Recommendation method and device, equipment and medium

Abstract

The invention provides a recommendation method and device, equipment and a medium, and relates to the technical field of artificial intelligence, in particular to the technical field of intelligent recommendation and software application. According to the implementation scheme, first behavior data and second behavior data of a target user are obtained, and the first behavior data comprise multiple user search texts historically searched by the target user and user behavior information corresponding to each user search text; the second behavior data comprises a plurality of recommended search texts historically interacted by the target user and user behavior information corresponding to each recommended search text; based on the first behavior data and the second behavior data, determining browsing behavior data of the target user browsing search result pages corresponding to the plurality of first search texts and content interaction behavior data of the target user aiming at search result pages corresponding to the plurality of second search texts; determining a target search text based on the browsing behavior data and the content interaction behavior data; and recommending the target search text to the target user.

Description

Technical Field

[0001] The present disclosure provides a recommendation method, apparatus, device, and medium, relating to the field of artificial intelligence technology, particularly to the fields of intelligent recommendation and software application technology, and specifically relates to a recommendation method, apparatus, electronic device, computer-readable storage medium, and computer program product. Background Art

[0002] Artificial intelligence is a discipline that studies how to make a computer simulate certain thinking processes and intelligent behaviors of humans (such as learning, reasoning, thinking, planning, etc.), including both hardware-level technologies and software-level technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, and big data processing; artificial intelligence software technologies mainly include several major directions such as computer vision technology, speech recognition technology, natural language processing technology, and machine learning / deep learning, big data processing technology, and knowledge graph technology.

[0003] With the development of computer technology and big data technology, behaviors such as consumption, entertainment, learning, and travel in people's lives are closely related to big data. During the operation of a software platform, it is usually necessary to actively recommend content to users. In current search software, relevant content is usually recommended to users based on the search text that the users have actively input.

[0004] The methods described in this section are not necessarily methods that have been previously conceived or adopted. Unless otherwise specified, any method described in this section should not be considered prior art solely because it is included in this section. Similarly, unless otherwise specified, the problems mentioned in this section should not be considered to have been recognized in any prior art. Summary of the Invention

[0005] The present disclosure provides a recommendation method, apparatus, electronic device, computer-readable storage medium, and computer program product.

[0006] According to one aspect of the present disclosure, there is provided a recommendation method, including: obtaining first behavior data and second behavior data of a target user, wherein the first behavior data includes a plurality of user search texts searched by the target user in history and user behavior information corresponding to each user search text, and the second behavior data includes a plurality of recommended search texts interacted by the target user in history and user behavior information corresponding to each recommended search text; determining browsing behavior data and content interaction behavior data based on the first behavior data and the second behavior data, wherein the browsing behavior data includes a plurality of first search texts and browsing behavior information of the target user browsing a search result page corresponding to each first search text, and the content interaction behavior data includes a plurality of second search texts and behavior information of the target user performing an interaction operation on a search result page corresponding to each second search text; determining a target search text based on the browsing behavior data and the content interaction behavior data; and recommending the target search text to the target user.

[0007] According to one aspect of the present disclosure, there is provided a recommendation model, including: a first interoperability network for outputting a first interoperability result based on browsing behavior data of a target user and a candidate search text set input to the first interoperability network, wherein the browsing behavior data includes a plurality of first search texts and browsing behavior information of the target user browsing a search result page corresponding to each first search text; a second interoperability network for outputting a second interoperability result based on content interaction behavior data of the target user and the candidate search text set input to the second interoperability network, wherein the content interaction behavior data includes a plurality of second search texts and behavior information of the target user performing an interaction operation on a search result page corresponding to each second search text; a prediction network for determining first prediction information capable of indicating the probability that the target user browses a search result page corresponding to each candidate search text in the candidate search text set and second prediction information capable of indicating the probability that the target user performs an interaction operation on a search result page corresponding to each candidate search text in the candidate search text set based on the first interoperability result, the second interoperability result, the browsing behavior data, the content interaction behavior data, and the candidate search text set input to the prediction network; and an output layer for determining a target search text for recommending to the target user from the candidate search text set based on the first prediction information and the second prediction information.

[0008] According to one aspect of the present disclosure, a recommendation device is provided, including: a first acquisition unit configured to acquire first behavior data and second behavior data of a target user, wherein the first behavior data includes a plurality of user search texts historically searched by the target user and user behavior information corresponding to each user search text, and the second behavior data includes a plurality of recommended search texts historically interacted with by the target user and user behavior information corresponding to each recommended search text; a first determination unit configured to determine browsing behavior data and content interaction behavior data based on the first behavior data and the second behavior data, wherein the browsing behavior data includes a plurality of first search texts and browsing behavior information of the target user browsing the search result pages corresponding to each first search text, and the content interaction behavior data includes a plurality of second search texts and behavior information of the target user performing interaction operations on the search result pages corresponding to each second search text; a second determination unit configured to determine a target search text based on the browsing behavior data and the content interaction behavior data; and a recommendation unit configured to recommend the target search text to the target user.

[0009] According to one aspect of the present disclosure, an electronic device is provided, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is enabled to execute the above-mentioned recommendation method.

[0010] According to one aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to execute the above-mentioned recommendation method.

[0011] According to one aspect of the present disclosure, a computer program product is provided, including a computer program, wherein when the computer program is executed by a processor, the above-mentioned recommendation method can be implemented.

[0012] According to one or more embodiments of the present disclosure, the first behavior data in the active search scenario and the second behavior data in the recommendation interaction scenario of the user can be fused for recommendation to improve the recommendation accuracy.

[0013] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it used to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. Description of the Drawings

[0014] The accompanying drawings exemplarily illustrate embodiments and constitute a part of the specification, and are used together with the written description of the specification to explain the exemplary embodiments of the embodiments. The illustrated embodiments are for illustrative purposes only and do not limit the scope of the claims. In all the drawings, the same reference numerals refer to similar but not necessarily identical elements.

[0015] Figure 1 FIG. shows a schematic diagram of an exemplary system in which various methods described herein can be implemented according to an exemplary embodiment of the present disclosure;

[0016] Figure 2 FIG. shows a flowchart of a recommendation method according to an exemplary embodiment of the present disclosure;

[0017] Figure 3 FIG. shows a schematic diagram of a recommendation process according to an exemplary embodiment of the present disclosure;

[0018] Figure 4 FIG. shows a schematic diagram of a recommendation model according to an exemplary embodiment of the present disclosure;

[0019] Figure 5 FIG. shows a schematic diagram of a recall network according to an exemplary embodiment of the present disclosure;

[0020] Figure 6 FIG. shows a schematic diagram of a ranking network according to an exemplary embodiment of the present disclosure;

[0021] Figure 7 FIG. shows a structural block diagram of a recommendation device according to an exemplary embodiment of the present disclosure;

[0022] Figure 8 FIG. shows a structural block diagram of an exemplary electronic device that can be used to implement the embodiments of the present disclosure. Detailed Embodiments

[0023] The following describes exemplary embodiments of the present disclosure with reference to the accompanying drawings. Various details of the embodiments of the present disclosure are included to facilitate understanding, and they should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for clarity and conciseness, descriptions of well-known functions and structures are omitted below.

