CN116628235B - Data recommendation method, device, equipment and medium - Google Patents

Data recommendation method, device, equipment and medium Download PDF

Info

Publication number
CN116628235B
CN116628235B CN202310889981.9A CN202310889981A CN116628235B CN 116628235 B CN116628235 B CN 116628235B CN 202310889981 A CN202310889981 A CN 202310889981A CN 116628235 B CN116628235 B CN 116628235B
Authority
CN
China
Prior art keywords
data
recommendation
information
recommended
user
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202310889981.9A
Other languages
Chinese (zh)
Other versions
CN116628235A (en
Inventor
范谨麒
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Alipay Hangzhou Information Technology Co Ltd
Original Assignee
Alipay Hangzhou Information Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Alipay Hangzhou Information Technology Co Ltd filed Critical Alipay Hangzhou Information Technology Co Ltd
Priority to CN202310889981.9A priority Critical patent/CN116628235B/en
Publication of CN116628235A publication Critical patent/CN116628235A/en
Application granted granted Critical
Publication of CN116628235B publication Critical patent/CN116628235B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/40—Information retrieval; Database structures therefor; File system structures therefor of multimedia data, e.g. slideshows comprising image and additional audio data
    • G06F16/43—Querying
    • G06F16/435—Filtering based on additional data, e.g. user or group profiles
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

The specification provides a data recommendation method, device, equipment and medium. The server determines a plurality of recommendation data from the total candidate recommendation data based on real-time feedback information of the user on the terminal, so that the recommendation data acquired from the server are real-time, sequencing results of the plurality of recommendation data and recommendation information of each recommendation data are acquired from the server, the plurality of recommendation data are reordered based on the recommendation information of each recommendation data, the real-time feedback information and the application platform type corresponding to the terminal, the application platform type information of the terminal is fully utilized, inaccurate sequencing results caused by distribution differences of the data on different types of application platforms are avoided, real-time adjustment of the display sequence of the recommendation data is realized, the plurality of recommendation data are displayed according to the recommendation sequencing results, the display sequence of the recommendation data is ensured to be more in line with current interest preference of the user, and the accuracy of data recommendation is improved.

