resource;

It should be noted that: when the resource information display device provided in the above embodiment displays resource information, only the division of the above functional modules is used for illustration, in practical application, the above functional allocation may be completed by different functional modules according to needs, that is, the internal structure of the terminal is divided into different functional modules, so as to complete all or part of the functions described above. In addition, the resource information display device and the resource information display method provided in the foregoing embodiments belong to the same concept, and detailed implementation processes of the resource information display device and the resource information display method are detailed in the method embodiments and are not repeated here.

Fig. 13 is a schematic structural diagram of a server according to an embodiment of the present invention, where the server 1300 may have a relatively large difference due to different configurations or performances, and may include one or more processors (central processing units, CPU) 1301 and one or more memories 1302, where the memories 1302 store at least one instruction, and the at least one instruction is loaded and executed by the processor 1301 to implement the methods provided in the foregoing method embodiments. Of course, the server may also have a wired or wireless network interface, a keyboard, an input/output interface, and other components for implementing the functions of the device, which are not described herein.

In an exemplary embodiment, a computer-readable storage medium storing a computer program, for example, a memory storing a computer program, which when processed and executed implements the resource information pushing method shown in the above embodiment, is also provided. For example, the above-mentioned computer readable storage medium may be a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a compact disc Read-Only Memory (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, and the like.

Fig. 14 is a schematic structural diagram of a terminal 1400 according to an embodiment of the present invention. Terminal 1400 may be: a smart phone, a tablet computer, an MP3 player (Moving Picture Experts Group Audio Layer III, motion picture expert compression standard audio plane 3), an MP4 (Moving Picture Experts Group Audio Layer IV, motion picture expert compression standard audio plane 4) player, a notebook computer, or a desktop computer. Terminal 1400 may also be referred to as a user device, a portable terminal, a laptop terminal, a desktop terminal, and the like.

In general, terminal 1400 includes: a processor 1401 and a memory 1402.

Processor 1401 may include one or more processing cores, such as a 4-core processor, an 8-core processor, and the like. The processor 1401 may be implemented in at least one hardware form of DSP (Digital Signal Processing ), FPGA (Field-Programmable Gate Array, field programmable gate array), PLA (Programmable Logic Array ). The processor 1401 may also include a main processor, which is a processor for processing data in an awake state, also called a CPU (Central Processing Unit ), and a coprocessor; a coprocessor is a low-power processor for processing data in a standby state. In some embodiments, the processor 1401 may be integrated with a GPU (Graphics Processing Unit, image processor) for rendering and rendering of content required to be displayed by the display screen. In some embodiments, the processor 1401 may also include an AI (Artificial Intelligence ) processor for processing computing operations related to machine learning.

Memory 1402 may include one or more computer-readable storage media, which may be non-transitory. Memory 1402 may also include high-speed random access memory, as well as non-volatile memory, such as one or more magnetic disk storage devices, flash memory storage devices. In some embodiments, a non-transitory computer readable storage medium in memory 1402 is used to store at least one instruction for execution by processor 1401 to implement the resource information presentation methods provided by the method embodiments herein.

In some embodiments, terminal 1400 may optionally further include: a peripheral interface 1403 and at least one peripheral. The processor 1401, memory 1402, and peripheral interface 1403 may be connected by a bus or signal lines. The individual peripheral devices may be connected to the peripheral device interface 1403 via buses, signal lines or a circuit board. Specifically, the peripheral device includes: at least one of radio frequency circuitry 1404, a touch display screen 1405, a camera 1406, audio circuitry 1407, a positioning component 1408, and a power source 1409.

Peripheral interface 1403 may be used to connect at least one Input/Output (I/O) related peripheral to processor 1401 and memory 1402. In some embodiments, processor 1401, memory 1402, and peripheral interface 1403 are integrated on the same chip or circuit board; in some other embodiments, either or both of processor 1401, memory 1402, and peripheral interface 1403 may be implemented on separate chips or circuit boards, which is not limited in this embodiment.