[0024] In the present disclosure, unless otherwise specified, the terms "first", "second", etc. are used to describe various elements and are not intended to limit the positional relationship, timing relationship, or importance relationship of these elements. Such terms are only used to distinguish one element from another. In some examples, the first element and the second element may refer to the same instance of the element, and in certain cases, based on the context description, they may also refer to different instances.

[0025] In the description of the various examples in this disclosure, the terms used are for the purpose of describing specific examples only and are not intended to be limiting. Unless the context clearly indicates otherwise, if the number of elements is not specifically defined, the element can be one or more. In addition, the term "and / or" used in this disclosure covers any one of the listed items and all possible combinations thereof.

[0026] Embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0027] Figure 1 FIG. shows a schematic diagram of an exemplary system 100 in which the various methods and apparatuses described herein can be implemented according to an embodiment of the present disclosure. Referring Figure 1 , the system 100 includes one or more client devices 101, 102, 103, 104, 105 and 106, a server 120, and one or more communication networks 110 that couple the one or more client devices to the server 120. The client devices 101, 102, 103, 104, 105 and 106 can be configured to execute one or more applications.

[0028] In an embodiment of the present disclosure, the server 120 can run one or more services or software applications that enable the execution of a recommendation method.

[0029] In certain embodiments, the server 120 can also provide other services or software applications, which can include non-virtual environments and virtual environments. In certain embodiments, these services can be provided as web-based services or cloud services, for example, provided to users of the client devices 101, 102, 103, 104, 105 and / or 106 under a software as a service (SaaS) model.

[0030] In Figure 1 the configuration shown, the server 120 can include one or more components that implement the functions performed by the server 120. These components can include software components, hardware components, or a combination thereof that can be executed by one or more processors. Users operating the client devices 101, 102, 103, 104, 105 and / or 106 can in turn utilize one or more client applications to interact with the server 120 to utilize the services provided by these components. It should be understood that various different system configurations are possible, which can be different from the system 100. Therefore, Figure 1 is an example of a system for implementing the various methods described herein and is not intended to be limiting.

[0031] Users can use client devices 101, 102, 103, 104, 105, and / or 106 to send recommendation requests or perform interactive operations on recommended content. The client devices can provide an interface that enables the users of the client devices to interact with the client devices. The client devices can also output information to the users via this interface. Although Figure 1 only six client devices are depicted, those skilled in the art will be able to understand that the present disclosure can support any number of client devices.

[0032] Client devices 101, 102, 103, 104, 105, and / or 106 can include various categories of computer devices, such as portable handheld devices, general-purpose computers (such as personal computers and laptop computers), workstation computers, wearable devices, smart screen devices, self-service terminal devices, service robots, gaming systems, thin clients, various messaging devices, sensors, or other sensing devices, etc. These computer devices can run various categories and versions of software applications and operating systems, such as MICROSOFT Windows, APPLE iOS, UNIX-like operating systems, Linux, or Linux-like operating systems (such as GOOGLE Chrome OS); or include various mobile operating systems, such as MICROSOFT WindowsMobile OS, iOS, Windows Phone, Android. Portable handheld devices can include cellular phones, smartphones, tablets, personal digital assistants (PDAs), etc. Wearable devices can include head-mounted displays (such as smart glasses) and other devices. Gaming systems can include various handheld gaming devices, Internet-enabled gaming devices, etc. The client devices are capable of executing various different applications, such as various Internet-related applications, communication applications (such as email applications), short message service (SMS) applications, and can use various communication protocols.

[0033] Network 110 can be any category of network known to those skilled in the art, which can support data communication using any one of a variety of available protocols (including but not limited to TCP / IP, SNA, IPX, etc.). By way of example only, one or more networks 110 can be a local area network (LAN), an Ethernet-based network, token ring, wide area network (WAN), the Internet, virtual network, virtual private network (VPN), intranet, extranet, blockchain network, public switched telephone network (PSTN), infrared network, wireless network (such as Bluetooth, WIFI), and / or any combination of these and / or other networks.

[0034] Server 120 may include one or more general-purpose computers, dedicated server computers (such as PC (Personal Computer) servers, UNIX servers, midrange servers), blade servers, mainframes, server clusters, or any other suitable arrangement and / or combination. Server 120 may include one or more virtual machines running a virtual operating system, or other computing architectures involving virtualization (such as one or more flexible pools of logical storage devices that can be virtualized to maintain virtual storage devices for the server). In various embodiments, server 120 may run one or more services or software applications that provide the functions described below.

[0035] The computing units in server 120 may run one or more operating systems including any of the above operating systems as well as any commercially available server operating systems. Server 120 may also run any one of a variety of additional server applications and / or middleware applications, including HTTP servers, FTP servers, CGI servers, JAVA servers, database servers, etc.

[0036] In some embodiments, server 120 may include one or more applications to analyze and merge data feeds and / or event updates received from users of client devices 101, 102, 103, 104, 105, and 106. Server 120 may also include one or more applications to display data feeds and / or real-time events via one or more display devices of client devices 101, 102, 103, 104, 105, and 106.

[0037] In some embodiments, server 120 may be a server of a distributed system, or a server incorporating a blockchain. Server 120 may also be a cloud server, or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology. A cloud server is a host product in the cloud computing service system, which addresses the defects of high management difficulty and weak business scalability existing in traditional physical hosts and virtual private server (VPS, Virtual Private Server) services.

[0038] System 100 may also include one or more databases 130. In some embodiments, these databases may be used to store data and other information. For example, one or more of the databases 130 may be used to store information such as audio files and video files. The databases 130 may reside in various locations. For example, the databases used by the server 120 may be local to the server 120, or may be remote from the server 120 and may communicate with the server 120 via a network-based or dedicated connection. The databases 130 may be of different categories. In some embodiments, the databases used by the server 120 may be, for example, relational databases. One or more of these databases may store, update, and retrieve data to and from the databases in response to commands.

[0039] In some embodiments, one or more of the databases 130 may also be used by an application to store application data. The databases used by the application may be different categories of databases, such as key-value repositories, object repositories, or conventional repositories supported by a file system.

[0040] Figure 1 The system 100 can be configured and operated in various ways to enable the application of the various methods and apparatuses described in this disclosure.

[0041] In the related art, in current search software, different page display components are usually used to separately display search results and recommended content to users, and the search system and the recommendation system are decoupled. In this case, when it is necessary to recommend search text to a user, the active search behavior information of the user in the search page is usually directly input into the recommendation model so that the recommendation model can output recommended content related to the user's active search content. This method fails to comprehensively and accurately capture the user's interest in the recommended content, and the recommendation accuracy is limited.

[0042] Based on this, the present disclosure provides a recommendation method, which obtains the first behavior data of the user in the active search scenario and the second behavior data of the user in the recommendation scenario, and then groups and extracts the behavior characteristics of different behavior categories to obtain the browsing behavior data representing the user's search and browsing content and the content interaction behavior data representing the user's interaction operations on the content in the search result page, that is, the integration of the user's behavior information in different scenarios is realized, and the characteristic data of different behavior categories are used to more accurately indicate the user's interest, and more accurate recommendations are realized based on this.