Description

Data recommendation method, device, equipment and medium
Technical Field
One or more embodiments of the present disclosure relate to the field of computer technologies, and in particular, to a data recommendation method, apparatus, device, and medium.
Background
With the rapid development of internet technology, the number of internet enterprises and internet users has been rapidly increased, and at the same time, the amount of data in a network environment has been greatly increased, and in this case, both the internet enterprises and the internet users face a great challenge. For an internet enterprise as a data producer, how to make the data produced by the enterprise stand out from mass data and receive the attention of more users is a very difficult matter; it is also a very difficult matter for internet users, which are data consumers, to find data of interest to themselves from mass data.
In order to help users to quickly screen out interesting information from massive network data, data recommendation technology has been developed. Data recommendation techniques may enable an association between a user and data to recommend data to a user who may be interested in it. However, the current data recommendation technology mainly carries out recommendation based on historical preference of the user, and the display sequence of the recommended data often cannot meet the interest preference of the user, so that the recommendation effect is poor.
Disclosure of Invention
In view of this, one or more embodiments of the present disclosure provide a data recommendation method, apparatus, device, and medium.
In order to achieve the above object, one or more embodiments of the present disclosure provide the following technical solutions:
according to a first aspect of one or more embodiments of the present disclosure, a data recommendation method is provided and applied to a terminal, where the method includes:
acquiring a plurality of recommendation data, a sequencing result of the plurality of recommendation data and recommendation information of each recommendation data, wherein the recommendation information of the plurality of recommendation data and each recommendation data is determined and obtained from the total candidate recommendation data by a server based on real-time feedback information of a user on a terminal;
based on the recommendation information and the real-time feedback information of each recommendation data and the type of the application platform corresponding to the terminal, reordering the plurality of recommendation data to obtain recommendation ordering results of the plurality of recommendation data on the application platform corresponding to the terminal;
and displaying the plurality of recommended data according to the recommended sequencing result.
According to a second aspect of one or more embodiments of the present specification, there is provided a data recommendation device, applied to a terminal, the device comprising:
the acquisition module is used for acquiring a plurality of recommendation data, a sequencing result of the plurality of recommendation data and recommendation information of each recommendation data, wherein the recommendation information of the plurality of recommendation data and each recommendation data is determined and obtained from the total candidate recommendation data by the server based on real-time feedback information of a user on the terminal;
The sequencing module is used for reordering the plurality of recommended data based on the recommended information of each recommended data, the real-time feedback information and the application platform type corresponding to the terminal so as to obtain a recommended sequencing result of the plurality of recommended data on the application platform corresponding to the terminal;
and the display module is used for displaying the plurality of recommended data according to the recommended sequencing result.
According to a third aspect of one or more embodiments of the present specification, there is provided a terminal comprising:
a processor;
a memory for storing processor-executable instructions;
wherein the processor implements the method as described in the first aspect above by executing executable instructions.
According to a fourth aspect of one or more embodiments of the present description, there is provided a computer readable storage medium having stored thereon computer instructions which, when executed by a processor, implement a method as described in the first aspect above.
According to the method and the system, the server determines the plurality of recommendation data from the total candidate recommendation data based on the real-time feedback information of the user on the terminal, so that the recommendation data acquired from the server are real-time and more in line with the current interest preference of the user, and the ranking results of the plurality of recommendation data and the recommendation information of each recommendation data are acquired from the server, so that the plurality of recommendation data are reordered based on the recommendation information of each recommendation data, the real-time feedback information and the application platform type corresponding to the terminal, the application platform type information of the terminal is utilized more fully, inaccurate ranking results caused by the distribution difference of the data on different types of application platforms are avoided, the recommendation ranking results of the plurality of recommendation data on the application platform corresponding to the terminal are obtained, the real-time adjustment of the display sequence of the recommendation data is realized, the recommendation data can be displayed according to the recommendation ranking results, the display sequence of the recommendation data is more in line with the current interest preference of the user is guaranteed, and the accuracy of data recommendation is improved.
Drawings
Fig. 1 is a flowchart of a data recommendation method according to an exemplary embodiment.
Fig. 2 is a graph of a double-ended total play provided by an exemplary embodiment.
Fig. 3 is a graph of a double-ended average person play volume provided by an exemplary embodiment.
Fig. 4 is a graph of a double-ended total play duration provided by an exemplary embodiment.
Fig. 5 is a graph of a double-ended average play duration provided by an exemplary embodiment.
Fig. 6 is a schematic diagram of a model architecture of a hybrid expert model according to an exemplary embodiment.
FIG. 7 is a flow chart of a video recommendation process provided by an exemplary embodiment.
Fig. 8 is a schematic block diagram of a terminal according to an exemplary embodiment.
Fig. 9 is a block diagram of a data recommendation device according to an exemplary embodiment.
Detailed Description
Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with one or more embodiments of the present specification. Rather, they are merely examples of apparatus and methods consistent with aspects of one or more embodiments of the present description as detailed in the accompanying claims.
It should be noted that: in other embodiments, the steps of the corresponding method are not necessarily performed in the order shown and described in this specification. In some other embodiments, the method may include more or fewer steps than described in this specification. Furthermore, individual steps described in this specification, in other embodiments, may be described as being split into multiple steps; while various steps described in this specification may be combined into a single step in other embodiments.
User information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, presented data, etc.) referred to in this specification are both information and data authorized by the user or sufficiently authorized by the parties, and the collection, use and processing of relevant data requires compliance with relevant laws and regulations and standards of the relevant country and region, and is provided with corresponding operation portals for the user to choose authorization or denial.
The specification provides a data recommendation method, which is used for ensuring the instantaneity of the acquired recommendation data so as to ensure that the acquired recommendation data better accords with the current interest preference of a user, and can reorder the acquired recommendation data when the user triggers a data reordering instruction. When the data are reordered, the application platform type information of the terminal can be fully utilized, inaccurate ordering results caused by distribution differences of the data on different types of application platforms are avoided, the obtained recommendation ordering results are more matched with the data distribution of the current application platform, the accuracy of the determined recommendation ordering results is improved, when the recommendation ordering results are used for displaying the recommendation data, the recommendation data which are possibly interested by a user can be displayed at a position which is more forward, and therefore the display sequence of the recommendation data can be guaranteed to be more in line with the current interest preference of the user, and the accuracy of data recommendation is improved.
The recommended data may be different types of data, for example, the recommended data may be multimedia data, or the recommended data may be a page delivery application, etc., and the data type of the recommended data is not limited in this specification.
Taking the recommendation data as multimedia data as an example, the multimedia data may be video data, audio data, image data, etc., and the specific type of the multimedia data is not limited in this specification. Taking multimedia data as video data as an example, the data recommendation method provided in the present specification may be used to provide a video recommendation function for a short video function in a video playing application (such as a short video application) or a payment application. By the data recommendation method provided by the specification, the real-time adjustment of the video recommendation sequence can be realized on the terminal side by utilizing the real-time feedback information of the user on various video data, so that the user is promoted to watch more videos, the user experience can be improved on one hand, the on-terminal resources can be utilized to the greatest extent on the other hand, and the video data displayed in front of the user can be the video data of interest to the user, so that the loss of the user and the waste of video resources are avoided.
Taking the recommended data as an example of the page delivery application, the application function to be delivered in the functional page (such as the application home page) of the application program can be determined by the data recommendation method provided by the specification, alternatively, the application program can be a plurality of application programs such as a payment application, a life service application and the like, and the specific type of the application program is not limited in the specification. When the application function is put in the function page of the application program, the application function can be put in a corner mark form. In the related art, after an application function is launched into a page in the form of a corner mark, a terminal adjusts the launched application according to the fatigue degree principle, namely if a user does not click on the corner mark of the launched application, the corner mark is always exposed, so that the launched corner mark is exposed for a plurality of times, but the user does not consider the condition of click-through conversion to cause the waste of launched resources, and through the data recommendation method provided by the specification, a rearrangement mechanism can be introduced at the terminal side, user feedback is perceived in real time, the launched application is timely adjusted according to the perception result, the user click is promoted, and the condition of long-tailed exposure is reduced.
The data recommendation method may be executed by a terminal, where the terminal may be a desktop computer, a portable computer, a notebook computer, a smart phone, a tablet computer, a smart watch, or the like, and optionally, the terminal may be another type of device, and the device type of the terminal is not limited in this specification.