The Radio Frequency circuit 1404 is configured to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The radio frequency circuit 1404 communicates with a communication network and other communication devices via electromagnetic signals. The radio frequency circuit 1404 converts an electrical signal into an electromagnetic signal for transmission, or converts a received electromagnetic signal into an electrical signal. Optionally, the radio frequency circuit 1404 includes: antenna systems, RF transceivers, one or more amplifiers, tuners, oscillators, digital signal processors, codec chipsets, subscriber identity module cards, and so forth. The radio frequency circuit 1404 may communicate with other terminals via at least one wireless communication protocol. The wireless communication protocol includes, but is not limited to: metropolitan area networks, various generations of mobile communication networks (2G, 3G, 4G, and 5G), wireless local area networks, and/or WiFi (Wireless Fidelity ) networks. In some embodiments, the radio frequency circuit 1404 may also include NFC (Near Field Communication, short range wireless communication) related circuits, which are not limited in this application.

The display screen 1405 is used to display UI (User Interface). The UI may include graphics, text, icons, video, and any combination thereof. When the display screen 1405 is a touch display screen, the display screen 1405 also has the ability to collect touch signals at or above the surface of the display screen 1405. The touch signal may be input to the processor 1401 as a control signal for processing. At this time, the display 1405 may also be used to provide virtual buttons and/or a virtual keyboard, also referred to as soft buttons and/or a soft keyboard. In some embodiments, the display 1405 may be one, providing a front panel of the terminal 1400; in other embodiments, the display 1405 may be at least two, respectively disposed on different surfaces of the terminal 1400 or in a folded design; in still other embodiments, the display 1405 may be a flexible display disposed on a curved surface or a folded surface of the terminal 1400. Even more, the display 1405 may be arranged in a non-rectangular irregular pattern, i.e. a shaped screen. The display 1405 may be made of LCD (Liquid Crystal Display ), OLED (Organic Light-Emitting Diode) or other materials.

The camera component 1406 is used to capture images or video. Optionally, camera assembly 1406 includes a front camera and a rear camera. Typically, the front camera is disposed on the front panel of the terminal and the rear camera is disposed on the rear surface of the terminal. In some embodiments, the at least two rear cameras are any one of a main camera, a depth camera, a wide-angle camera and a tele camera, so as to realize that the main camera and the depth camera are fused to realize a background blurring function, and the main camera and the wide-angle camera are fused to realize a panoramic shooting and Virtual Reality (VR) shooting function or other fusion shooting functions. In some embodiments, camera assembly 1406 may also include a flash. The flash lamp can be a single-color temperature flash lamp or a double-color temperature flash lamp. The dual-color temperature flash lamp refers to a combination of a warm light flash lamp and a cold light flash lamp, and can be used for light compensation under different color temperatures.

The audio circuitry 1407 may include a microphone and a speaker. The microphone is used for collecting sound waves of users and the environment, converting the sound waves into electric signals, and inputting the electric signals to the processor 1401 for processing, or inputting the electric signals to the radio frequency circuit 1404 for voice communication. For purposes of stereo acquisition or noise reduction, a plurality of microphones may be provided at different portions of the terminal 1400, respectively. The microphone may also be an array microphone or an omni-directional pickup microphone. The speaker is used to convert electrical signals from the processor 1401 or the radio frequency circuit 1404 into sound waves. The speaker may be a conventional thin film speaker or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, not only the electric signal can be converted into a sound wave audible to humans, but also the electric signal can be converted into a sound wave inaudible to humans for ranging and other purposes. In some embodiments, audio circuitry 1407 may also include a headphone jack.

The locating component 1408 is used to locate the current geographic location of the terminal 1400 to enable navigation or LBS (Location Based Service, location-based services).

A power supply 1409 is used to power the various components in terminal 1400. The power supply 1409 may be an alternating current, a direct current, a disposable battery, or a rechargeable battery. When the power supply 1409 includes a rechargeable battery, the rechargeable battery may support wired or wireless charging. The rechargeable battery may also be used to support fast charge technology.