[0043] Figure 2 shows a flowchart of a recommendation method according to an exemplary embodiment of the present disclosure. As Figure 2 shown, the method 200 includes:

[0044] Step S201: Obtain the first behavior data and the second behavior data of the target user. Among them, the first behavior data includes multiple user search texts searched by the target user historically and the corresponding user behavior information for each user search text, and the second behavior data includes multiple recommended search texts interacted with by the target user historically and the corresponding user behavior information for each recommended search text;

[0045] Step S202: Based on the first behavior data and the second behavior data, determine the browsing behavior data and the content interaction behavior data. Among them, the browsing behavior data includes multiple first search texts and the browsing behavior information of the target user browsing the corresponding search result pages for each first search text, and the content interaction behavior data includes multiple second search texts and the behavior information of the target user performing interaction operations on the corresponding search result pages for each second search text;

[0046] Step S203: Based on the browsing behavior data and the content interaction behavior data, determine the target search text; and

[0047] Step S204: Recommend the target search text to the target user.

[0048] By applying the above-mentioned recommendation method 200, it is possible to obtain the first behavior data of the user in the active search scenario and the second behavior data of the user in the recommendation scenario, and then group and extract the behavior characteristics of different behavior categories to obtain the browsing behavior data representing the user's search and browsing content and the content interaction behavior data representing the user's interaction operations on the content in the search result pages, that is, the integration of the user behavior information in different scenarios is realized, and the feature data of different behavior categories are used to more accurately indicate the user's interests, and more accurate recommendations are realized based on this.

[0049] In some examples, the interaction operations of the user on the search result page indicated by the content interaction behavior data may include one or more of various interaction operations such as clicking, selecting content, sharing content, copying content, etc. For example, it may be behaviors such as clicking on the link of the search result page, selecting the text or picture of the search result page, copying the text or picture of the search result page, and sharing the search result page to other pages or sessions or software platforms. Based on the user's content interaction behavior information, the degree of the user's interest in the search text can be more accurately indicated, and more accurate recommendations can be realized based on this.

[0050] In some examples, the first line of data in step S201 can be the behavior sequence of the user actively entering the user search text and conducting a search on the search page, which can specifically include information such as the user search text, the search time corresponding to the user search text, and the browsing duration of the user on the search result page. In some examples, the behavior sequence can be structured. For example, it can include the user search text entered by the user in the search box and the search moment or interaction duration corresponding to the user search text (for example, it can be the browsing duration of the search result page).

[0051] In some examples, the second line of data in step S201 can be the behavior sequence of the user performing operations such as browsing and clicking on the recommended search text within a historical time period. The behavior sequence included in the second line of data can be structured. For example, it can include a sequence of behavior category information, interaction content information, and time information. Behavior categories can include behaviors such as triggering the search process based on the recommended search text, browsing the search result page of the recommended search text, clicking or sharing the search result items in the search result page, etc. Interaction content information can be, for example, the display content targeted by each interaction behavior (such as web pages, documents, pictures, etc. in the search result page, or the user search text entered by the user in the search box). Time information can include, for example, the moment when the user performs the interaction behavior or the interaction duration (such as the browsing duration of the search result page).

[0052] As mentioned above, the first line of data contains the user search text and the user behavior information corresponding to the search category, and the second line of data contains the recommended search text and the user behavior information corresponding to categories such as triggering the search process and clicking. In this case, in step S202, the first line of data and the second line of data can be grouped based on the behavior category, and then the grouped results can be fused based on the information of the search text to obtain the browsing behavior data corresponding to the search browsing category and the content interaction behavior data corresponding to the click category. In the actual application scenario, different interaction behavior categories correspond to different degrees of user interest. For example, when the user performs a click operation on a search result item for a certain search text, it indicates that the user has a relatively high degree of interest in the search text. When the user only browses the search result page of a certain search text for a short time, it indicates that the user has a relatively low degree of interest in the search text. By grouping and extracting and fusing the behavior characteristics of different behavior categories in different scenarios, the integration of user behavior information in different scenarios is achieved, and the characteristic data of different behavior categories can be used to more accurately indicate the degree of user interest, based on which more accurate recommendations can be realized.

[0053] In one example, when the first row of data includes the search text A and the browsing duration in the active search scenario corresponding to the search text A, and the second row of data includes the search text A, the browsing duration in the recommended scenario corresponding to the search text A, and the click behavior information corresponding to the search text B, the browsing duration information corresponding to the browsing behavior and the click behavior information can be extracted respectively based on the behavior category. The information of the search text A and its corresponding browsing durations in different scenarios is added to the browsing behavior data, and the search text A and its corresponding click behavior information are added to the content interaction behavior data.

[0054] By querying the user behavior information of the specific behavior category corresponding to each search text (including the user search text and the recommended search text) based on the first row of data and the second row of data and fusing the query results, the user behavior information corresponding to the browsing behavior category (i.e., the browsing behavior data) and the user behavior information corresponding to the content interaction behavior category (such as click, select content, share content, copy content) of each search text can be determined. Based on the user behavior data of different categories, the interest of the user in the search text can be indicated more accurately.

[0055] According to some embodiments, in step S203, based on the browsing behavior data and the content interaction behavior data, determining the target search text includes: determining a candidate search text set; based on the browsing behavior data, determining first prediction information capable of indicating the probability that the target user browses the search result page corresponding to each candidate search text in the candidate search text set; based on the content interaction behavior data, determining second prediction information capable of indicating the probability that the target user performs an interaction operation in the search result page corresponding to each candidate search text in the candidate search text set; and based on the first prediction information and the second prediction information, determining the target search text from the candidate search text set. Thus, different user behaviors can be predicted respectively based on the feature data of different behavior categories, the prediction accuracy can be improved, and further the recommendation accuracy can be improved.

[0056] In some examples, the step of determining the target search text based on the browsing behavior data and the content interaction behavior data can also be implemented in other ways. For example, the relevance between the first search text and the second search text included in the browsing behavior data and the content interaction behavior data and the candidate search text can be calculated. Based on the content relevance between the candidate search text and the search text included in the behavior data and the user behavior information included in the behavior data, the recommendation weight of each candidate search text can be determined, and based on this, the target search text can be determined. In this case, the influence degree of the browsing behavior data on the recommendation weight of the candidate search text should be less than the influence degree of the content interaction behavior data on the recommendation weight of the candidate search text, that is, different types of user behavior data can be used to indicate different degrees of user interest, and the correlation degree between the content interaction behavior data and the user interest is stronger than that of the browsing behavior data. Based on this, more accurate recommendations can be achieved.

[0057] According to some embodiments, the first prediction information for determining the probability that the target user browses the search result page corresponding to each candidate search text in the candidate search text set based on the browsing behavior data includes: inputting the browsing behavior data and each candidate search text in the candidate search text set into a first interoperability network; and determining the first prediction information based on the output result of the first interoperability network. By using the interoperability network to implement the interoperability calculation between the user's browsing behavior data (including the search text and browsing behavior information browsed by the user in history) and the candidate search text, the relevance between the search text browsed by the user in history and the candidate search text can be indicated based on the interoperability result, and thus the user behavior can be predicted more accurately, improving the recommendation accuracy.

[0058] In some examples, the second prediction information can be determined in a similar manner, that is, the content interaction behavior data and each candidate search text in the candidate search text set are input into a second interoperability network, and then the second prediction information is determined based on the output result of the second interoperability network. The present disclosure will not elaborate on this.