The foregoing is merely exemplary descriptions about application scenarios of the present specification, and does not limit the application scenarios of the present specification, and in more possible implementations, the solution provided in the present specification may be applied to other more scenarios, and the present specification does not limit specific application scenarios.
After the application scenario of the present specification is described, a description is next given of a specific implementation procedure of the present specification.
Referring to fig. 1, fig. 1 is a flowchart of a data recommendation method provided in an exemplary embodiment, where the data recommendation method may be performed by a terminal, as shown in fig. 1, and the method includes:
and 101, acquiring a plurality of recommendation data, a sequencing result of the plurality of recommendation data and recommendation information of each recommendation data, wherein the recommendation information of the plurality of recommendation data and each recommendation data is determined and obtained from the total candidate recommendation data by the server based on real-time feedback information of a user on the terminal.
It should be noted that, when there is a data recommendation requirement, the terminal may acquire a plurality of recommendation data and a sorting result of the plurality of recommendation data from the server, so that the plurality of recommendation data may be displayed based on the acquired sorting result. The plurality of recommendation data and the sequencing result of the plurality of recommendation data are determined by the service based on the real-time feedback information of the user, so that the recommendation data and the sequencing result acquired from the server are guaranteed to be stronger in real-time performance and more accord with the current interest preference of the user. Alternatively, the score of each recommended data may be returned by the server, and for any recommended data, the higher the score, the higher the likelihood that the recommended data is of interest to the user, the lower the score, the lower the likelihood that the recommended data is of interest to the user, so that the terminal can determine the ranking result of each recommended data based on the score of each recommended data.
Based on the obtained recommendation data with stronger real-time performance and the sequencing result, recommendation information of the recommendation data can be obtained, so that after the recommendation data is displayed based on the obtained data, the recommendation data can be reordered based on the recommendation information and other more information, the display sequence of the recommendation data can be timely adjusted according to real-time feedback of a user and other real-time data of a terminal, and the accuracy of the display sequence of the recommendation result is improved.
And 102, reordering the plurality of recommended data based on the recommended information of each recommended data, the real-time feedback information and the application platform type corresponding to the terminal to obtain a recommended sequencing result of the plurality of recommended data on the application platform corresponding to the terminal.
Generally, the types (or versions) of applications that need to be installed by terminals that mount different operating systems will be different for the same application, and thus, the type of application platform that the terminal corresponds to may be determined based on the type of operating system that the terminal mounts. For example, the operating system carried by the terminal may be an Android (Android) system or a iOS (iPhone OS) system, and correspondingly, the type of the application platform corresponding to the terminal may also be an Android system or an iOS system.
Through data analysis, the performance difference of the same kind of data on different types of application platforms is larger, and the description considers the type of the application platform corresponding to the terminal when reordering the recommended data, so that the reordered sequencing result can more meet the current use situation of the terminal, and the sequencing accuracy is improved.
And 103, displaying the plurality of recommended data according to the recommended sequencing result.
According to the method and the system, the server determines the plurality of recommendation data from the total candidate recommendation data based on the real-time feedback information of the user on the terminal, so that the recommendation data acquired from the server are real-time and more in line with the current interest preference of the user, and the ranking results of the plurality of recommendation data and the recommendation information of each recommendation data are acquired from the server, so that the plurality of recommendation data are reordered based on the recommendation information of each recommendation data, the real-time feedback information and the application platform type corresponding to the terminal, the application platform type information of the terminal is utilized more fully, inaccurate ranking results caused by the distribution difference of the data on different types of application platforms are avoided, the recommendation ranking results of the plurality of recommendation data on the application platform corresponding to the terminal are obtained, the real-time adjustment of the display sequence of the recommendation data is realized, the recommendation data can be displayed according to the recommendation ranking results, the display sequence of the recommendation data is more in line with the current interest preference of the user is guaranteed, and the accuracy of data recommendation is improved.
The foregoing is merely an introduction to a basic implementation procedure related to the present specification, and a detailed description of a data recommendation method provided in the present specification is provided below in connection with an alternative embodiment of the present specification.
It should be noted that, the terminal may detect the operation of the user on the terminal in real time, so as to trigger a recommendation process of obtaining the plurality of recommendation data, the sorting result of the plurality of recommendation data, and the recommendation information of each recommendation data when the preset operation is detected. The preset operation for triggering the recommendation process may be different for different types of data, and the preset operation for triggering the recommendation process in the application scene may be determined according to the actual application scene of the different types of data.
For example, the scene in which data recommendation is required may be a multimedia data browsing scene (such as a short video browsing scene), where the data to be recommended is multimedia data, the preset operation may be a multimedia data switching operation, that is, a process of acquiring a plurality of recommended data, a sorting result of the plurality of recommended data, and recommendation information of each recommended data may be triggered when the user switches the multimedia data currently being played.
Taking a short video browsing scene as an example, a multimedia data browsing scene needing to be subjected to data recommendation, in the short video browsing scene, a user can switch a currently playing short video in a video sliding manner, that is, can respond to the operation of the user for sliding the video, and trigger a process of acquiring a plurality of recommended data, a sequencing result of the plurality of recommended data and recommendation information of each recommended data.
For another example, the scene that needs to be recommended for data may be an application function page display scene, where a function page of an application (e.g., a target page in an application, such as an application top page) may display a plurality of corner marks corresponding to an application function, and a user may trigger a corner mark corresponding to a certain application function to implement use of a corresponding application function. For an application program, more application functions may be included, and the corner mark corresponding to each application function cannot be displayed in the function page, and at this time, the recommended application function needs to be determined, so that the corner mark of the recommended application function can be displayed in the function page; in addition, for an application program, even if the functional page can completely display the corner marks corresponding to all application functions, the display sequence of the corner marks corresponding to the application functions is also considered, so that the corner marks of the application functions possibly interested by the user are preferentially displayed, the use experience of the user can be improved, and the utilization efficiency of page display resources can be improved.
In a scenario that a corner mark corresponding to an application function needs to be put on a target page of an application program, the application program can comprise a plurality of functional pages, a user can skip among the plurality of functional pages to view different functional pages, at this time, data to be recommended can be put on the page, a preset operation can be an operation that the user returns the target page, optionally, the user returns the target page can comprise two cases that the user enters the target page of the application program for the first time and skips from other functional pages of the application program to the target page, that is, when the user enters the target page of the application program for the first time or skips from other functional pages of the application program to the target page, a process of acquiring a plurality of recommended data, a sequencing result of the recommended data and recommendation information of each recommended data can be triggered.
In summary, for step 101, when obtaining the plurality of recommended data, the sorting result of the plurality of recommended data, and the recommended information of each recommended data, there may be two possible trigger occasions as follows:
in one possible implementation manner, in the case that the data to be recommended is multimedia data, a plurality of recommendation data, a sorting result of the plurality of recommendation data and recommendation information of each recommendation data are acquired in response to a multimedia data switching operation of a user.
In another possible implementation manner, in the case that the data to be recommended is a page delivery application, the plurality of recommended data, the sorting result of the plurality of recommended data and the recommendation information of each recommended data are obtained in response to the operation of returning the target page by the user.
The above embodiments mainly describe the timing of triggering the acquisition of the plurality of recommended data, the ranking result of the plurality of recommended data, and the recommendation information of each recommended data, and the detailed process of acquiring the plurality of recommended data, the ranking result of the plurality of recommended data, and the recommendation information of each recommended data will be described below.
In some embodiments, for step 101, when obtaining the plurality of recommended data, the sorting result of the plurality of recommended data, and the recommendation information of each recommended data, the following steps may be implemented:
Step 1011, generating a data recommendation request based on the real-time feedback information.
Optionally, in the case that the data to be recommended is multimedia data, a data recommendation request may be generated based on real-time feedback information in response to a multimedia data switching operation of the user; and under the condition that the data to be recommended is a page delivery application, responding to the operation of returning the target page by the user, and generating a data recommendation request based on the real-time feedback information.
It should be noted that, the terminal may collect the user operation on the terminal in real time, so as to generate real-time feedback information based on the collected operation, so that the generation of the data recommendation request may be performed based on the real-time feedback information, so that the generated data recommendation request may include the real-time feedback information.