In some embodiments, terminal 1400 also includes one or more sensors 1410. The one or more sensors 1410 include, but are not limited to: acceleration sensor 1411, gyro sensor 1412, pressure sensor 1413, optical sensor 1415, and proximity sensor 1416.

The acceleration sensor 1411 may detect the magnitudes of accelerations on three coordinate axes of a coordinate system established with the terminal 1400. For example, the acceleration sensor 1411 may be used to detect components of gravitational acceleration in three coordinate axes. The processor 1401 may control the touch display 1405 to display a user interface in a landscape view or a portrait view according to the gravitational acceleration signal acquired by the acceleration sensor 1411. The acceleration sensor 1411 may also be used for the acquisition of motion data of a game or a user.

The gyro sensor 1412 may detect a body direction and a rotation angle of the terminal 1400, and the gyro sensor 1412 may collect a 3D motion of the user to the terminal 1400 in cooperation with the acceleration sensor 1411. The processor 1401 may implement the following functions based on the data collected by the gyro sensor 1412: motion sensing (e.g., changing UI according to a tilting operation by a user), image stabilization at shooting, game control, and inertial navigation.

Pressure sensor 1413 may be disposed on a side frame of terminal 1400 and/or on an underlying layer of touch screen 1405. When the pressure sensor 1413 is provided at a side frame of the terminal 1400, a grip signal of the terminal 1400 by a user can be detected, and the processor 1401 performs right-and-left hand recognition or quick operation according to the grip signal collected by the pressure sensor 1413. When the pressure sensor 1413 is disposed at the lower layer of the touch screen 1405, the processor 1401 realizes control of the operability control on the UI interface according to the pressure operation of the user on the touch screen 1405. The operability controls include at least one of a button control, a scroll bar control, an icon control, and a menu control.

The optical sensor 1415 is used to collect the ambient light intensity. In one embodiment, the processor 1401 may control the display brightness of the touch screen 1405 based on the intensity of ambient light collected by the optical sensor 1415. Specifically, when the intensity of the ambient light is high, the display brightness of the touch display screen 1405 is turned up; when the ambient light intensity is low, the display brightness of the touch display screen 1405 is turned down. In another embodiment, the processor 1401 may also dynamically adjust the shooting parameters of the camera assembly 1406 based on the ambient light intensity collected by the optical sensor 1415.

A proximity sensor 1416, also referred to as a distance sensor, is typically provided on the front panel of terminal 1400. The proximity sensor 1416 is used to collect the distance between the user and the front of the terminal 1400. In one embodiment, when the proximity sensor 1416 detects that the distance between the user and the front surface of the terminal 1400 gradually decreases, the processor 1401 controls the touch display 1405 to switch from the bright screen state to the off screen state; when the proximity sensor 1416 detects that the distance between the user and the front surface of the terminal 1400 gradually increases, the touch display 1405 is controlled by the processor 1401 to switch from the off-screen state to the on-screen state.

Those skilled in the art will appreciate that the structure shown in fig. 14 is not limiting and that terminal 1400 may include more or less components than those illustrated, or may combine certain components, or employ a different arrangement of components.

In an exemplary embodiment, a computer-readable storage medium storing a computer program, for example, a memory storing a computer program, which when processed and executed implements the resource information presentation method shown in the above embodiment, is also provided. For example, the above-mentioned computer readable storage medium may be a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a compact disc Read-Only Memory (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, and the like.