[0059] According to some embodiments, the first interoperability network is an attention network. Thus, it is possible to more accurately capture the relevance between the first search text browsed by the user in history and the candidate search text based on the attention mechanism, improving the prediction accuracy.

[0060] In some examples, the first interoperability network can also be in other forms, that is, the interoperability calculation between the user behavior data and the candidate search text can be implemented based on other forms. For example, the relevance between the first search text and the second search text included in the browsing behavior data and the content interaction behavior data and the candidate search text can be calculated, and the calculation result of the relevance is used as the interoperability result to improve the efficiency of the interoperability calculation.

[0061] According to some embodiments, determining the first prediction information based on the output result of the first interoperability network includes: inputting the output result of the first interoperability network, the browsing behavior data, the content interaction behavior data, and the candidate search text set into a prediction network; and determining the first prediction information based on the output result of the prediction network. Thereby, on the basis of inputting the prediction information regarding user behavior into the recommendation prediction network, the user behavior data indicating user interests and the information of candidate content can be transmitted into the recommendation prediction network together, so that the prediction network can determine the prediction result based on more comprehensive information, thereby improving the prediction accuracy.

[0062] In some examples, the recommendation prediction network can be constructed based on various types of neural networks, and various model training methods can be used to tune and train the first interoperability network and the recommendation prediction network. For example, the first interoperability network and the recommendation prediction network can be supervised-trained using labeled data, or they can also be trained based on reinforcement learning strategies and other methods. As long as it can be realized to predict the user's behavior regarding the candidate search text based on the information of user behavior data and candidate search text, the present disclosure does not limit the specific structural types and training methods of the interoperability network or the prediction network.

[0063] As mentioned above, the second prediction information can be determined based on a similar method, that is, inputting each candidate search text in the content interaction behavior data and the candidate search text set into the second interoperability network. In this case, the output result of the first interoperability network, the output result of the second interoperability network, the browsing behavior data, the content interaction behavior data, and the candidate search text set can be input into the prediction network, so that the prediction network can output the first prediction information and the second prediction information based on the input information, thereby realizing end-to-end user behavior prediction using the recommendation model framework including the first interoperability network, the second interoperability network, and the prediction network.

[0064] In some examples, the prediction network may include multiple prediction towers, and each prediction tower is used to output a type of user behavior metric based on input information. The user behavior metrics may include, for example, the predicted probability that the target user views the search result page corresponding to the candidate search text (i.e., the probability of triggering the search process based on the candidate search text), the predicted duration that the target user views the search result page corresponding to the candidate search text, and the predicted probability that the target user performs an interaction operation (such as clicking on a search result item) on the search result page corresponding to the candidate search text. By constructing multiple prediction towers to predict multiple user behavior metrics, the prediction accuracy can be improved, which is convenient for targeted parameter tuning and optimization during the model training process. In some examples, the multiple prediction towers may include one or more shared network layers to more accurately fit the correlation between different user behaviors (for example, the user must first trigger the search process based on a certain candidate search text before performing an interaction operation on the search result page corresponding to the candidate search text), and based on this, the prediction accuracy is improved.

[0065] According to some embodiments, method 200 further includes: determining information about the display page for displaying the target search text to the target user, where the information about the display page includes the trigger behavior information of the target user entering the display page; and the input of the prediction network further includes the information about the display page. Thus, it is possible to indicate the user's interest based on the trigger behavior information for triggering the display of the recommended search text, so as to more accurately determine the recommended content.

[0066] In some examples, the trigger behavior of the target user may include: actively entering a search text in the search box to trigger the search result display page, clicking on a search entry in the popular search content list to quickly trigger the search result display page, and clicking on a search entry in the recommended card to quickly trigger the search result display page. In this case, the target search text can be displayed while showing the search results to the user, which is convenient for the user to further trigger the search based on the target search text, so as to show the search result page corresponding to the target search text and achieve accurate recommendation in the search scenario.

[0067] It should be understood that the trigger method corresponding to the display page can indicate the user's interest to a certain extent. For example, when the user triggers the search by clicking on the popular search content list, it indicates that the user may be more sensitive to the timeliness or popularity of the search content. Furthermore, the recommended search text that can better meet the user's needs can be determined based on this to improve the recommendation effect.

[0068] In some examples, the information of the display page may further include other content. For example, it may include display time information, and then content that better meets the user's needs can be determined based on the display time information. For example, when the display time is a public holiday, content related to holiday travel can be recommended to the user to improve the accuracy of recommendations.

[0069] According to some embodiments, the determining of the candidate search text set includes: recalling a plurality of candidate search texts in the candidate search text set from a search text library based on at least one of the first behavior data and the second behavior data. By using the original user behavior sequence in the recall step, a more comprehensive and rich candidate search text set can be determined, thereby improving the recommendation accuracy.

[0070] In some examples, other methods may also be used to determine the candidate search text set. For example, the candidate search text set can be preset manually, or multiple popular search texts during the current period can be used as the candidate search text set to fully meet the requirements in the actual recommendation scenario.

[0071] According to some embodiments, the recalling of a plurality of candidate search texts in the candidate search text set from the search text library based on at least one of the first behavior data and the second behavior data includes: determining the interest characteristics of the target user based on at least one of the first behavior data and the second behavior data; and recalling the plurality of candidate search texts from the search text library based on the interest characteristics. Thus, the interest characteristics can be extracted based on the behavior data to more accurately indicate the user's interests, reduce the number of parameters, and improve the recommendation efficiency.

[0072] In some examples, the user's interest characteristics can be extracted by inputting the user behavior data into an attention network to improve the efficiency and accuracy of interest characteristic extraction.

[0073] In some examples, other types of neural networks may also be applied to extract the user's interest characteristics, or other processing methods may be used to determine the user's interest characteristics. For example, the search texts with a higher degree of user interest can be obtained based on the statistical characteristics of the user behavior information and used as explicit user interest characteristics. The present disclosure does not limit the specific method for obtaining the user interest characteristics.

[0074] According to some embodiments, method 200 further includes: determining a plurality of user interest words based on at least one of the plurality of user search texts and the plurality of recommended search texts; wherein, the recalling of the plurality of candidate search texts from the search text library based on the interest feature includes: recalling the plurality of candidate search texts from the search text library based on the interest feature and the plurality of user interest words. Thereby, it is possible to determine user interest words and each set of user interest words based on the user's historical search or the searched text of historical interaction, so as to more comprehensively and accurately represent the user's interest and improve the recommendation accuracy.

[0075] According to some embodiments, the plurality of user interest words include at least one of the following: a plurality of historical search words obtained by performing a word segmentation operation on at least one of the plurality of user search texts and the plurality of recommended search texts; and content label words corresponding to at least one of the plurality of user search texts and the plurality of recommended search texts. By determining user interest words based on the corresponding word segmentation results of historical search texts or the content label words of historical search texts, the convenience and efficiency of determining user interest words can be improved, and further the recommendation efficiency can be improved.