It should be noted that different types of data may correspond to different real-time feedback information. Taking recommendation data as multimedia data and page delivery application as an example, and in the case that the recommendation data is multimedia data, the real-time feedback information comprises at least one of user real-time preference information, data browsing condition information, user preference change information, browsing condition change information and terminal characteristic information; and under the condition that the recommended data is a page delivery application, the real-time feedback information comprises real-time browsing condition information and user state prediction information of the user.
The real-time feedback information corresponding to the different types of recommended data is described in detail below.
In the case that the recommendation data is multimedia data, the user real-time preference information may be a data type of user preference predicted based on real-time feedback of the user; the data browsing condition information can be the degree of interest predicted based on the browsing time of the user on the browsed recommended data; the user preference change information may be difference information determined based on the ranking result of the plurality of recommendation data and the user real-time preference information; the browsing condition change information may include at least one of difference information determined based on recommended data currently browsed by the user and browsed recommended data, and difference information determined based on a current time and a time period for the user to browse the recommended data; the terminal characteristic information may include at least one of a cached length of each recommended data on the terminal, power information of the terminal, and memory information of the terminal.
The user real-time preference information may be predicted by the feedback information of the user on each multimedia data that has been browsed currently, that is, by predicting what type of multimedia data the user may prefer to watch through the feedback information, and optionally, the feedback information may include a playing duration, a playing proportion, whether to perform forward feedback (such as praise, comment, collection, attention, sharing, etc.) of each browsed multimedia data.
When the data browsing condition information is acquired, the method can determine how many multimedia data are currently browsed by the user according to the real-time condition of the user browsing the multimedia data, and can determine the interested degree of the user on the browsed video by combining the browsing time length of the user. For example, a user quickly browses 10 videos, which indicates that the user may not be interested in the video types of the 10 videos, does not watch the video content carefully at all, or switches videos immediately after the user brushes the videos to cause the user to run off, at this time, in addition to the videos consistent with the 10 video types, more importantly, some popular videos need to be pushed to stay the user, so that the user experience is improved.
In a real-time scenario, the real-time preference information of the user and the ranking result of the plurality of recommended videos pushed by the server are not matched, or the real-time preference information of the user and the score of each recommended video by the server may not be proportional, for example, the user may be interested in a certain video, but the score of the video by the server may be low, so that the ranking result of the video is relatively rear, which not only results in poor recommendation accuracy, but also adversely affects the subsequent reordering process. In this case, the change of the real-time interests of the user can be captured by sensing the offset characteristics between the real-time preference information of the user and the server score as the user preference change information, so that the real-time property and accuracy of the ranking result can be ensured.
For obtaining the browsing condition change information, the difference information determined based on the currently browsed recommendation data and the browsed recommendation data of the user may be the difference information determined based on the position of the currently browsed recommendation data in the history playlist formed by the browsed recommendation data and the position of the currently browsed recommendation data in the current playlist, for example, the difference information may be the position difference value of the currently browsed recommendation data in the history playlist and the current playlist, and through the information, the overall influence weight of the recommendation data recently played by the user on the recommendation data to be rearranged later may be perceived, so as to improve the accuracy of the reordering process. The difference information determined based on the current time and the time length of browsing the recommended data by the user can be the deviation between the current time and the time length of browsing the recommended data by the user, and is used for sensing the influence weight of the time factor on the historical playing video.
For the terminal characteristic information such as the cache length of the recommended data, each time the terminal pulls the recommended multimedia data from the server, each multimedia data is preloaded (namely, a part of each multimedia data is cached in advance), but the cache lengths of different multimedia data are different due to the influence of various factors such as the current network signal strength of a user and the data quantity of the whole multimedia data to be loaded, under the condition that the current network signal strength of the user is poor, if the user recommends the multimedia data with the short cache length, the rest uncached part can not be successfully loaded under the current network environment, so that the playing of the multimedia data is blocked, and the user is lost.
And for terminal characteristic information such as terminal electric quantity information and terminal memory information, under the condition of higher terminal electric quantity and memory pressure (namely, under the condition of lower terminal electric quantity and larger memory occupation), if the multimedia data with higher consumption of resources is pushed, the terminal electric quantity and the memory pressure can be further increased, so that the normal use of the terminal by a user is influenced, at the moment, the multimedia data with lower consumption of resources can be pushed preferentially so as to ensure that the user can still use the terminal more smoothly, and the user experience is improved.
Under the condition that the recommended data is a page delivery application, the real-time browsing condition information of the user comprises at least one of an application sequence accessed by the user in a target page and an application sequence accessed by the user in a set page area of the target page; the user state prediction information is a user state determined based on the sensing data acquired by the terminal sensor.
The application sequence accessed by the user in the target page may be an application sequence (may be referred to as a global application sequence) determined based on the application access behavior of the user in the global domain of the target page (i.e., the entire target page). The user can interact frequently with the application program when using the application program, the behavior habit of the user in the current context can be perceived through various interaction behaviors, the conversion of which step is usually carried out by the user after each interaction is perceived, for example, in the application program capable of providing various life services, when the user has a wish to purchase a movie ticket, the user is enhanced in the wish to purchase the movie ticket by putting a corner mark of an application function capable of providing the movie ticket purchasing capability in a target page, and the click conversion of the user is promoted; on the other hand, when the user is in an idle state, according to the determined universal application sequence, application marks which are not paid attention to before the user is put in the user, and the user is helped to open a new field of view by using the diversity of the marks so as to promote the permeability of the application.
The application sequence that the user accesses in the set page area of the target page may be an application sequence (may be referred to as an area application sequence) determined based on the application access behavior of the user in the set page area of the target page. For a page, the page may include a plurality of functional areas, the display positions of different functional areas in the page are different, the possibility that the functional areas with more obvious display positions are focused by a user is higher, the areas with higher focusing possibility can be used as set page areas, the user access behaviors in the set page areas can be prioritized, and the application sequences are formed by utilizing the applications and the access sequence of clicking access of the user in the set page areas, so that the capturing of the real-time behavior preferences of the user is realized. For example, if the user accesses an application function of a public welfare class such as shopping, the user may be interested in the public welfare class application at this time, and the user may be considered to be pushed with more corner marks of the public welfare class application.
Alternatively, the terminal sensor may be a terminal gyroscope for detecting a terminal attitude. By acquiring the terminal gesture, whether the user is currently playing games, paying and the like by using the mobile phone can be determined, so that whether the user is currently idle or not can be determined, and when the user is idle, some popular application corner marks with high conversion possibility can be pushed to the user, and the exposure rate and conversion rate of the corner marks are increased.
The foregoing is merely a few exemplary descriptions of real-time feedback information, and in many possible implementations, the real-time feedback information may include many more types of information, and the specific types of real-time feedback information are not limited in this specification.
Step 1012, a data recommendation request is sent to a server, and the server is used for determining a plurality of recommendation data from the total candidate recommendation data based on the data recommendation request, and determining recommendation information of the plurality of recommendation data so as to return the plurality of recommendation data and recommendation information of each recommendation data to the terminal.
It should be noted that, the cloud may store a plurality of published candidate recommendation data, which is used as a data source in the data recommendation process. Optionally, a fine-ranking model may be deployed on the server, and is configured to obtain recommendation data actually to be pushed to the user from the total candidate recommendation data stored in the cloud, and rank the obtained recommendation data.
Optionally, when the server receives the data recommendation request, the server may input real-time feedback information carried by the received data recommendation request and the plurality of candidate recommendation data stored in the cloud end into the fine-ranking model, so as to output, through the fine-ranking model, the plurality of recommendation data and the score of each recommendation data, which are ordered before the preset position.
The fine-ranking model may be a Click-Through Rate (CTR) model, or the fine-ranking model may be another type of machine learning model, and the specific type of the fine-ranking model is not limited in this specification.
It should be noted that after determining the scores of the respective recommended data, the plurality of recommended data may be ranked according to the order of the determined scores from top to bottom, so as to obtain the ranking result of the plurality of recommended data.
In addition, the cloud end may further store recommendation information of each candidate recommendation data, and after determining recommendation data from the total candidate recommendation data, the cloud end may further acquire recommendation information of each recommendation data, so that recommendation information of each recommendation data may be sent to the terminal when the recommendation data and the sorting result of each recommendation data are sent to the terminal.