Claims

1. The method for pushing the resource information is characterized by comprising the following steps:

inputting the resource information and the use preference information of the target user into a multi-stage model, outputting preference degree information of the target user for at least one use stage of the resource, wherein the multi-stage model is used for predicting preference degree information of the target user for different use stages of the resource according to content information of the different use stages of the resource, the multi-stage model comprises a plurality of sub-models of the use stages, each sub-model of the use stage is used for outputting semantic information of the corresponding use stage according to the content information of the corresponding use stage, and when the semantic information output by the sub-model of each use stage is matched with the use preference information, the sub-model of the next use stage is triggered to calculate the semantic information;

2. The method according to claim 1, wherein the inputting the resource information and the usage preference information of the target user into the multi-stage model, outputting preference degree information of the target user for at least one usage stage of the resource, comprises:

Inputting the content information of the current use stage into the sub-model of the current use stage in the multi-stage model, and outputting the semantic information of the current use stage;

calculating preference degree information of the current use stage according to semantic information and use preference information of the current use stage;

outputting preference degree information of the current use stage;

and when the preference degree information of the current use stage accords with a preset condition, inputting the content information of the next use stage into the submodel of the next use stage.

3. The method according to claim 2, wherein inputting the content information of the current usage stage into the sub-model of the current usage stage, outputting the semantic information of the current usage stage, comprises:

inputting the content information of the current use stage into a deep neural network model of a sub-model of the current use stage, and outputting a vector corresponding to the content information, wherein the deep neural network model is obtained by training the content information of the corresponding use stage;

and inputting the vector corresponding to the content information and the additional information of the current use stage into a linear model of a sub-model of the current use stage, outputting semantic information of the current use stage, and training the linear model according to the additional information of the corresponding use stage.

4. The method of claim 3, wherein the step of,

the linear model is a preset function, and the input parameters of the preset function comprise content information of the current use stage and use additional information; or alternatively, the first and second heat exchangers may be,

the linear model includes a plurality of candidate functions, and when the linear model calculates the content information of the current use stage, the candidate functions adopted are determined based on the content information of the current use stage and the use additional information.

5. The method of claim 1, wherein the training process of the multi-stage model comprises:

6. The method of claim 1, wherein the plurality of usage phases includes a first usage phase and a second usage phase, the first usage phase being a phase of displaying at least one of a title, a thumbnail, a summary of a resource, and the second usage phase being a phase of displaying a full-text of the resource.

7. The method according to any one of claims 1 to 6, wherein the multi-stage model comprises a plurality of sub-models of usage stages, each sub-model of usage stage further being adapted to output semantic information of a corresponding usage stage based on content information of the corresponding usage stage and usage additional information.

8. The method of claim 7, wherein the usage additional information includes at least one of an operation duration of a usage period, attribute information of the target user, irrational information, social information, impression information;

The irrational information comprises at least one of user emotion information, user health information, weather information, traffic information, date event information and labor state information, and the impression information is used for indicating the impression of the matching degree between the content information of the corresponding use stage of the provider of the resource information and the content information of the next use stage of the target user.

9. The method of claim 1, wherein the process of obtaining the usage preference information comprises:

10. The method according to claim 9, wherein before the obtaining the usage preference information of the target user according to the operation behavior of the target user on the content information of the at least first usage stage of the history resource, the method further comprises:

when the browsing behavior of the target user on the content information of any use stage of the historical resources is detected, acquiring the duration of the browsing behavior; and/or the number of the groups of groups,

when detecting the listening behavior of the target user to the content information in any use stage of the historical resources, acquiring the duration of the listening behavior; and/or the number of the groups of groups,

When the watching behavior of the content information of any using stage of the historical resources by the target user is detected, acquiring the duration of the watching behavior; and/or the number of the groups of groups,

when detecting a game behavior of content information of any use stage of a historical resource by the target user, acquiring at least one of duration, operation times and operation frequency of the game behavior; and/or the number of the groups of groups,

recording a confirmation action when the confirmation action of the target user on the content information of any use stage of the historical resources is detected, wherein the confirmation action refers to the action of confirming to enter the next use stage; and/or the number of the groups of groups,

and when the interaction behavior of the target user on the content information of any use stage of the historical resources is detected, recording the interaction behavior, wherein the interaction behavior refers to the behavior that the target user interacts with other users or the provider of the resource information.