[0076] In some examples, the content label words corresponding to the historical search text may be label words annotated by the search engine for each search text that can indicate the content field corresponding to the search text. For example, they may be labels such as "current affairs news", "sports", "education", etc. By applying the content label words as user interest words, the user's interest in the content field can be more accurately indicated to obtain recommended content that better meets the user's needs.

[0077] In some examples, the extraction of the plurality of user interest words can also be implemented by other means. For example, an operation of extracting keywords can be performed on each historical search text to obtain user interest words that can represent the key semantics of the historical search text. For another example, a language model can also be used to determine a plurality of user interest words based on a plurality of historical search texts. As long as it is possible to determine words that can indicate the user's interest and are more concise than the original search text information based on the historical search text, the present disclosure does not limit the specific extraction method of the user interest words.

[0078] According to some embodiments, method 200 further includes: determining an interest weight for each user interest word among the plurality of user interest words based on at least one of the first behavior data and the second behavior data, wherein the recalling the plurality of candidate search texts from the search text library based on the interest feature includes: recalling the plurality of candidate search texts from the search text library based on the interest feature, the plurality of user interest words, and the interest weight of each user interest word. Thereby, it is possible to determine the interest word weight based on the interaction behavior information in the behavior data to more accurately indicate the degree of user interest, and based on this, more accurate recommendations can be realized.

[0079] In some examples, it may be to determine the statistical result of the user behavior information corresponding to each user interest word (such as the total number of views, the sum of the viewing durations, the total number of clicks) for each user interest word, and based on this, determine the interest weight of the user interest word to more accurately indicate the degree of interest of the target user in each user interest word, and based on this, more accurate recommendations can be realized.

[0080] In some examples, the interest weight of each user interest word can also be determined based on the time interval between the user browsing time or the user interaction time corresponding to the user interest word and the current time. The longer the time interval between the corresponding user browsing time or user interaction time of each user interest word and the current time, the lower the interest weight of the user interest word, that is, it corresponds to the degree of attenuation of the user interest over time. Based on this, it is possible to determine more content that conforms to the user interest for recommendation to improve the recommendation effect.

[0081] According to some embodiments, method 200 further includes: determining at least one popular search text from the search text library based on the user interaction data of each search text in the search text library; and adding the at least one popular search text to the candidate search text set. By adding the popular content recalled based on the rules to the candidate set, the richness of the candidate search text set can be improved, the user's requirements for the popularity and timeliness of the content can be met, and the recommendation effect can be improved.

[0082] According to some embodiments, method 200 further includes: determining the historical number of times of recommending the target search text to the target user; and in response to the historical number of times exceeding the number threshold, re-determining the target search text based on the browsing behavior data and the content interaction behavior data. By checking the historical number of times of recommending the target search text, it is possible to avoid frequently recommending duplicate content to the user and improve the user experience.

[0083] According to some embodiments, method 200 further includes: in response to determining that the target search text is the same as any one of the multiple user search texts and the multiple recommended search texts, re-determining the target search text based on the browsing behavior data and the content interaction behavior data. By adjusting the recommended content according to the information actively searched by the user, it is possible to avoid recommending the content that has been interacted with by the user and improve the user experience.

[0084] Figure 3 FIG. shows a schematic diagram of a recommendation process according to an exemplary embodiment of the present disclosure. As Figure 3 shown, the recommendation process can be implemented based on the collaborative operation of the client 310 and the server 320. In this example, the user interaction module 312 in the client 310 is used to display the target search text to the user, and the data collection module 311 is used to collect the interaction behavior data of the user (for example, it may include the above-mentioned first behavior data and second behavior data) and transmit it to the server 320.

[0085] The feature processing module 321 can perform operations such as cleaning, behavior category classification, and fusion based on the user behavior data collected at the front end. For example, it can implement the operations of determining the browsing behavior data and the content interaction behavior data based on the first behavior data and the second behavior data described above. In the above example, the feature processing module 321 processes based on the personalized behavior data of the target user to obtain feature data that can indicate the personalized preferences of the target user. In some examples, the feature processing module 321 can also determine statistical features based on the user behavior data of multiple users to indicate the features of the candidate content. For example, it can count statistical information such as the user search frequency, user browsing volume, and user click volume corresponding to each search text to indicate the popularity of each search text. The feature processing module 321 can also determine features related to the recommendation scenario based on the user behavior data of multiple users, such as the information of the display page described above. For example, based on the trigger behavior information of multiple users entering the search result display page and the interaction behavior of multiple users in the search result display page, the correlation between the search trigger behavior and the user interaction behavior can be statistically analyzed, and thus the user interest indicated by the trigger method corresponding to the search result display page can be fully considered in the recommendation process to improve the recommendation effect.

[0086] In the recall model service 322, a recall model can be deployed to determine recall content based on the user behavior data input into the recall model and the recallable search texts, that is, to obtain a candidate search text set and pass it into the sorting model. The recallable search texts can be obtained from the search text library 325. In some examples, the search text library 325 can be pre-configured manually or configured based on the historical search data of multiple users. In one example, multiple popular search texts within the current period can also be added to the search text library 325 to meet the user's requirements for the popularity and timeliness of search texts.

[0087] In the sorting model service 323, a sorting model can be deployed to sort based on the user behavior data input into the sorting model and the candidate search text set, and then determine the target search text for recommendation to the user. After obtaining the target search text using the recall model and the sorting model, the push module 324 can check whether the target search text meets the push conditions based on the pre-configured push policy. For example, it can check whether the historical recommendation times of the target search text exceed the times threshold or whether the target search text is repeated with the search texts historically interacted with by the target user based on the steps described above, to determine whether to call the recall model or the sorting model to re-determine a new target search text. After obtaining the target search text that meets the push conditions, the push module 324 can send it to the user interaction module 312 to be displayed to the target user.

[0088] In some examples, rules for checking based on the content tag words of the search text can be further configured in the push module 324. As mentioned above, the content tag words corresponding to the search text can be tag words annotated by the search engine for each search text to indicate the content field corresponding to the search text. For example, they can be tag words such as "Current Affairs News", "Sports", "Education", etc. In one example, when the number of times of pushing search texts related to a certain content field to the target user has exceeded a certain threshold, a filtering operation based on the content tag words can be performed in the push module 324 to ensure that search texts related to this content field are no longer recommended to the target user, so as to avoid affecting the user experience due to a large number of repeated recommendations of similar content.

[0089] In some examples, the presentation and interaction of the target search text on the display page can be implemented based on the following methods:

[0090] When receiving the initial search text sent by the user, the search result entries corresponding to the initial search text and the target search text can be displayed on the first page (i.e., recommend the target search text to the user on the search result page triggered by the user); in response to receiving the first interaction operation of the target user, and the content volume of the search result entries corresponding to the initial search text that have been displayed is less than the content volume threshold, the second page is continuously displayed on the basis of the first page, and the second page may include the search result entries corresponding to the initial search text that have not been displayed on the first page; in response to receiving the first interaction operation of the target user, and when the content volume of the search result entries corresponding to the initial search text that have been displayed is greater than or equal to the content volume threshold, the third page including the search result entries corresponding to the first recommended search text is displayed (i.e., trigger the search process based on the target search text and display the results to the user).