It should be noted that different types of data may correspond to different recommendation information. Taking recommendation data as multimedia data and page delivery application as an example, and under the condition that the recommendation data is multimedia data, the recommendation information comprises recommendation data time length, recommendation data type, recommendation data score and statistical characteristic values determined based on browsed conditions of the recommendation data; and under the condition that the recommendation data is a page delivery application, the recommendation information comprises a recommendation data identifier, a recommendation data type and a statistical characteristic value determined based on the clicked condition of the recommendation data.
In the case that the recommended data is multimedia data, the recommended data duration is the duration of the multimedia data, the recommended data type is the multimedia data type (taking the multimedia data as video data as an example, the recommended data type may include life video, science popularization video, entertainment video, and the like), the recommended data score may include an operation score and a model score, the operation score may be a score (including a quality score, an author, and the like) set by a professional for the recommended data, the model score is a score determined by a fine ranking model, the statistical feature value determined based on the browsed condition of the recommended data may be a popular degree of the multimedia data determined based on the historical playing condition such as the total duration, the praise amount, and the like of the recommended data played in a preset time period, and the preset time period may be the past week, the past half month, and the like.
In the case where the recommended data is a page delivery application, the recommended data identifier may be an application Identifier (ID), and the recommended data type may be an application category ID. The statistical characteristic value determined based on the clicked condition of the recommended data can comprise the clicked times of a certain application for the full-quantity user and the clicked times of the application for the certain user, and the hot degree of the application can be determined based on the clicked times of the certain application for the full-quantity user so as to realize the identification of the hot application; based on the clicked times of a certain application to a certain user, whether the application is a common application of the user can be determined, and the corresponding common application is different for different users.
Step 1013, receiving the plurality of recommended data returned by the server, the sorting result of the plurality of recommended data, and the recommendation information of each recommended data.
After the plurality of recommended data, the sorting result of the plurality of recommended data, and the recommended information of each recommended data are obtained through the above embodiment, the plurality of recommended data may be reordered according to the recommended information of each recommended data, the real-time feedback information, and the application platform type corresponding to the terminal through step 102, so as to obtain the recommended sorting result of the plurality of recommended data on the application platform corresponding to the terminal.
It should be noted that, according to the data analysis, the data amount and the data distribution of the same application program in different application platforms at present have a large difference, taking the Android system and the iOS system as the application platforms as examples, the overall play amount (or called total play amount), the average play amount, the total play duration, and the performance of the average play duration in the two ends of the video play application may be referred to the graphs shown in fig. 2 to 5, fig. 2 is a graph of the total play amount in the two ends provided by an exemplary embodiment, fig. 3 is a graph of the average play amount in the two ends provided by an exemplary embodiment, fig. 4 is a graph of the total play duration in the two ends provided by an exemplary embodiment, and fig. 5 is a graph of the average play duration in the two ends provided by an exemplary embodiment.
It can be seen that the data expression of the same application program in the Android system and the iOS system is greatly different, and the Android system and the iOS system are only two exemplary application platforms, and the data expression difference may be larger under the condition that the application platforms are more.
In view of this, to enable accurate predictions across different application platforms, a hybrid expert (Mixture of Experts, MOE) model may be employed to reorder the plurality of recommended data.
The mixed Expert model may include a plurality of Expert networks (Expert), a Gate network, and a multi-layer perceptron (Multilayer Perceptron, MLP), where the plurality of Expert networks correspond to different application platform types, and for any Expert network in the mixed Expert model, the Expert network may be configured to determine ranking characteristics of the plurality of recommended data on the application platform corresponding to the Expert network. Taking an application platform as an Android system and an iOS system as an example, the hybrid expert model may include two expert networks, where one expert network is used to determine ranking features of the plurality of recommended data in the Android system, and the other expert network is used to determine ranking features of the plurality of recommended data in the iOS system.
Alternatively, the hybrid expert model may be pre-trained by the server based on a training sample set. The training sample set may include a plurality of training samples, each of which may include one sample data, scores of the sample data on different application platforms, and user feedback information, so that model training may be performed based on the obtained training sample set to obtain a trained hybrid expert model.
After model training is completed, the trained mixed expert model can be issued to the terminal, so that the terminal can realize the reordering processing of a plurality of recommended data directly through the mixed expert model deployed at the local end when reordering requirements exist.
In some embodiments, for step 102, when reordering the plurality of recommended data based on the recommended information of each recommended data, the real-time feedback information, and the application platform type corresponding to the terminal to obtain the recommendation ordering result of the plurality of recommended data on the application platform corresponding to the terminal, the following steps may be implemented:
and 1021, processing the plurality of recommended data respectively through a plurality of expert networks included in the mixed expert model based on the recommended information and the real-time feedback information of each recommended data to obtain a plurality of sequencing features.
Alternatively, a feed forward neural network (Feed Forward Networks, FFN) may be employed as the expert network, or other types of neural networks may be employed as the expert network, which is not limited in this specification.
It should be noted that, for different expert networks in the same hybrid expert model, the network type and the network structure are the same, but the network parameters used in the different expert networks are different.
The acquisition of the ordering characteristics of the same recommended data on different application platforms can be realized through a plurality of expert networks included in the mixed expert model.
Step 1022, determining, from the plurality of ranking features, a target ranking feature matching the application platform type corresponding to the terminal through a gating network included in the hybrid expert model.
In one possible implementation manner, a weight may be allocated to each expert network in the hybrid expert model through a gating network included in the hybrid expert model based on an application platform type corresponding to the terminal; and based on the weights corresponding to the expert networks, carrying out weighted summation processing on the ordering features output by the expert networks to obtain the target ordering features.
Optionally, the weight of the expert network with the corresponding application platform type consistent with the application platform type corresponding to the terminal is set to be 1 through a gating network included in the hybrid expert model, and the weight of the expert network with the corresponding application platform type consistent with the application platform type corresponding to the terminal is set to be 0, so that multi-terminal selection output is realized, the final output target ordering characteristic is ensured to be consistent with the data distribution characteristic of the application platform corresponding to the terminal, and the problem of inconsistent double-terminal data distribution is solved through one model.
Step 1023, processing the target ordering feature through a multi-layer perceptron included in the mixed expert model to obtain a recommended ordering result.
In one possible implementation, the target ranking features may be processed by a multi-layer perceptron included in the hybrid expert model to obtain a re-score for each of the recommended data, such that the plurality of recommended data may be reordered based on the re-score for each of the recommended data.
Referring to fig. 6, fig. 6 is a schematic diagram of a model architecture of a hybrid Expert model provided by an exemplary embodiment, taking an example that an application platform includes an Android system and an iOS system, where the hybrid Expert model may include two Expert networks, one Expert network corresponds to the Android system (denoted as Android Expert), the other Expert network corresponds to the iOS system (denoted as iOSExpert), and recommendation information and real-time feedback information of recommendation data may be simultaneously input into the two Expert networks, and each Expert network may output a ranking feature, so that different weights may be given to ranking features output by the two Expert networks through a gating network, so that weighted summation may be performed on the two ranking features based on the given weights, so as to implement double-end selection output, obtain a target ranking feature matched with the application platform corresponding to the terminal, and thus may implement reordering based on the target ranking feature through MLP.
For easy understanding, the following describes a complete flow of the data recommendation method provided in the present specification by taking a video recommendation scenario as an example. Referring to fig. 7, fig. 7 is a schematic flow chart of a video recommendation process according to an exemplary embodiment, as shown in fig. 7, a hybrid expert network (that is, xNN export) may be trained in advance in the cloud, and the trained hybrid expert network may be deployed to an end (that is, a terminal). When the terminal has data recommendation requirements, online reasoning can be performed through a server side model (namely a fine-ranking model) to obtain a plurality of candidate videos, and the server side pushes the plurality of candidate videos to the terminal so that the terminal can obtain the candidate videos. After the candidate videos are obtained, the terminal can reorder the candidate videos based on the real-time feedback information in multiple aspects through the mixed expert model to obtain reordered results of the plurality of recommended videos, and therefore the recommended videos can be displayed based on the reordered results. In addition, the terminal can collect feedback information of the recommended video displayed based on the reordering result in real time, and the obtained real-time feedback information is transmitted back to the cloud end so as to assist the cloud end to carry out video fine-ranking and mixed expert model training.
According to the data recommendation method provided by the specification, the problem of large data expression difference of different application platforms can be solved through the mixed expert model, so that the situation of inconsistent multi-terminal data distribution can be compatible in the terminal rearrangement scene, and the sequencing accuracy is improved. In addition, the data recommendation method provided by the specification can greatly mine and process the available information on the terminal and the server, make different decisions by utilizing different information of the data in different scenes, maximize the effect of the characteristic information and improve the sequencing accuracy.