11. The method of claim 10, wherein the recording the interaction behavior comprises:

recording comment content when comment behaviors of the target user on content information of any use stage of the historical resources are detected; and/or the number of the groups of groups,

When the sharing behavior of the target user on the content information of any use stage of the historical resources is detected, the sharing times or the number of the shared objects are recorded; and/or the number of the groups of groups,

recording the score of scoring behavior when the scoring behavior of the target user on the content information of any using stage of the historical resources is detected; and/or the number of the groups of groups,

recording the amount of the rewarding action when detecting the rewarding action of the target user on the content information of any using stage of the history resource; and/or the number of the groups of groups,

when the gift sending behavior of the target user to the content information of any using stage of the history resource is detected, recording at least one of the quantity, the type and the amount of the transmitted virtual gift; and/or the number of the groups of groups,

when detecting the purchasing behavior of the content information of any using stage of the historical resources by the target user, recording at least one of the quantity and the amount of the purchasing behavior; and/or the number of the groups of groups,

and recording the content of the transmitted barrage when the barrage transmitting behavior of the content information of any using stage of the historical resources by the target user is detected.

12. The method of claim 10, wherein the confirmation behavior comprises at least one of a click behavior, a slide behavior, a long press behavior, a voice confirmation behavior.

13. The method of any one of claims 1-6, wherein the resource is an article;

the content information of the first reading stage of the article comprises at least one of a title, a thumbnail and a abstract of the article, the content information of the second reading stage of the article is a pilot-viewing fragment of the article, and the content information of the third reading stage of the article is the full text of the article.

14. The method of any one of claims 1-6, wherein the resource is video;

15. The method of any one of claims 1-6, wherein the resource is music;

16. The method of any one of claims 1-6, wherein the resource is a commodity;

17. The method of any one of claims 1-6, wherein the resource is a game;

18. The method of claim 1, wherein pushing the resource information based on the at least one preference degree information comprises:

determining an order in which the resource information is pushed based on at least one preference degree information; and/or the number of the groups of groups,

determining at least one of the position, the sequence and the occupied page proportion of the resource information in the push page based on at least one preference degree information; and/or the number of the groups of groups,

based on at least one preference degree information, it is determined whether to push the resource information.

19. The method of claim 18, wherein pushing the resource information based on the at least one preference degree information comprises:

pushing resource information based on preference degree information of the first use stage; and/or the number of the groups of groups,

pushing resource information based on at least one statistic value of preference degree information, wherein the statistic value is a weighted sum value, an average value or a maximum value; and/or the number of the groups of groups,

And pushing the resource information based on the preference degree information of the designated using stage.

20. The method of claim 19, wherein the step of determining the position of the probe comprises,

the pushing the resource information based on the preference degree information of the first use stage comprises the following steps:

pushing content information of the first use stage of the resource when the preference degree information of the first use stage is larger than a first threshold value; or pushing the content information of the first use stage of the resource when the preference degree information of the first use stage is larger than a first threshold value and not larger than a second threshold value; or pushing content information of at least one use stage of the resource when the preference degree information of the first use stage is larger than a second threshold value, wherein the second threshold value is larger than the first threshold value; and/or the number of the groups of groups,

the step of pushing the resource information based on the statistical value of the at least one preference degree information comprises the following steps:

pushing content information of the first use stage of the resource when the statistic value is larger than a third threshold value; or pushing the content information of the first use stage of the resource when the statistical value is larger than a third threshold value and not larger than a fourth threshold value; or pushing content information of at least one using stage of the resource when the statistic value is larger than a fourth threshold value; and/or the number of the groups of groups,

The step of pushing the resource information based on the preference degree information of the appointed using stage comprises the following steps:

pushing content information of the appointed use stage of the resource when the preference degree information of the appointed use stage is larger than a fifth threshold value; or pushing content information of the specified use stage of the resource when the preference degree information of the specified use stage is larger than a fifth threshold value and not larger than a sixth threshold value; or pushing content information of at least one use stage of the resource when the preference degree information of the designated use stage is larger than a sixth threshold.