[0091] By applying the above page display steps, it is possible to determine whether to jump and trigger the target search text recommended to the target user based on the content volume that has been browsed when the user browses the search result page corresponding to the initial search text by using the first interaction operation (i.e., the result page actively triggered by the user), so that the user can more conveniently trigger the display of the recommended content, achieve lightweight activation, improve the click-through rate of the recommended content, and improve the recommendation effect. In some examples, the user may actively search on a web platform or a search software platform, and then trigger the display of the search result page by sending the initial search text to the search engine. The first page for displaying the search result entries may be a web page or the display interface of the search software. In this case, the search engine can recall multiple search result entries based on the initial search text sent by the user, and then display a certain number of search result entries on the first page.

[0092] In some examples, the second behavior data applied in the foregoing step S201 may be collected during the above page display process, so as to be able to indicate the user's preference for the recommended search text in the search scenario, and based on this, more accurate recommendations can be realized.

[0093] According to one aspect of the present disclosure, a recommendation model is further provided. Figure 4 The schematic diagram of the recommendation model 400 according to an exemplary embodiment of the present disclosure is shown. As Figure 4 shown, the recommendation model 400 includes:

[0094] A first interoperability network 410, configured to output a first interoperability result based on the browsing behavior data of the target user input to the first interoperability network and the candidate search text set, where the browsing behavior data includes a plurality of first search texts and the browsing behavior information of the target user browsing the search result pages corresponding to each first search text;

[0095] A second interoperability network 420, configured to output a second interoperability result based on content interaction behavior data of a target user input to the second interoperability network and a candidate search text set, where the content interaction behavior data includes a plurality of second search texts and behavior information of the target user performing interaction operations on corresponding search result pages for each second search text;

[0096] A prediction network 430, configured to determine first prediction information capable of indicating the probability that the target user browses the search result page corresponding to each candidate search text in the candidate search text set and second prediction information capable of indicating the probability that the target user performs an interaction operation on the search result page corresponding to each candidate search text in the candidate search text set based on the first interoperability result, the second interoperability result, the browsing behavior data, the content interaction behavior data, and the candidate search text set input to the prediction network; and

[0097] An output layer 440, configured to determine a target search text for recommending to the target user from the candidate search text set based on the first prediction information and the second prediction information.

[0098] By constructing multiple network layers for implementing interoperability in the recommendation model, interoperability between behavior data of different behavior categories (specifically, browsing behavior data and content interaction behavior data) and candidate content can be achieved in each interoperability network layer, so as to more accurately predict different categories of interaction behaviors of users, and more accurate recommendations can be implemented based on this.

[0099] According to some embodiments, the recommendation model 400 further includes: a recall network, configured to recall a plurality of candidate search texts in the candidate search text set from a search text library based on at least one of the first behavior data and the second behavior data of the target user, where the first behavior data includes a plurality of user search texts historically searched by the target user and user behavior information corresponding to each user search text, and the second behavior data includes a plurality of recommended search texts historically interacted by the target user and user behavior information corresponding to each recommended search text.

[0100] Figure 5 Shows a schematic diagram of a recall network 500 according to an exemplary embodiment of the present disclosure. As Figure 5As shown, the recall network 500 includes an embedding layer 501, a multi-layer perceptron network 502, an attention network 503, a multi-layer perceptron network 504, a feature processing network 505, and a recall layer 506. In this example, the input information of the recall network includes user-side information (specifically user features and user behavior data) and candidate content-side information (such as search text information in a search text library). The embedding layer 501 is used to perform an embedding representation on the input information of the recall network to obtain an embedded feature vector mapped to the semantic space, facilitating the model network to further learn and process the semantic features of the input information.

[0101] See Figure 5 , in this example, the multi-layer perceptron network 502 and the multi-layer perceptron network 504 are respectively used to process user features and search text information, and the attention network can perform attention mechanism-based feature extraction based on the output result of the multi-layer perceptron network 502 (the feature extraction result corresponding to user features) and user behavior data to obtain user interest features. On this basis, the feature processing network 505 and the recall layer 506 can determine the recalled content based on the output result of the attention network 503 (the user interest feature obtained based on user-side information) and the output result of the multi-layer perceptron network 504 (the candidate content feature obtained based on candidate content-side information), that is, obtain a candidate search text set that can enter the sorting step.

[0102] Figure 6 shows a schematic diagram of a sorting network 600 according to an exemplary embodiment of the present disclosure. As Figure 6 shown, the sorting network 600 includes an embedding layer 601, a fully connected network 602, an embedding layer 603, a multi-layer perceptron network 604, an attention network 605, an attention network 606, a prediction network 607, and an output layer 608. In this example, the input information of the sorting network includes user-side information (specifically user features, browsing behavior data grouped according to behavior categories, and content interaction behavior data) and information of the candidate search text set. Similar to the embedding representation process in the recall network described above, the embedding layer 601 and the embedding layer 602 are respectively used to perform an embedding representation on the user features and the candidate search text set to obtain an embedded feature vector mapped to the semantic space, facilitating the model network to further learn and process the semantic features of the input information.

[0103] See Figure 6, in this example, the fully-connected network 602 and the multi-layer perceptron network 604 are respectively used to perform more comprehensive feature extraction and processing based on the user feature information and the information of the candidate search text set. The attention network 605 and the attention network 606 can respectively perform attention interoperability between the user behavior data and the candidate search text based on the browsing behavior data and the content interaction behavior data, so as to more accurately extract the correlation features between the user's historical browsing behavior and the candidate search text and the correlation features between the user's historical content interaction behavior and the candidate search text. On this basis, the prediction network 607 can make predictions based on the output results of the fully-connected network 602 (corresponding to user features) and the attention network, so as to obtain the first prediction information that can indicate the probability that the target user browses the search result page corresponding to the candidate search text and the second prediction information that can indicate the probability that the target user performs an interaction operation on the search result page corresponding to the candidate search text. The output layer 608 can then sort the multiple candidate search texts based on the first prediction information and the second prediction information, and output the target search text that can be used to recommend to the target user based on the sorting result.

[0104] According to an aspect of the present disclosure, there is also provided a recommendation device 700. Figure 7 The structural block diagram of the recommendation device 700 according to an exemplary embodiment of the present disclosure is shown. As Figure 7 shown, the recommendation device 700 includes:

[0105] A first acquisition unit 701, configured to acquire the first behavior data and the second behavior data of the target user, where the first behavior data includes a plurality of user search texts searched by the target user history and the user behavior information corresponding to each user search text, and the second behavior data includes a plurality of recommended search texts interacted by the target user history and the user behavior information corresponding to each recommended search text;

[0106] A first determination unit 702, configured to determine the browsing behavior data and the content interaction behavior data based on the first behavior data and the second behavior data, where the browsing behavior data includes a plurality of first search texts and the browsing behavior information of the target user browsing the search result page corresponding to each first search text, and the content interaction behavior data includes a plurality of second search texts and the behavior information of the target user performing an interaction operation on the search result page corresponding to each second search text;

[0107] A second determination unit 703, configured to determine the target search text based on the browsing behavior data and the content interaction behavior data; and

[0108] A recommendation unit 704, configured to recommend the target search text to the target user.

[0109] According to some embodiments, the second determination unit 703 includes: a first determination subunit, configured to determine a candidate search text set; a second determination subunit, configured to determine first prediction information capable of indicating the probability that the target user browses the search result page corresponding to each candidate search text in the candidate search text set based on the browsing behavior data; a third determination subunit, configured to determine second prediction information capable of indicating the probability that the target user performs an interaction operation in the search result page corresponding to each candidate search text in the candidate search text set based on the content interaction behavior data; and a fourth determination subunit, configured to determine the target search text from the candidate search text set based on the first prediction information and the second prediction information.

[0110] According to some embodiments, the second determination subunit includes: a first input module, configured to input the browsing behavior data and each candidate search text in the candidate search text set into a first interoperability network; and a first determination module, configured to determine the first prediction information based on the output result of the first interoperability network.

[0111] According to some embodiments, the first interoperability network is an attention network.

[0112] According to some embodiments, the first determination module includes: a second input module, configured to input the output result of the first interoperability network, the browsing behavior data, the content interaction behavior data, and the candidate search text set into a prediction network; and a second determination module, configured to determine the first prediction information based on the output result of the prediction network.

[0113] According to some embodiments, the apparatus 700 further includes: a third determination unit, configured to determine information about a display page for displaying the target search text to the target user, where the information about the display page includes trigger behavior information for the target user to enter the display page, and the input to the prediction network further includes the information about the display page.

[0114] According to some embodiments, the first determination subunit is configured to: recall a plurality of candidate search texts in the candidate search text set from a search text library based on at least one of the first behavior data and the second behavior data.

[0115] According to some embodiments, the first determination subunit includes: a third determination module, configured to determine an interest feature of the target user based on at least one of the first behavior data and the second behavior data; and a recall module, configured to recall the plurality of candidate search texts from the search text library based on the interest feature.

[0116] According to some embodiments, the apparatus 700 further includes: a fourth determination unit configured to determine a plurality of user interest words based on at least one of the plurality of user search texts and the plurality of recommended search texts; wherein, the recall module is configured to: recall the plurality of candidate search texts from the search text library based on the interest features and the plurality of user interest words.

[0117] According to some embodiments, the plurality of user interest words include at least one of the following: a plurality of historical search words obtained by performing a word segmentation operation on at least one of the plurality of user search texts and the plurality of recommended search texts; and content label words corresponding to at least one of the plurality of user search texts and the plurality of recommended search texts.

[0118] According to some embodiments, the apparatus 700 further includes: a fifth determination unit configured to determine an interest weight of each user interest word in the plurality of user interest words based on at least one of the first behavior data and the second behavior data, wherein, the recall module is configured to: recall the plurality of candidate search texts from the search text library based on the interest features, the plurality of user interest words, and the interest weight of each user interest word.

[0119] According to some embodiments, the apparatus 700 further includes: a sixth determination unit configured to determine at least one popular search text from the search text library based on the user interaction data of each search text in the search text library; and an adding unit configured to add the at least one popular search text to the candidate search text set.

[0120] According to some embodiments, the apparatus 700 further includes: a seventh determination unit configured to determine the historical recommendation times of recommending the target search text to the target user, wherein the second determination unit 703 is further configured to, in response to the historical recommendation times exceeding a times threshold, re-determine the target search text based on the browsing behavior data and the content interaction behavior data.

[0121] According to some embodiments, the second determination unit 703 is further configured to, in response to determining that the target search text is the same as any one of the plurality of user search texts and the plurality of recommended search texts, re-determine the target search text based on the browsing behavior data and the content interaction behavior data.

[0122] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision, and disclosure of the user's personal information involved all comply with the provisions of relevant laws and regulations and do not violate public order and good customs.

[0123] According to one aspect of the present disclosure, there is also provided an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the above-mentioned recommendation method.

[0124] According to one aspect of the present disclosure, there is also provided a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the computer to execute the above-mentioned recommendation method.

[0125] According to one aspect of the present disclosure, there is also provided a computer program product, including a computer program, wherein the computer program implements the above-mentioned recommendation method when executed by a processor.

[0126] Referring to Figure 8 , a block diagram of an electronic device 800 that can be a server or a client of the present disclosure will now be described. It is an example of a hardware device that can be applied to various aspects of the present disclosure. The electronic device is intended to represent various forms of digital electronic computer devices, such as, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, a personal digital processor, a cellular phone, a smart phone, a wearable device, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0127] As Figure 8 shown, the device 800 includes a computing unit 801, which can execute various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other through a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0128] Multiple components in device 800 are connected to I / O interface 805, including: input unit 806, output unit 807, storage unit 808, and communication unit 809. The input unit 806 can be any type of device capable of inputting information into device 800. The input unit 806 can receive input digital or character information, and generate key signal inputs related to user settings and / or function controls of the electronic device, and can include, but are not limited to, a mouse, keyboard, touch screen, trackpad, trackball, joystick, microphone, and / or remote control. The output unit 807 can be any type of device capable of presenting information, and can include, but are not limited to, a display, speaker, video / audio output terminal, vibrator, and / or printer. The storage unit 808 can include, but is not limited to, magnetic disks and optical discs. The communication unit 809 allows device 800 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks, and can include, but are not limited to, a modem, network card, infrared communication device, wireless communication transceiver, and / or chipset, such as a Bluetooth™ device, 802.11 device, WiFi device, WiMax device, cellular communication device, and / or the like.

[0129] The computing unit 801 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 executes the various methods and processes described above, such as the recommendation method. For example, in some embodiments, the recommendation method can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 808. In some embodiments, part or all of the computer program can be loaded and / or installed onto device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the recommendation method described above can be executed. Alternatively, in other embodiments, the computing unit 801 can be configured to execute the recommendation method in any other suitable manner (e.g., by means of firmware).

[0130] The various embodiments of the systems and techniques described above in this specification can be implemented in digital electronic circuitry, integrated circuit systems, field programmable gate arrays (FPGA), application specific integrated circuits (ASIC), application specific standard products (ASSP), systems-on-chip (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be a special-purpose or general-purpose programmable processor that receives data and instructions from, and transmits data and instructions to, a storage system, at least one input device, and at least one output device.

[0131] The program code for implementing the methods of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing device, such that the program codes, when executed by the processor or controller, cause the functions / operations specified in the flowchart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine, or entirely on the remote machine or server.

[0132] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0133] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and a pointing device (e.g., a mouse or a trackball), by which the user can provide input to the computer. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, speech input, or tactile input).

[0134] The systems and techniques described herein can be implemented in a computing system including backend components (e.g., as a data server), or a computing system including middleware components (e.g., an application server), or a computing system including frontend components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system including any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected to each other by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), the Internet, and blockchain networks.

[0135] A computer system can include a client and a server. The client and the server are generally far from each other and usually interact through a communication network. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, or a server of a distributed system, or a server incorporating blockchain.

[0136] It should be understood that various forms of the processes shown above can be used, with steps reordered, added, or deleted. For example, the steps recited in this disclosure can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and no limitations are imposed herein.

[0137] Although embodiments or examples of the present disclosure have been described with reference to the accompanying drawings, it should be understood that the above methods, systems, and devices are merely exemplary embodiments or examples, and the scope of the present invention is not limited by these embodiments or examples, but is only defined by the authorized claims and their equivalent scope. Various elements in the embodiments or examples may be omitted or replaced by their equivalent elements. In addition, the various steps may be performed in an order different from that described in the present disclosure. Further, the various elements in the embodiments or examples may be combined in various ways. Importantly, with the evolution of technology, many of the elements described herein may be replaced by equivalent elements that emerge after the present disclosure.

Claims

1. A recommendation method, comprising: Acquire first behavior data and second behavior data of the target user, wherein the first behavior data includes multiple user search texts of the target user's historical searches and user behavior information corresponding to each user search text, and the second behavior data includes multiple recommended search texts of the target user's historical interactions and user behavior information corresponding to each recommended search text; Determine browsing behavior data and content interaction behavior data based on the first behavior data and the second behavior data, wherein the browsing behavior data includes a plurality of first search texts and browsing behavior information of the target user browsing the search result page corresponding to each first search text, and the content interaction behavior data includes a plurality of second search texts and behavior information of the target user performing interactive operations on the search result page corresponding to each second search text; Determining a target search text based on the browsing behavior data and the content interaction behavior data; and The target search text is recommended to the target user.

2. The method of claim 1, wherein: The determining the target search text based on the browsing behavior data and the content interaction behavior data includes: determining a candidate search text set; Based on the browsing behavior data, determining first prediction information that can indicate the probability of the target user browsing a search result page corresponding to each candidate search text in the candidate search text set; Based on the content interaction behavior data, determining second prediction information that can indicate the probability of the target user performing an interaction operation in a search result page corresponding to each candidate search text in the candidate search text set; and The target search text is determined from the candidate search text set based on the first prediction information and the second prediction information.

3. The method of claim 2, wherein: The determining, based on the browsing behavior data, first prediction information that can indicate the probability of the target user browsing the search result page corresponding to each candidate search text in the candidate search text set includes: Inputting the browsing behavior data and each candidate search text in the candidate search text set into a first interoperability network; and The first prediction information is determined based on an output result of the first interoperation network.

4. The method of claim 3, wherein: The first interoperation network is an attention network.

5. The method according to claim 3 or 4, wherein: The determining the first prediction information based on the output result of the first interoperability network includes: inputting the output result of the first interoperability network, the browsing behavior data, the content interaction behavior data, and the candidate search text set into a prediction network; and The first prediction information is determined based on an output result of the prediction network.

6. The method of claim 5, further comprising: Determine information of a display page for displaying the target search text to the target user, wherein the information of the display page includes triggering behavior information of the target user entering the display page; The input of the prediction network also includes information of the display page.

7. The method according to any one of claims 2 to 6, wherein: Determining a candidate search text set includes: Based on at least one of the first behavior data and the second behavior data, a plurality of candidate search texts in the candidate search text set are recalled from a search text library.

8. The method of claim 7, wherein: The step of recalling a plurality of candidate search texts in the candidate search text set from a search text library based on at least one of the first behavior data and the second behavior data comprises: Determining an interest feature of the target user based on at least one of the first behavior data and the second behavior data; and Based on the interest feature, the plurality of candidate search texts are recalled from the search text library.

9. The method of claim 8, further comprising: determining a plurality of user interest words based on at least one of the plurality of user search texts and the plurality of recommended search texts; The step of recalling the plurality of candidate search texts from the search text library based on the interest feature includes: Based on the interest feature and the plurality of user interest words, the plurality of candidate search texts are recalled from the search text library.

10. The method of claim 9, wherein: The plurality of user interest words include at least one of the following: A plurality of historical search words obtained by performing a word segmentation operation on at least one of the plurality of user search texts and the plurality of recommended search texts; and The content tag word corresponding to at least one of the multiple user search texts and the multiple recommended search texts.

11. The method according to claim 9 or 10, further comprising: determining an interest weight of each user interest word among the plurality of user interest words based on at least one of the first behavior data and the second behavior data, The step of recalling the plurality of candidate search texts from the search text library based on the interest feature includes: The plurality of candidate search texts are recalled from the search text library based on the interest feature, the plurality of user interest words and the interest weight of each user interest word.

12. The method according to any one of claims 7 to 11, further comprising: determining at least one hot search text from the search text library based on user interaction data of each search text in the search text library; as well as The at least one hot search text is added to the candidate search text set.

13. The method of any one of claims 1 to 12, further comprising: Determine the historical number of times the target search text is recommended to the target user; as well as In response to the historical recommendation times exceeding a times threshold, the target search text is re-determined based on the browsing behavior data and the content interaction behavior data.

14. The method of any one of claims 1 to 13, further comprising: In response to determining that the target search text is the same as any one of the multiple user search texts and the multiple recommended search texts, the target search text is re-determined based on the browsing behavior data and the content interaction behavior data.

15. A recommendation model, comprising: A first interoperability network, configured to output a first interoperability result based on browsing behavior data of a target user input into the first interoperability network and a set of candidate search texts, wherein the browsing behavior data includes a plurality of first search texts and browsing behavior information of the target user browsing a search result page corresponding to each first search text; A second interoperability network, configured to output a second interoperability result based on content interaction behavior data of a target user input into the second interoperability network and a set of candidate search texts, wherein the content interaction behavior data includes a plurality of second search texts and behavior information of the target user performing an interaction operation on a search result page corresponding to each second search text; A prediction network, for determining first prediction information capable of indicating a probability that the target user browses a search result page corresponding to each candidate search text in the candidate search text set and second prediction information capable of indicating a probability that the target user performs an interaction operation in the search result page corresponding to each candidate search text in the candidate search text set based on the first interoperation result, the second interoperation result, the browsing behavior data, the content interaction behavior data and the candidate search text set input into the prediction network; and The output layer is used to determine a target search text for recommendation to the target user from the candidate search text set based on the first prediction information and the second prediction information.

16. The recommendation model according to claim 15, further comprising: A recall network is used to recall multiple candidate search texts in the candidate search text set from a search text library based on at least one of the first behavior data and the second behavior data of the target user, wherein the first behavior data includes multiple user search texts historically searched by the target user and user behavior information corresponding to each user search text, and the second behavior data includes multiple recommended search texts historically interacted with by the target user and user behavior information corresponding to each recommended search text.

17. A recommendation device, comprising: A first acquisition unit is configured to acquire first behavior data and second behavior data of a target user, wherein the first behavior data includes a plurality of user search texts historically searched by the target user and user behavior information corresponding to each user search text, and the second behavior data includes a plurality of recommended search texts historically interacted by the target user and user behavior information corresponding to each recommended search text; A first determination unit is configured to determine browsing behavior data and content interaction behavior data based on the first behavior data and the second behavior data, wherein the browsing behavior data includes a plurality of first search texts and browsing behavior information of the target user browsing a search result page corresponding to each first search text, and the content interaction behavior data includes a plurality of second search texts and behavior information of the target user performing an interactive operation on a search result page corresponding to each second search text; A second determining unit is configured to determine a target search text based on the browsing behavior data and the content interaction behavior data; and The recommendation unit is configured to recommend the target search text to the target user.

18. An electronic device, comprising: at least one processor; as well as a memory communicatively coupled to the at least one processor; in The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 14.

19. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to make a computer execute the method according to any one of claims 1-14.

20. A computer program product comprising a computer program, wherein: The computer program implements the method according to any one of claims 1 to 14 when executed by a processor.

Images