Corresponding to the embodiments of the method described above, the present description also provides corresponding device embodiments.
Referring to fig. 8, fig. 8 is a schematic block diagram of a terminal according to an exemplary embodiment. Referring to fig. 8, at a hardware level, a terminal may include a processor 802, an internal bus 804, a network interface 806, a memory 808, and a nonvolatile memory 810, although other tasks may be performed by the terminal. One or more embodiments of the present description may be implemented in a software-based manner, such as by the processor 802 reading a corresponding computer program from the non-volatile memory 810 into the memory 808 and then running. Of course, in addition to software implementation, one or more embodiments of the present disclosure do not exclude other implementation manners, such as a logic device or a combination of software and hardware, etc., that is, the execution subject of the following processing flow is not limited to each logic unit, but may also be hardware or a logic device.
The present disclosure further provides a data recommending apparatus, please refer to fig. 9, fig. 9 is a block diagram of a data recommending apparatus provided in an exemplary embodiment, and the data recommending apparatus may be applied to a terminal shown in fig. 8 to implement the technical solution of the present disclosure. The data recommending device may include:
the acquiring module 901 is configured to acquire a plurality of recommendation data, a sorting result of the plurality of recommendation data, and recommendation information of each recommendation data, where the recommendation information of the plurality of recommendation data and each recommendation data is determined by the server from a total of candidate recommendation data based on real-time feedback information of a user on the terminal;
the ranking module 902 is configured to reorder the plurality of recommended data based on the recommendation information, the real-time feedback information, and the application platform type corresponding to the terminal, so as to obtain a recommendation ranking result of the plurality of recommended data on the application platform corresponding to the terminal;
the display module 903 is configured to display a plurality of recommended data according to the recommendation ordering result.
In some embodiments, the ranking module 902 is configured to, when configured to reorder the plurality of recommended data based on the recommendation information of each recommended data, the real-time feedback information, and the application platform type corresponding to the terminal, obtain a recommendation ranking result of the plurality of recommended data on the application platform corresponding to the terminal, perform:
Based on the recommendation information and the real-time feedback information of each recommendation data, respectively processing the plurality of recommendation data through a plurality of expert networks included in the mixed expert model to obtain a plurality of sequencing features;
determining target sequencing features matched with the application platform types corresponding to the terminals from a plurality of sequencing features through a gating network included in the mixed expert model;
and processing the target ordering characteristics through a multi-layer perceptron included in the mixed expert model to obtain a recommended ordering result.
In some embodiments, the types of application platforms corresponding to the plurality of expert networks in the hybrid expert model are different, and for any expert network in the hybrid expert model, the expert network is used for determining ranking features of the plurality of recommendation data on the application platform corresponding to the expert network;
the ranking module 902 is configured to, when determining, from a plurality of ranking features through a gating network included in the hybrid expert model, a target ranking feature that matches an application platform type corresponding to the terminal, to:
based on the type of the application platform corresponding to the terminal, distributing weights to each expert network in the mixed expert model through a gating network included in the mixed expert model;
And based on the weights corresponding to the expert networks, carrying out weighted summation processing on the ordering features output by the expert networks to obtain the target ordering features.
In some embodiments, the recommendation data is multimedia data or a page delivery application;
the acquisition module 901, when configured to acquire a plurality of recommendation data, a ranking result of the plurality of recommendation data, and recommendation information of each recommendation data, is configured to:
when the recommended data is multimedia data, responding to multimedia data switching operation of a user, and acquiring a plurality of recommended data, a sequencing result of the plurality of recommended data and recommendation information of each recommended data;
and under the condition that the recommended data is a page delivery application, responding to the operation of returning the target page by the user, and acquiring a plurality of recommended data, a sequencing result of the plurality of recommended data and recommendation information of each recommended data.
In some embodiments, in the case where the recommended data is multimedia data, the real-time feedback information includes at least one of user real-time preference information, data browsing condition information, user preference change information, browsing condition change information, and terminal feature information;
The user real-time preference information is a data type of user preference predicted based on real-time feedback of the user; the data browsing condition information is the degree of interest predicted based on the browsing time of the user on the browsed recommended data; the user preference change information is difference information determined based on the sequencing result of the plurality of recommendation data and the user real-time preference information; the browsing condition change information comprises at least one item of difference information determined based on the current browsed recommended data and browsed recommended data of the user and difference information determined based on the current moment and the time length of browsing the recommended data of the user; the terminal characteristic information comprises at least one of the cached length of each recommended data on the terminal, the electric quantity information of the terminal and the memory information of the terminal;
in the case that the recommended data is multimedia data, the recommended information includes a recommended data duration, a recommended data type, a recommended data score, and a statistical feature value determined based on the browsed condition of the recommended data.
In some embodiments, in the case where the recommendation data is a page delivery application, the real-time feedback information includes real-time browsing condition information and user state prediction information for the user;
The user real-time browsing condition information comprises at least one of an application sequence accessed by the user in a target page and an application sequence accessed by the user in a set page area of the target page; the user state prediction information is a user state determined based on the sensing data acquired by the terminal sensor;
and under the condition that the recommendation data is a page delivery application, the recommendation information comprises a recommendation data identifier, a recommendation data type and a statistical characteristic value determined based on the clicked condition of the recommendation data.
In some embodiments, the obtaining module 901, when configured to obtain a plurality of recommended data, a ranking result of the plurality of recommended data, and recommendation information of each recommended data, is configured to:
generating a data recommendation request based on the real-time feedback information;
the method comprises the steps that a data recommendation request is sent to a server, and the server is used for determining a plurality of recommendation data from the total candidate recommendation data based on the data recommendation request and determining recommendation information of the plurality of recommendation data so as to return the plurality of recommendation data and recommendation information of each recommendation data to a terminal;
and receiving the plurality of recommended data returned by the server, the sequencing result of the plurality of recommended data and the recommendation information of each recommended data.
For the device embodiments, reference is made to the description of the method embodiments for the relevant points, since they essentially correspond to the method embodiments. The apparatus embodiments described above are merely illustrative, in which the modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical, i.e., may be located in one place, or may be distributed over a plurality of network modules. Some or all of the modules may be selected according to actual needs to achieve the purposes of the present description. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
The system, apparatus, module or unit set forth in the above embodiments may be implemented in particular by a computer chip or entity, or by a product having a certain function. A typical implementation device is a computer, which may be in the form of a personal computer, laptop computer, cellular telephone, camera phone, smart phone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or a combination of any of these devices.
In a typical configuration, a computer includes one or more processors (Central Processing Unit, CPU), input/output interfaces, network interfaces, and memory.
The Memory may include non-volatile Memory in a computer readable medium, random access Memory (Random Access Memory, RAM) and/or non-volatile Memory, etc., such as Read-Only Memory (ROM) or flash RAM. Memory is an example of computer-readable media.
Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change Memory (Phase Change Random Access Memory, PRAM), static random access Memory (Static Random Access Memory, SRAM), dynamic random access Memory (Dynamic Random Access Memory, DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically erasable programmable read Only Memory (Electrically Erasable Programmable Read-Only Memory, EEPROM), flash Memory or other Memory technology, read Only optical disk read Only Memory (Compact Disc Read-Only Memory, CD-ROM), digital versatile disks (Digital Video Disc, DVD) or other optical storage, magnetic cassettes, magnetic disk storage, quantum Memory, graphene-based storage media or other magnetic storage devices, or any other non-transmission media, that can be used to store information that can be accessed by a computing device. Computer-readable Media, as defined herein, does not include Transitory computer-readable Media (transmission Media), such as modulated data signals and carrier waves.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises the element.
The foregoing describes specific embodiments of the present disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
The terminology used in the one or more embodiments of the specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of the specification. As used in this specification, one or more embodiments and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any or all possible combinations of one or more of the associated listed items.
It should be understood that although the terms first, second, third, etc. may be used in one or more embodiments of the present description to describe various information, these information should not be limited to these terms. These terms are only used to distinguish one type of information from another. For example, first information may also be referred to as second information, and similarly, second information may also be referred to as first information, without departing from the scope of one or more embodiments of the present description. The word "if" as used herein may be interpreted as "at … …" or "at … …" or "responsive to a determination", depending on the context.
The foregoing description of the preferred embodiment(s) is (are) merely intended to illustrate the embodiment(s) of the present invention, and it is not intended to limit the embodiment(s) of the present invention to the particular embodiment(s) described.