21. The method for pushing the resource information is characterized by comprising the following steps:

pushing the resource information based on the preference degree information;

The first stage model is used for predicting preference degree information of a user on the resource according to content information of a first use stage of the resource, the first stage model is similar to a sub-model of the first use stage in the multi-stage model, the multi-stage model is used for predicting preference degree information of a target user on different use stages of the resource according to content information of different use stages of the resource, the multi-stage model comprises a plurality of sub-models of the use stages, the sub-model of each use stage is used for outputting semantic information of the corresponding use stage according to content information of the corresponding use stage, and when the semantic information output by the sub-model of each use stage is matched with the preference information, the sub-model of the next use stage is triggered to calculate the semantic information.

22. The method of claim 21, wherein pushing the resource information based on the preference degree information comprises:

pushing content information of a first use stage of the resource when the preference degree information is larger than a first threshold value; or alternatively, the first and second heat exchangers may be,

pushing content information of a first use stage of the resource when the preference degree information is larger than a first threshold value and not larger than a second threshold value; or alternatively, the first and second heat exchangers may be,

And pushing content information of at least one using stage of the resource when the preference degree information is larger than a second threshold value, wherein the second threshold value is larger than the first threshold value.

23. The method of claim 21, wherein the operational behavior of the content information of the first usage stage refers to a browsing behavior of the content information of the first usage stage of the history resource, the content information of the first usage stage of the history resource including at least one of a title, a thumbnail, and a summary of the history resource.

24. The method of claim 21, wherein prior to obtaining the usage preference information of the target user based on the operational behavior of the target user with respect to the content information of the first usage stage of the history resource, the method further comprises:

and receiving the duration of the browsing behavior of the target user on the content information of the first use stage of the historical resources, which is sent by the terminal.

25. A method for displaying resource information, the method comprising:

when detecting the operation behavior of the target user on the content information of at least the first using stage of the historical resources, transmitting the operation behavior to a server, so that the server executes the following steps: acquiring the use preference information of the target user according to the operation behavior of the target user on the content information of the first use stage of the historical resources; acquiring resource information, wherein the resource information comprises content information of a plurality of different use stages of the resource; inputting the resource information and the usage preference information into a multi-stage model, outputting preference degree information of at least one usage stage of the resource by the target user, wherein the multi-stage model is used for predicting preference degree information of different usage stages of the resource by the target user according to content information of the different usage stages of the resource, the multi-stage model comprises a plurality of sub-models of the usage stages, each sub-model of the usage stage is used for outputting semantic information corresponding to the usage stage according to the content information of the corresponding usage stage, and when the semantic information output by each sub-model of the usage stage is matched with the usage preference information, the sub-model of the next usage stage is triggered to calculate the semantic information; pushing the resource to the target user based on at least one preference degree information;

displaying content information of a first use stage of the resource;

wherein the resource is determined by the server based on content information for a plurality of different usage phases of the historical resource and at least one preference degree information.

26. A server comprising a processor and a memory, wherein the memory has stored therein at least one instruction that is loaded and executed by the processor to implement the operations performed by the resource information pushing method of any of claims 1 to 20 or to implement the operations performed by the resource information pushing method of any of claims 21 to 24.

27. A terminal comprising a processor and a memory having at least one instruction stored therein, the instruction being loaded and executed by the processor to perform the operations performed by the resource information presentation method of claim 25.

28. A resource information pushing apparatus, characterized in that the apparatus comprises:

The input-output module is used for inputting the resource information and the use preference information of the target user into a multi-stage model and outputting preference degree information of at least one use stage of the resource by the target user, wherein the multi-stage model is used for predicting preference degree information of different use stages of the resource by the target user according to content information of the different use stages of the resource, the multi-stage model comprises a plurality of use stage sub-models, each use stage sub-model is used for outputting semantic information corresponding to the use stage according to the content information of the corresponding use stage, and when the semantic information output by each use stage sub-model is matched with the use preference information, the sub-model of the next use stage is triggered to calculate the semantic information;

29. The apparatus of claim 28, wherein the input-output module comprises:

the input/output sub-module is further used for outputting preference degree information of the current use stage;

and the input and output sub-module is also used for inputting the content information of the next use stage into the sub-model of the next use stage when the preference degree information of the current use stage accords with a preset condition.

30. The apparatus of claim 29, wherein the input-output sub-module is further configured to:

31. The apparatus of claim 30, wherein the device comprises a plurality of sensors,

32. The apparatus of claim 28, wherein the training process of the multi-stage model comprises:

33. The apparatus of claim 28, wherein the plurality of usage phases comprises a first usage phase and a second usage phase, the first usage phase being a phase of displaying at least one of a title, a thumbnail, a summary of a resource, and the second usage phase being a phase of displaying a full-text of the resource.

34. The apparatus of any one of claims 28 to 33, wherein the multi-stage model comprises a plurality of sub-models of usage stages, each sub-model of usage stage further configured to output semantic information of a corresponding usage stage based on content information of the corresponding usage stage and usage additional information.

35. The apparatus of claim 28, wherein the usage additional information includes at least one of an operation duration of a usage period, attribute information of the target user, irrational information, social information, impression information;

36. The apparatus of claim 28, wherein the process of obtaining the usage preference information comprises:

37. The apparatus according to claim 36, wherein before the obtaining the usage preference information of the target user according to the operation behavior of the target user on the content information of the at least first usage stage of the history resource, the obtaining module is further configured to:

38. The apparatus of claim 37, wherein the means for obtaining is further configured to:

39. The apparatus of claim 37, wherein the confirmation behavior comprises at least one of a click behavior, a slide behavior, a long press behavior, a voice confirmation behavior.

40. The apparatus of any one of claims 28-33, wherein the resource is an article;

41. The apparatus of any one of claims 28-33, wherein the resource is video;

42. The apparatus of any one of claims 28-33, wherein the resource is music;

43. The apparatus of any one of claims 28-33, wherein the resource is a commodity;

44. The apparatus of any one of claims 28-33, wherein the resource is a game;

45. The apparatus of claim 28, wherein the pushing module is configured to:

46. The apparatus of claim 45, wherein the pushing module is configured to:

47. The apparatus of claim 46, wherein the device comprises,

the pushing module is used for:

48. A resource information pushing apparatus, characterized in that the apparatus comprises:

49. The apparatus of claim 48, wherein the pushing module is configured to:

50. The apparatus of claim 48, wherein the operational behavior of the content information of the first usage stage refers to browsing behavior of the content information of the first usage stage of the history resource, the content information of the first usage stage of the history resource including at least one of a title, a thumbnail, and a summary of the history resource.

51. The apparatus of claim 48, further comprising:

and the receiving module is used for receiving the duration time of the browsing behavior of the target user on the content information of the first use stage of the historical resources, which is sent by the terminal.

52. A resource information display device, the device comprising:

the sending module is used for sending the operation behavior to the server when detecting the operation behavior of the target user on the content information of at least the first use stage of the historical resources, so that the server executes the following steps: acquiring the use preference information of the target user according to the operation behavior of the target user on the content information of the first use stage of the historical resources; acquiring resource information, wherein the resource information comprises content information of a plurality of different use stages of the resource; inputting the resource information and the usage preference information into a multi-stage model, outputting preference degree information of at least one usage stage of the resource by the target user, wherein the multi-stage model is used for predicting preference degree information of different usage stages of the resource by the target user according to content information of the different usage stages of the resource, the multi-stage model comprises a plurality of sub-models of the usage stages, each sub-model of the usage stage is used for outputting semantic information corresponding to the usage stage according to the content information of the corresponding usage stage, and when the semantic information output by each sub-model of the usage stage is matched with the usage preference information, the sub-model of the next usage stage is triggered to calculate the semantic information; pushing the resource to the target user based on at least one preference degree information;