Claims (9)

1. The data recommendation method is applied to a terminal and comprises the following steps:
acquiring a plurality of recommendation data, a sequencing result of the plurality of recommendation data and recommendation information of each recommendation data, wherein the recommendation information of the plurality of recommendation data and each recommendation data is determined and obtained from the total candidate recommendation data by a server based on real-time feedback information of a user on the terminal;
based on the recommendation information of each recommendation data, the real-time feedback information and the application platform type corresponding to the terminal, reordering the plurality of recommendation data to obtain recommendation ordering results of the plurality of recommendation data on the application platform corresponding to the terminal;
displaying the plurality of recommended data according to the recommended sorting result;
the recommendation data are multimedia data or page delivery application; when the recommended data is multimedia data, the real-time feedback information comprises at least one of user real-time preference information, data browsing condition information, user preference change information, browsing condition change information and terminal characteristic information; when the recommended data is a page delivery application, the real-time feedback information comprises user real-time browsing condition information and user state prediction information;
The acquiring the plurality of recommendation data, the sorting result of the plurality of recommendation data and the recommendation information of each recommendation data comprises any one of the following steps:
when the recommended data are multimedia data, responding to multimedia data switching operation of a user, and acquiring a plurality of recommended data, a sequencing result of the plurality of recommended data and recommendation information of each recommended data;
and under the condition that the recommended data is a page delivery application, responding to the operation of returning to the target page by a user, and acquiring a plurality of recommended data, a sequencing result of the plurality of recommended data and recommendation information of each recommended data.
2. The method of claim 1, wherein the reordering the plurality of recommended data based on the recommended information of each recommended data, the real-time feedback information, and the application platform type corresponding to the terminal to obtain a recommendation ordering result of the plurality of recommended data on the application platform corresponding to the terminal, comprises:
based on the recommendation information of each recommendation data and the real-time feedback information, respectively processing the plurality of recommendation data through a plurality of expert networks included in the mixed expert model to obtain a plurality of sequencing features;
Determining target sequencing features matched with the application platform types corresponding to the terminal from the plurality of sequencing features through a gating network included in the hybrid expert model;
and processing the target ranking features through a multi-layer perceptron included in the mixed expert model to obtain the recommended ranking result.
3. The method of claim 2, wherein application platforms corresponding to a plurality of expert networks in the hybrid expert model are different in type, and for any expert network in the hybrid expert model, the expert network is configured to determine ranking features of the plurality of recommendation data on the application platform corresponding to the expert network;
the determining, by the gating network included in the hybrid expert model, a target ranking feature that matches the application platform type corresponding to the terminal from the plurality of ranking features includes:
based on the type of the application platform corresponding to the terminal, distributing weights to each expert network in the mixed expert model through a gating network included in the mixed expert model;
and based on the weights corresponding to the expert networks, carrying out weighted summation processing on the ranking features output by the expert networks to obtain the target ranking features.
4. The method of claim 1, the user real-time preference information being a data type of user preference predicted based on real-time feedback of a user; the data browsing condition information is the degree of interest predicted based on the browsing time length of the browsed recommended data by the user; the user preference change information is difference information determined based on the sorting result of the plurality of recommendation data and the user real-time preference information; the browsing condition change information comprises at least one item of difference information determined based on the current browsed recommended data and browsed recommended data of the user and difference information determined based on the current moment and the time length of browsing the recommended data of the user; the terminal characteristic information comprises at least one of the cached length of each recommended data on the terminal, the electric quantity information of the terminal and the memory information of the terminal;
and when the recommended data are multimedia data, the recommended information comprises recommended data duration, recommended data type, recommended data score and statistical characteristic values determined based on browsed conditions of the recommended data.
5. The method of claim 1, wherein the real-time browsing condition information of the user comprises at least one of an application sequence accessed by the user in the target page and an application sequence accessed by the user in a set page area of the target page; the user state prediction information is a user state determined based on the sensing data acquired by the terminal sensor;
and under the condition that the recommendation data is a page delivery application, the recommendation information comprises a recommendation data identification, a recommendation data type and a statistical characteristic value determined based on the clicked condition of the recommendation data.
6. The method of claim 1, the obtaining a plurality of recommendation data, a ranking result of the plurality of recommendation data, and recommendation information of each recommendation data, comprising:
generating a data recommendation request based on the real-time feedback information;
the data recommendation request is sent to a server, and the server is used for determining the plurality of recommendation data from the total candidate recommendation data based on the data recommendation request and determining recommendation information of the plurality of recommendation data so as to return the plurality of recommendation data and recommendation information of each recommendation data to the terminal;
And receiving the plurality of recommended data returned by the server, the sequencing result of the plurality of recommended data and the recommendation information of each recommended data.
7. A data recommendation device is applied to a terminal and comprises:
the acquisition module is used for acquiring a plurality of recommendation data, a sequencing result of the plurality of recommendation data and recommendation information of each recommendation data, wherein the recommendation information of the plurality of recommendation data and each recommendation data is determined and obtained from the total candidate recommendation data by the server based on real-time feedback information of a user on the terminal;
the sequencing module is used for sequencing the plurality of recommended data based on the recommended information of each recommended data, the real-time feedback information and the application platform type corresponding to the terminal so as to obtain a recommended sequencing result of the plurality of recommended data on the application platform corresponding to the terminal;
the display module is used for displaying the plurality of recommended data according to the recommended sorting result;
the recommendation data are multimedia data or page delivery application; when the recommended data is multimedia data, the real-time feedback information comprises at least one of user real-time preference information, data browsing condition information, user preference change information, browsing condition change information and terminal characteristic information; when the recommended data is a page delivery application, the real-time feedback information comprises user real-time browsing condition information and user state prediction information;
The acquisition module is specifically configured to: when the recommended data are multimedia data, responding to multimedia data switching operation of a user, and acquiring a plurality of recommended data, a sequencing result of the plurality of recommended data and recommendation information of each recommended data;
and under the condition that the recommended data is a page delivery application, responding to the operation of returning to the target page by a user, and acquiring a plurality of recommended data, a sequencing result of the plurality of recommended data and recommendation information of each recommended data.
8. A terminal, comprising:
a processor;
a memory for storing processor-executable instructions;
wherein the processor is configured to implement the method of any of claims 1-6 by executing the executable instructions.
9. A computer readable storage medium having stored thereon computer instructions which, when executed by a processor, implement the steps of the method of any of claims 1-6.
CN202310889981.9A 2023-07-19 2023-07-19 Data recommendation method, device, equipment and medium Active CN116628235B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310889981.9A CN116628235B (en) 2023-07-19 2023-07-19 Data recommendation method, device, equipment and medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310889981.9A CN116628235B (en) 2023-07-19 2023-07-19 Data recommendation method, device, equipment and medium

Publications (2)

Publication Number Publication Date
CN116628235A CN116628235A (en) 2023-08-22
CN116628235B true CN116628235B (en) 2023-11-03

Family

ID=87602872

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310889981.9A Active CN116628235B (en) 2023-07-19 2023-07-19 Data recommendation method, device, equipment and medium

Country Status (1)

Country Link
CN (1) CN116628235B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116881575B (en) * 2023-09-08 2023-11-21 腾讯科技（深圳）有限公司 Content pushing method, device, computer equipment and storage medium

Citations (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106326391A (en) * 2016-08-17 2017-01-11 合智能科技（深圳）有限公司 Method and device for recommending multimedia resources
WO2018090793A1 (en) * 2016-11-18 2018-05-24 腾讯科技（深圳）有限公司 Multimedia recommendation method and device
CN109033299A (en) * 2018-07-16 2018-12-18 深圳市谷熊网络科技有限公司 It is a kind of by terminal applies to the method, device and equipment of user's recommendation information
CN113641894A (en) * 2021-07-20 2021-11-12 北京三快在线科技有限公司 Information recommendation method and device
CN113946745A (en) * 2021-09-01 2022-01-18 同济大学 General personalized recommendation method, recommendation system and training method
CN114186121A (en) * 2021-02-02 2022-03-15 寿带鸟信息科技（苏州）有限公司 Mixed recommendation algorithm system based on service record
CN114741583A (en) * 2021-01-07 2022-07-12 腾讯科技（深圳）有限公司 Information recommendation method and device based on artificial intelligence and electronic equipment
CN115062215A (en) * 2022-05-25 2022-09-16 中国联合网络通信集团有限公司 Multimedia content recommendation method, device and storage medium
CN115080835A (en) * 2021-03-10 2022-09-20 阿里巴巴新加坡控股有限公司 Information recommendation method and device, user side and equipment
CN115145455A (en) * 2022-06-07 2022-10-04 北京达佳互联信息技术有限公司 Data interaction method and device, electronic equipment and storage medium
CN115878839A (en) * 2021-09-28 2023-03-31 腾讯科技（深圳）有限公司 Video recommendation method and device, computer equipment and computer program product
CN116186413A (en) * 2023-03-10 2023-05-30 东北林业大学 Method and system for multi-modal feature recommendation
CN116233554A (en) * 2023-01-17 2023-06-06 北京达佳互联信息技术有限公司 Video playing method, device, electronic equipment and storage medium
CN116340374A (en) * 2023-02-20 2023-06-27 北京钰兔科技有限公司 Personalized task recommendation method and system

Patent Citations (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106326391A (en) * 2016-08-17 2017-01-11 合智能科技（深圳）有限公司 Method and device for recommending multimedia resources
WO2018090793A1 (en) * 2016-11-18 2018-05-24 腾讯科技（深圳）有限公司 Multimedia recommendation method and device
CN109033299A (en) * 2018-07-16 2018-12-18 深圳市谷熊网络科技有限公司 It is a kind of by terminal applies to the method, device and equipment of user's recommendation information
CN114741583A (en) * 2021-01-07 2022-07-12 腾讯科技（深圳）有限公司 Information recommendation method and device based on artificial intelligence and electronic equipment
CN114186121A (en) * 2021-02-02 2022-03-15 寿带鸟信息科技（苏州）有限公司 Mixed recommendation algorithm system based on service record
CN115080835A (en) * 2021-03-10 2022-09-20 阿里巴巴新加坡控股有限公司 Information recommendation method and device, user side and equipment
CN113641894A (en) * 2021-07-20 2021-11-12 北京三快在线科技有限公司 Information recommendation method and device
CN113946745A (en) * 2021-09-01 2022-01-18 同济大学 General personalized recommendation method, recommendation system and training method
CN115878839A (en) * 2021-09-28 2023-03-31 腾讯科技（深圳）有限公司 Video recommendation method and device, computer equipment and computer program product
CN115062215A (en) * 2022-05-25 2022-09-16 中国联合网络通信集团有限公司 Multimedia content recommendation method, device and storage medium
CN115145455A (en) * 2022-06-07 2022-10-04 北京达佳互联信息技术有限公司 Data interaction method and device, electronic equipment and storage medium
CN116233554A (en) * 2023-01-17 2023-06-06 北京达佳互联信息技术有限公司 Video playing method, device, electronic equipment and storage medium
CN116340374A (en) * 2023-02-20 2023-06-27 北京钰兔科技有限公司 Personalized task recommendation method and system
CN116186413A (en) * 2023-03-10 2023-05-30 东北林业大学 Method and system for multi-modal feature recommendation

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
Parham Hosseinzadeh Namin ; Crystal Roma ; Roberto Muscedere ; Majid Ahmadi.Efficient VLSI Implementation of a Sequential Finite Field Multiplier Using Reordered Normal Basis in Domino Logic. IEEE Transactions on Very Large Scale Integration (VLSI) Systems.2018,第2542 - 2552页. *
基于向量化标签的视频推荐算法研究与实现;许良武;;无线互联科技(12);第138-140+171页 *
基于点击反馈模型的内容推荐算法研究;石慧霞;机床与液压;第129-135页 *

Also Published As

Publication number Publication date
CN116628235A (en) 2023-08-22

Similar Documents

Publication Publication Date Title
CN108446410B (en) Information recommendation method, device, system, equipment and readable storage medium
CN107451199B (en) Question recommendation method, device and equipment
US10891679B2 (en) Generating media recommendations based upon beats per minute
CN110110203A (en) Resource information method for pushing and server, resource information methods of exhibiting and terminal
WO2023051102A1 (en) Video recommendation method, apparatus, and device, and medium
US20120096088A1 (en) System and method for determining social compatibility
US20180365709A1 (en) Personalized creator recommendations
CN111723237B (en) Media content access control method
CN109558544B (en) Sorting method and device, server and storage medium
CN116628235B (en) Data recommendation method, device, equipment and medium
US8725795B1 (en) Content segment optimization techniques
CN108846030B (en) method, system, electronic device and storage medium for visiting official website
CN114168843A (en) Search word recommendation method, device and storage medium
CN109446415A (en) A kind of application recommendation, acquisition methods and equipment
CN111988642B (en) Method, device, server and storage medium for recommending videos
CN115103232B (en) Video playing method, device, equipment and storage medium
CN112785391B (en) Recommendation processing method and device, intelligent equipment and storage medium
CN112035740A (en) Project use duration prediction method, device, equipment and storage medium
CN114861033A (en) Live broadcast recommendation method, device, equipment and computer readable storage medium
CN112256719A (en) Entity query method, device, readable medium and electronic equipment
CN113052664B (en) Service information recall method and device, electronic equipment and storage medium
CN113987360B (en) Object recommendation method and device, electronic equipment and storage medium
CN114707075B (en) Cold start recommendation method and device
CN112307313B (en) Information recommendation method, device and storage medium
CN112740203B (en) Data processing method, device, server and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant