Construction method of generative heterogeneous graph model, agent distribution method and related device

Abstract

The invention provides a construction method of a generative heterogeneous graph model, an agent distribution method and a related device, and relates to the technical field of artificial intelligence such as generative models, generative search, graph models, agents and object distribution. The method comprises the following steps: extracting a first interaction action chain containing agent distribution behaviors exceeding a preset satisfaction degree from historical search behaviors and search behavior feedback of a first sample user; extracting a second interaction action chain containing agent distribution behaviors exceeding the preset satisfaction from historical recommended behaviors and recommendation behavior feedback of a second sample user; and on the basis of the first interaction action chain and the second interaction action chain, by taking the user, the behavior object and the agent as different nodes and taking the interaction action as an edge for connecting the corresponding nodes, constructing a generative heterogeneous graph model. The method comprises the following steps: acquiring user related information of a target user; inputting the user related information into a generative heterogeneous graph model obtained by pre-training to obtain a target agent output by the generative heterogeneous graph model; and distributing the target agent to the target user.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing, specifically to the fields of artificial intelligence technology such as generative models, graphical models, intelligent agents, and object distribution, and in particular to a method for constructing a generative heterogeneous graphical model and an intelligent agent distribution method, as well as devices, electronic devices, computer-readable storage media, and computer program products corresponding to the methods. Background Art

[0002] The mainstream content distribution method is to use a single search system or recommendation system to achieve it, that is, to recall, sort, and mix heterogeneous results of related web pages through a multi-layer system funnel. Each strategic funnel sorts web pages based on basic relevance, user feedback behavior, authority and other information and truncates the output to the next layer.

[0003] That is, in the search scenario, the search system is used to understand the content of the search terms or search statements entered by the user, and match the search results according to the content relevance; in the recommendation scenario, the recommendation system is used to form a user profile based on the user's historical behavior or user preferences, and push content that is considered to be of interest to the user based on the user profile. Summary of the invention

[0004] The embodiments of the present disclosure propose a method for constructing a generative heterogeneous graph model and an agent distribution method, as well as a device, an electronic device, a computer-readable storage medium, and a computer program product corresponding to the method.

[0005] In a first aspect, an embodiment of the present disclosure proposes a method for constructing a generative heterogeneous graph model, comprising: extracting a first interactive action chain including agent distribution behaviors exceeding a preset satisfaction level from a first sample user's historical search behaviors and search behavior feedback; extracting a second interactive action chain including agent distribution behaviors exceeding a preset satisfaction level from a second sample user's historical recommended behaviors and recommendation behavior feedback; constructing a generative heterogeneous graph model based on the first interactive action chain and the second interactive action chain, with users, behavior objects, and agents as different nodes and interactive actions as edges connecting corresponding nodes.

[0006] In the second aspect, an embodiment of the present disclosure proposes a device for constructing a generative heterogeneous graph model, including: a first interactive action chain extraction unit, configured to extract a first interactive action chain including an agent distribution behavior exceeding a preset satisfaction level from a first sample user's historical search behavior and search behavior feedback; a second interactive action chain extraction unit, configured to extract a second interactive action chain including an agent distribution behavior exceeding a preset satisfaction level from a second sample user's historical recommended behavior and recommendation behavior feedback; a generative heterogeneous graph model construction unit, configured to construct a generative heterogeneous graph model based on the first interactive action chain and the second interactive action chain, with users, behavior objects, and agents as different nodes and interactive actions as edges connecting corresponding nodes.

[0007] In a third aspect, an embodiment of the present disclosure proposes an agent distribution method, including: obtaining user-related information of a target user; inputting the user-related information into a pre-trained generative heterogeneous graph model to obtain a target agent output by the generative heterogeneous graph model, wherein the generative heterogeneous graph model is obtained according to the method for constructing a generative heterogeneous graph model as described in the first aspect; and distributing the target agent to the target user.

[0008] In a fourth aspect, an embodiment of the present disclosure proposes an agent distribution device, comprising: a user-related information acquisition unit, configured to acquire user-related information of a target user; a model usage unit, configured to input user-related information into a pre-trained generative heterogeneous graph model to obtain a target agent output by the generative heterogeneous graph model, wherein the generative heterogeneous graph model is obtained according to the construction device of the generative heterogeneous graph model described in the second aspect; and an agent distribution unit, configured to distribute the target agent to the target user.

[0009] In a fifth aspect, an embodiment of the present disclosure provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that when the at least one processor executes, it is able to implement the method for constructing a generative heterogeneous graph model as described in the first aspect or the agent distribution method as described in the third aspect.

[0010] In a sixth aspect, an embodiment of the present disclosure provides a non-transitory computer-readable storage medium storing computer instructions, which are used to enable a computer to implement the method for constructing a generative heterogeneous graph model as described in the first aspect or the agent distribution method as described in the third aspect when executed.

[0011] In the seventh aspect, an embodiment of the present disclosure provides a computer program product comprising a computer program, which, when executed by a processor, can implement the method for constructing a generative heterogeneous graph model as described in the first aspect or the agent distribution method as described in the third aspect.

[0012] The construction scheme of the generative heterogeneous graph model provided by the embodiment of the present disclosure extracts a first interactive action chain including agent distribution behaviors that are considered satisfactory by users from historical search behaviors and search behavior feedback in search scenarios, and extracts a second interactive action chain including agent distribution behaviors that are considered satisfactory by users from historical recommended behaviors and recommendation behavior feedback in recommendation scenarios. Furthermore, different nodes played by users, behavior objects and agents and edges played by interactive actions can be extracted from each interactive action chain covering the search scenario and the recommendation scenario, and finally a generative heterogeneous graph model can be constructed based on these nodes and edges connecting different nodes, so that the constructed generative heterogeneous graph model integrates the association relationship between different users and different agents in the search scenario and the recommendation scenario in the form of a graph structure.

[0013] The agent distribution method provided by the embodiment of the present disclosure can, after obtaining user-related information of a target user, directly input the user-related information into a generative heterogeneous graph model constructed in the above manner, so as to quickly obtain a target agent that matches the target user without distinguishing between scene types. That is, the generative heterogeneous graph model can be used to distribute matching agents to users in all scenarios.

[0014] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Other features, objects and advantages of the present disclosure will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0016] Figure 1 is an exemplary system architecture in which the present disclosure may be applied;

[0017] Figure 2 A flowchart of a method for constructing a generative heterogeneous graph model provided in an embodiment of the present disclosure;

[0018] Figure 3 A flowchart of a method for determining a first interactive action chain corresponding to a search scenario provided by an embodiment of the present disclosure;

[0019] Figure 4A flowchart of a method for determining a second interactive action chain corresponding to a recommended scenario provided by an embodiment of the present disclosure;

[0020] Figure 5 A flowchart of a method for training a generative graph model to be trained to obtain a generative heterogeneous graph model provided in an embodiment of the present disclosure;

[0021] Figure 6 A flow chart of an agent distribution method provided in an embodiment of the present disclosure;

[0022] Figure 7 A flowchart of a method for distributing matching target agents to users according to acquired user search requirements provided by an embodiment of the present disclosure;

[0023] Figure 8 A flowchart of a method for distributing matching target agents to users according to determined user preference information provided by an embodiment of the present disclosure;

[0024] Figure 9-1 and Figure 9-2 They are respectively a system architecture diagram and an example diagram provided in combination with specific application scenarios according to the embodiments of the present disclosure;

[0025] Fig.10 A structural block diagram of a device for constructing a generative heterogeneous graph model provided in an embodiment of the present disclosure;

[0026] Fig.11 A structural block diagram of an intelligent agent distribution device provided in an embodiment of the present disclosure;

[0027] Fig.12 A schematic diagram of the structure of an electronic device suitable for executing a method for constructing a generative heterogeneous graph model and / or an agent distribution method provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0028] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description. It should be noted that the embodiments in the present disclosure and the features in the embodiments may be combined with each other without conflict.

[0029] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0030] Figure 1 An exemplary system architecture 100 is shown to which the method for constructing a generative heterogeneous graph model, the agent distribution method, and corresponding devices, electronic devices, and computer-readable storage media of the present application can be applied.

[0031] like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0032] The user can use the terminal devices 101, 102, 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various applications for realizing information communication between the terminal devices 101, 102, 103 and the server 105 can be installed, such as information search applications, information recommendation applications, search and push integrated applications, model training or model building applications, intelligent agent distribution applications, etc.

[0033] Terminal devices 101, 102, 103 and server 105 can be hardware or software. When terminal devices 101, 102, 103 are hardware, they can be various electronic devices with display screens, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, etc.; when terminal devices 101, 102, 103 are software, they can be installed in the electronic devices listed above, which can be implemented as multiple software or software modules, or as a single software or software module, and no specific limitation is made here. When server 105 is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or as a single server; when the server is software, it can be implemented as multiple software or software modules, or as a single software or software module, and no specific limitation is made here.

[0034] The server 105 can provide various services through various built-in applications. Taking the agent distribution application that can provide users with agent distribution services as an example, the server 105 can achieve the following effects when running the agent distribution application: first, receive user-related information transmitted by the target user through the terminal devices 101, 102, and 103 used by the target user through the network 104; then, input the user-related information into the pre-trained generative heterogeneous graph model to obtain the target agent output by the generative heterogeneous graph model; finally, distribute the target agent to the application used by the target user.

[0035] Among them, the generative heterogeneous graph model can be trained by the built-in model building application on the server 105 according to the following steps: first, from the historical search behavior and search behavior feedback of the first sample user, a first interactive action chain containing the intelligent agent distribution behavior exceeding the preset satisfaction is extracted; then, from the historical recommended behavior and recommendation behavior feedback of the second sample user, a second interactive action chain containing the intelligent agent distribution behavior exceeding the preset satisfaction is extracted; then, based on the first interactive action chain and the second interactive action chain, a generative heterogeneous graph model is constructed by treating users, behavior objects, and intelligent agents as different nodes and using interactive actions as edges connecting corresponding nodes.

[0036] Since training a generative heterogeneous graph model requires more computing resources and stronger computing power, the methods for constructing a generative heterogeneous graph model provided in the subsequent embodiments of the present application are generally performed by a server 105 having stronger computing power and more computing resources. Accordingly, the construction device of the generative heterogeneous graph model is generally also arranged in the server 105. However, it should also be pointed out that when the terminal devices 101, 102, and 103 also have computing power and computing resources that meet the requirements, the terminal devices 101, 102, and 103 can also complete the various operations assigned to the server 105 by the construction class application of the generative heterogeneous graph model installed thereon, and then output the same result as the server 105. Correspondingly, the construction device of the generative heterogeneous graph model can also be arranged in the terminal devices 101, 102, and 103. In this case, the exemplary system architecture 100 may also not include the server 105 and the network 104.

[0037] Of course, the server used to train the generative heterogeneous graph model may be different from the server that calls the trained generative heterogeneous graph model for use. In particular, the generative heterogeneous graph model trained by the server 105 can also be used to obtain a lightweight generative heterogeneous graph model suitable for being placed in the terminal devices 101, 102, and 103 by model distillation, that is, it is possible to flexibly choose to use the lightweight generative heterogeneous graph model in the terminal devices 101, 102, and 103 according to the recognition accuracy required in actual needs, or to choose to use the more complex generative heterogeneous graph model in the server 105.

[0038] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0039] Please refer to Figure 2 , Figure 2 A flowchart of a method for constructing a generative heterogeneous graph model provided in an embodiment of the present disclosure, wherein process 200 includes the following steps:

[0040] Step 201: extracting a first interactive action chain including an agent distribution behavior exceeding a preset satisfaction level from the historical search behavior and search behavior feedback of a first sample user;

[0041] This step is intended to be performed by the execution subject of the construction method of the generative heterogeneous graph model (e.g. Figure 1 The server 105 shown in the figure extracts the search interaction action chain including the agent distribution behavior exceeding the preset satisfaction level from the historical search behavior and search behavior feedback of the first sample user, and uses it as the first interaction action chain for subsequent use. Exceeding the preset satisfaction level is used to indicate that the user is satisfied and recognized with the actual agent distributed by the agent distribution behavior, and the satisfaction level or recognition level exceeds a predetermined threshold, that is, the actual agent distributed should be an agent that satisfies the user's expectations or intentions that the user wants to achieve through the search behavior. The threshold can be preset, and can usually be based on a feedback mechanism (such as user feedback rating, click-through rate, dwell time, etc.) to evaluate whether the performance of the actual agent distributed really meets the user's actual needs.

[0042] Among them, historical search behavior refers to all search records made by users during their previous interactions with the search system. These search records contain information such as the keywords used by users, pages viewed, links clicked, etc.; search behavior feedback refers to further operations or feedback that users may perform in search results (such as clicking on a certain result, how long they spend viewing a certain page, how they rate certain results, etc.). These behaviors can help the search system understand the user's satisfaction or preference for specific search results.

[0043] Among them, the agent in this disclosure refers to an entity that is encapsulated based on technologies such as generative large language models (such as GPT series models, BERT models, etc.) and can perform specific functions according to user needs. It is usually based on a generative large language model, and is encapsulated on this basis in combination with preset role settings or function settings or scene settings parameters. Specifically, these agents can provide users with information retrieval, problem answering, task execution and other services through natural language understanding, generation and reasoning capabilities. For example, a travel planning agent used to provide travel planning services to users, a poetry generation agent used to provide poetry generation services to users, a digital human agent used to provide immersive chat dialogue services to users, and so on.

[0044] Among them, the search interaction action chain refers to multiple interaction steps or operation processes between the user and the search system. These interaction actions start from the user initiating the query to the search system responding, and may include further user input, system feedback, and final result display, etc. The first interaction action chain specifically refers to the partial search interaction action chain that includes the intelligent agent distribution behavior that is satisfied or recognized by the user.

[0045] Step 202: extracting a second interactive action chain including an agent distribution behavior exceeding a preset satisfaction level from the historical recommended behaviors and recommended behavior feedback of the second sample user;

[0046] On the basis of step 201, this step aims to extract the recommended interactive action chain containing the agent distribution behavior exceeding the preset satisfaction level from the historical recommended behavior and recommendation behavior feedback of the second sample user, and use it as the second interactive action chain for subsequent use. Exceeding the preset satisfaction level is used to indicate that the user is satisfied and recognized with the actual agent distributed by the agent distribution behavior, and the satisfaction or recognition level exceeds a predetermined threshold, that is, the actual agent distributed should be an agent that better meets the user's potential needs. The threshold can be preset, and can usually be based on a feedback mechanism (which may be the user's satisfaction score, click, acceptance, subsequent behavior such as purchase, continued browsing, etc., or even explicit feedback such as rating, comment, like, etc.) to evaluate whether the performance of the actual agent distributed does meet the user's potential needs.

[0047] Among them, historical recommended behavior refers to the recommendation information that the user has previously received in the recommendation system. Specifically, it may be content recommended based on the user's interests, search history, clicks, purchases, and other behaviors. These behaviors can help the recommendation system understand the user's past interest trends and needs; recommendation behavior feedback refers to the feedback information given by the user after receiving the recommended content, which may be the user's satisfaction score, click, acceptance or not, subsequent behavior (such as purchase, continued browsing, etc.), or even clear feedback (such as ratings, comments, likes, etc.). This feedback information provides the system with important data on whether the recommendation is successful and whether it meets the user's expectations.

[0048] Among them, the recommendation interaction action chain usually refers to a feedback chain or behavior chain formed in the process of interaction between users and recommendation systems. This chain consists of multiple steps, each of which may involve a certain action of the user (for example, click, query, score, confirmation, etc.) and the response of the recommendation system (such as agent recommendation, feedback adjustment, etc.). Simply put, the recommendation interaction action chain includes the behavior sequence that the recommendation system adjusts and optimizes according to the feedback after the user gives feedback on the recommendation behavior. For example, the user clicks on a recommended agent, indicating interest, satisfaction or dissatisfaction with the recommendation result. The system further optimizes the recommendation algorithm based on this feedback to form a new recommendation or improved agent distribution behavior. The second interaction action chain specifically refers to the part of the recommendation interaction action chain that contains the agent distribution behavior that is satisfied or recognized by the user.

[0049] Step 203: Based on the first interaction action chain and the second interaction action chain, a generative heterogeneous graph model is constructed in the form of taking users, behavior objects, and agents as different nodes and taking interaction actions as edges connecting corresponding nodes.

[0050] Based on step 202, this step aims to have the above-mentioned execution subject extract the users, behavior objects, and intelligent agents contained in the first interaction action chain corresponding to the search scenario and the second interaction action chain corresponding to the recommendation scenario as different nodes, and then extract the interaction actions as edges connecting the corresponding nodes, and use these nodes and edges to construct a generative heterogeneous graph model.

[0051] Among them, the generative heterogeneous graph model is a technology that combines graph neural networks (GNN) and generative models (such as generative adversarial networks, GAN), which aims to learn the node representation of heterogeneous graphs and generate new nodes or edges. The following is an introduction to some key points of the generative heterogeneous graph model:

[0052] Heterogeneous Graph contains multiple types of nodes and edges, which can represent different entities and relationships. For example, in a movie recommendation system, nodes can be users, movies, directors, and edges can be users' ratings of movies, similarities between movies, etc.

[0053] Generative Model generates new data samples by learning the distribution of data. In the context of graphs, this means that the model can not only learn the representation of nodes, but also generate new nodes or edges, thereby expanding the structure of the graph.

[0054] GNN is a neural network that specializes in processing graph-structured data. It can capture the neighborhood information of nodes and update the representation of nodes. In heterogeneous graphs, GNN needs to process different types of nodes and edges, which is usually achieved through technologies such as heterogeneous graph attention networks.

[0055] The generative heterogeneous graph model that the present disclosure hopes to construct specifically is expected to be able to generate and represent the relationships between different types of nodes (such as users, behavior objects, agents) and edges (interaction actions), and to capture complex interaction patterns with the help of the graph-related characteristics of this model, so as to provide users with matching agent distribution in all scenarios.

[0056] The behavior object may include at least one of the following:

[0057] User-entered search terms / search statements: refers to the keywords or sentences that users enter into the search engine to retrieve information;

[0058] Selected target search results: The specific links or items selected by the user in the search results, which reflects the user's interests and needs;

[0059] Recommended search terms / search sentences: Keywords or sentences recommended by the recommendation system based on the user's historical behavior or current context;

[0060] Information or objects recommended to be presented in the information recommendation position: information, functional entities or products specially recommended to users in the user interface, such as advertisements, hot news, intelligent entities or products, etc.

[0061] In addition, behavior objects can also include the following other situations:

[0062] User-generated content: content posted by users on social platforms, forums or comment areas. These contents can be used as part of the behavior object; user historical interaction data: including user click history, browsing history, etc. These data can be used to build user portraits, which in turn affect the recommendation system; user location information: user location information can affect search results and recommended content, such as localized search and recommendations; user device information: the type of device used by the user (such as mobile phones, tablets, computers) and device settings. This information can be used to optimize user experience and content presentation; interactions in social networks: user interactive behaviors such as likes, shares, and comments in social networks. These behaviors can serve as social signals to affect the dissemination and recommendation of information.

[0063] The method for constructing a generative heterogeneous graph model provided by the embodiment of the present disclosure extracts a first interactive action chain including agent distribution behaviors that are considered satisfactory by users from historical search behaviors and search behavior feedback in search scenarios, and extracts a second interactive action chain including agent distribution behaviors that are considered satisfactory by users from historical recommended behaviors and recommendation behavior feedback in recommendation scenarios. Then, different nodes played by users, behavior objects, and agents and edges played by interactive actions can be extracted from each interactive action chain covering the search scenario and the recommendation scenario, and finally a generative heterogeneous graph model can be constructed based on these nodes and edges connecting different nodes, so that the constructed generative heterogeneous graph model integrates the association relationship between different users and different agents in the search scenario and the recommendation scenario in the form of a graph structure.

[0064] Please refer to Figure 3 , Figure 3 A flowchart of a method for determining a first interactive action chain corresponding to a search scenario provided by an embodiment of the present disclosure, that is, for Figure 2 Step 201 in the process 200 shown provides a specific implementation method, and the other steps in the process 200 are not adjusted. The specific implementation method provided in this embodiment is replaced by step 201 to obtain a new complete embodiment. The process 300 includes the following steps:

[0065] Step 301: Obtain historical search behavior and search behavior feedback of a first sample user;

[0066] When obtaining the historical search behavior and search behavior feedback of the first sample user, it can be performed specifically in a single historical search session as a unit, that is, the corresponding historical search behavior and search behavior feedback can be extracted from each historical search session.

[0067] Step 302: Filter the search interaction action chain containing the distribution behavior of any agent from the historical search behavior and the search behavior feedback to obtain the target search interaction action chain;

[0068] Based on step 301, this step aims to filter the search interaction action chain containing the distribution behavior of any agent from the historical search behavior and search behavior feedback by the above-mentioned execution subject to obtain the target search interaction action chain. That is, the search interaction action chain containing the distribution behavior of the agent is first filtered out and used as the target search interaction action chain for subsequent processing.

[0069] Step 303: determining the actual satisfaction of the first sample user with respect to the distributed agent according to the subsequent interaction action of the first sample user with respect to the distributed agent in the target search interaction action chain;

[0070] Based on step 302, this step aims to determine the actual satisfaction of the first sample user with the distributed intelligent agent according to the subsequent interaction actions of the first sample user with the distributed intelligent agent in the target search interaction action chain by the above-mentioned execution subject.

[0071] The subsequent interaction actions for determining the actual satisfaction level may include at least one of the following:

[0072] The duration of the first use of the distributed intelligent agent; the total usage time, total number of uses or average usage time of the distributed intelligent agent within a preset time period; whether the distributed intelligent agent has been collected or forwarded; the number of forwarding operations on the distributed intelligent agent or the forwarding channel; whether the distributed intelligent agent has been evaluated and the emotional tendency of the evaluation content, as well as other actions that can show the user's satisfaction with the actual distributed intelligent agent.

[0073] When multiple specific interactive actions mentioned above exist at the same time, each behavior can be processed into a quantitative score or rating through a quantitative algorithm, and the weighted weights converted from the influence of various interactive actions on the degree of satisfaction can be combined to calculate the comprehensive actual satisfaction using a weighted calculation method.

[0074] Step 304: Determine the target search interaction action chain with an actual satisfaction level exceeding a preset satisfaction level as the first interaction action chain.

[0075] This embodiment aims to provide a specific implementation scheme through steps 301-304, which is to evaluate the actual satisfaction of the distributed actual intelligent agents through two-stage screening and according to subsequent interactive actions, and then determine the first interactive action chain, so as to make the provided implementation method more practical and feasible, and to improve the accuracy of the determined first interactive action chain as much as possible by combining multiple evaluation items.

[0076] Please refer to Figure 4 , Figure 4 A flowchart of a method for determining a second interactive action chain corresponding to a recommended scenario provided by an embodiment of the present disclosure, that is, for Figure 2 Step 202 in the process 200 shown provides a specific implementation method, and the other steps in the process 200 are not adjusted. The specific implementation method provided in this embodiment is replaced by step 202 to obtain a new complete embodiment. The process 400 includes the following steps:

[0077] Step 401: Obtain historical recommended behaviors and recommended behavior feedback of a second sample user;

[0078] When obtaining the historical recommended behaviors and recommended behavior feedback of the second sample user, it can be specifically performed in a single historical recommendation session as a unit, that is, the corresponding historical recommended behaviors and recommended behavior feedback can be extracted from each historical recommendation session.

[0079] Step 402: Filter the recommended interactive action chains containing the distribution behaviors of any agent from the historical recommended behaviors and the recommended behavior feedback to obtain the target recommended interactive action chain;

[0080] Based on step 401, this step aims to have the above-mentioned execution subject screen the recommended interactive action chain containing the distribution behavior of any agent from the historical recommended behaviors and recommended behavior feedback to obtain the target recommended interactive action chain. That is, the recommended interactive action chain containing the distribution behavior of the agent is first screened out and used as the target recommended interactive action chain for subsequent processing.

[0081] Step 403: determining the actual satisfaction of the second sample user with respect to the distributed agent according to the subsequent interaction action of the second sample user with respect to the distributed agent in the target recommendation interaction action chain;

[0082] On the basis of step 402, this step aims to determine the actual satisfaction of the second sample user with the distributed agent according to the subsequent interaction actions of the second sample user with the distributed agent in the target recommendation interaction action chain by the above-mentioned execution subject.

[0083] The subsequent interactive actions for determining the actual satisfaction level may include at least one of the following:

[0084] The viewing time of the information recommendation position where the distributed agent is located; whether the distributed agent is run; the first running time of the distributed agent and the time interval between the second run and the first run; whether a search operation related to the function of the distributed agent is actively performed after running the distributed agent, and other actions that can show the user's satisfaction with the actual distributed agent.

[0085] When multiple specific interactive actions mentioned above exist at the same time, each behavior can be processed into a quantitative score or rating through a quantitative algorithm, and the weighted weights converted from the influence of various interactive actions on the degree of satisfaction can be combined to calculate the comprehensive actual satisfaction using a weighted calculation method.

[0086] Step 404: Determine the target recommended interactive action chain with an actual satisfaction level exceeding a preset satisfaction level as the second interactive action chain.

[0087] This embodiment aims to provide a specific implementation scheme through steps 401-404, which is to evaluate the actual satisfaction of the distributed actual intelligent agents through two-stage screening and according to subsequent interactive actions, and then determine the second interactive action chain, so as to make the provided implementation method more practical and feasible, and to improve the accuracy of the determined second interactive action chain as much as possible by combining multiple evaluation items.

[0088] Please refer to Figure 5 , Figure 5 A flowchart of a method for training a generative graph model to be trained to obtain a generative heterogeneous graph model provided in an embodiment of the present disclosure, that is, for Figure 2 Step 203 in the process 200 shown provides a specific implementation method, and the other steps in the process 200 are not adjusted. The specific implementation method provided in this embodiment is replaced by step 203 to obtain a new complete embodiment. The process 500 includes the following steps:

[0089] Step 501: Obtain a generative graph model to be trained;

[0090] The generative graph model to be trained can be a completely new model that has not been trained, or it can be a basic model obtained through preliminary training using the same training samples in similar scenarios.

[0091] Step 502: extracting user entities, behavior object entities, agent entities as different nodes, and interaction actions between different entities as edges connecting different nodes from the first interaction action chain and the second interaction action chain;

[0092] Step 503: The generative graph model to be trained is trained by utilizing the user entity, the behavior object entity, the intelligent agent entity and the interaction actions between different entities to obtain a generative heterogeneous graph model.

[0093] This embodiment provides a specific implementation method for training a generative heterogeneous graph model through steps 501-503: first, a generative graph model to be trained is required as a basic graph model, which will be used to learn the interaction patterns between users, behavior objects and agents, and finally generate new interactions; then, user entities, behavior object entities and agent entities are extracted from the first and second interaction action chains, and these entities will serve as nodes in the graph model, and interaction actions between these different entities need to be extracted, and these interaction actions will serve as edges connecting the nodes; next, the generative graph model is trained using the extracted entities and interaction actions. In this process, the model will learn how to represent these entities and the relationships between them, and the purpose of training is to enable the model to capture the complex interaction patterns between users, behavior objects and agents, and to generate new interaction actions or predict future interactions based on these patterns, and finally the generative heterogeneous graph model trained can not only understand the existing interaction data, but also generate new interaction data, thereby expanding and enriching the structure of the graph.

[0094] The above embodiments illustrate how to train a generative heterogeneous graph model from various aspects. In order to highlight the effect of the trained generative heterogeneous graph model from the actual usage scenario as much as possible, please refer to Figure 6 , Figure 6 A scheme for implementing agent distribution using a trained generative heterogeneous graph model is shown, wherein process 600 includes the following steps:

[0095] Step 601: Obtain user-related information of the target user;

[0096] The target user may be a user who has logged into the service scope provided by the execution subject among all users, or a user who has been determined to be in an active state or in use by the application installed by the execution subject. The user-related information of the user may be specifically expressed in different forms depending on whether the target user has actively initiated the search behavior.

[0097] Step 602: Inputting user-related information into a pre-trained generative heterogeneous graph model to obtain a target agent output by the generative heterogeneous graph model;

[0098] Step 603: Distribute the target agent to the target user.

[0099] Based on step 602, this step aims to distribute the target agent to the target user in a suitable form and at a suitable time by the above-mentioned execution subject.

[0100] The agent distribution method provided by the embodiment of the present disclosure can obtain the user-related information of the target user by directly inputting the user-related information into the generative heterogeneous graph model constructed in the above manner, and quickly obtain the target agent that matches the target user without distinguishing the scene type, that is, the generative heterogeneous graph model can be used to distribute the matching agent to the user in all scenes.

[0101] To deepen the understanding of the solutions provided by the above embodiments, the following will be respectively Figure 7 and Figure 8 , providing two different specific implementation methods for two different situations or scenarios, in which, Figure 7 The present disclosure provides a method for distributing a matching target agent to a user according to the acquired user search requirements, wherein the process 700 includes the following steps:

[0102] Step 701: in response to a target user actively initiating a search behavior in a current session, obtaining a search demand from the search behavior actively initiated by the target user;

[0103] Step 702: input the search requirements into the pre-trained generative heterogeneous graph model to obtain the target intelligent agent output by the production heterogeneous graph model;

[0104] Step 703: Distribute the target agent as the first candidate search result to the target user.

[0105] This embodiment aims at the situation where the target user actively initiates a search behavior in the current session. The search demand is obtained from the actively initiated search behavior, and the search demand is input into the generative heterogeneous graph model as the specific user-related information in this situation, so that the generative heterogeneous graph model can match a suitable target intelligent agent for the target user according to the input search demand.

[0106] Figure 8 A flowchart of a method for distributing a matching target agent to a user according to determined user preference information provided by an embodiment of the present disclosure, wherein the process 800 includes the following steps:

[0107] Step 801: in response to the target user not actively initiating a search behavior in the current session, obtaining user preference information from the target user's historical search session and / or historical recommendation session;

[0108] Step 802: Input the user preference information into the pre-trained generative heterogeneous graph model to obtain the target agent output by the generative heterogeneous graph model;

[0109] Step 803: Determine the target information recommendation position matching the target agent;

[0110] The display position of the target information recommendation position in the information flow presentation page is determined according to at least one of the following:

[0111] The total browsing time of all information by the target user in this information browsing behavior (i.e., inserting an interaction request at the corresponding information browsing position when the total browsing time reaches a certain set time threshold), the browsing duration of any information by the user in this information browsing behavior (i.e., inserting an interaction request at a suitable information browsing position in the information when the browsing duration of a single information reaches a certain set time threshold), the total number of information browsed by the user in this information browsing behavior (i.e., inserting an interaction request at a suitable information browsing position in the information browsing interface when the total number of browsed information reaches a certain set number threshold), the preset position for presenting the waterfall information flow (i.e., inserting an interaction request at some preset dynamic insertion position in the waterfall information flow), the end position of the information body of the browsed information, etc. Only a limited number of possible ways can be listed here, but they are not limited to the above-listed cases.

[0112] Step 804: Send the target intelligent agent to the target user in a manner of displaying it in the target information recommendation position.

[0113] This embodiment aims at the situation where the target user does not actively initiate a search behavior in the current session. Since there is no demand information actively expressed by the user, it is only possible to obtain user preference information from the target user's historical search sessions and / or historical recommendation sessions, and input the user preference information into the generative heterogeneous graph model as the specific user-related information in this case, in order for the generative heterogeneous graph model to match a suitable target intelligent agent for the target user according to the user preference information.

[0114] Based on any of the above embodiments, feedback information of the target user on the distributed target intelligent agent can also be obtained, and then a new interaction action chain can be generated based on the target user and the feedback information on the distributed target intelligent agent, so as to update the generative heterogeneous graph model by using the new interaction action chain.

[0115] To deepen understanding, this embodiment starts from the following difficulties in task completion in open fields and attempts to provide a search-and-push integrated solution based on a generative heterogeneous graph model:

[0116] 1) How to efficiently match agents and user needs in the capability space;

[0117] 2) How to use the same set of algorithms to efficiently model different scenarios;

[0118] 3) How to solve the distribution problem end-to-end and reduce multi-layer funnel losses.

[0119] Specifically, the proposed agent search and recommendation integrated distribution algorithm based on the generative graph model can be applied to various scenarios where agent distribution can be performed, such as search, recommendation, and content downloading. The solution realizes the precise distribution of agents and user needs based on the generative heterogeneous graph model. The solution is designed to have the following functions:

[0120] 1) Constructing generative graph models based on large-scale user feedback and agent interaction data;

[0121] 2) Using the user’s input and status as input, it performs walk-through recommendation on the graph model and generates distributed agent candidates;

[0122] 3) In different distribution scenarios, generative sorting modeling is performed based on the unified representation of the agent and user status in the graph model, the model is estimated from the perspective of the ability matching of the agent and the demand, and the agents with good output satisfaction are distributed.

[0123] It is expected to be applied in intelligent agent distribution systems in various scenarios such as search and information flow.

[0124] The implementation framework of this solution is as follows Figure 9-1 As shown, the input of the system includes: the intelligent agent distribution user behavior feedback logs in multiple scenarios (including search scenarios, recommendation scenarios, various content mounting scenarios, etc.), and the output is the intelligent agent distribution result in a given scenario.

[0125] The entire distribution system is based on a unified heterogeneous graph model and mainly consists of several parts:

[0126] 1) Offline trained agent unified heterogeneous graph model (Agent Heterogeneous Graph): to understand, decompose and connect tasks as a whole and give the final result;

[0127] 2) Online generative ranking: recall and ranking distribution based on graph models according to different scenario inputs.

[0128] Figure 9-2 It shows that in Figure 9-1 Under the provided implementation framework, a specific example of how to construct a generative heterogeneous graph model by using multiple interactive action chains related to poetry and poets (such as Li Bai) by multiple users is given, and the integrated graph structure is also shown.

[0129] The technical effects that can be achieved through the above-mentioned solution include:

[0130] 1. Unified graph modeling for multi-scenario data fusion

[0131] For current search engines that can perform information retrieval by matching requests with relevance, and recommendation systems that can perform content recommendations in non-query scenarios, the solution provided in this embodiment unifies models in scenarios such as search and recommendation mounting to form a unified intelligent graph model with multiple scenarios and multiple heterogeneous nodes to serve the online algorithm distribution process. The unified graph model can greatly integrate data from different scenarios, obtain more connection information, and further deeply explore the relationship between different nodes to achieve a more efficient characterization of nodes and edges.

[0132] 2. Online unified system distribution for different scenarios based on graph model + generative model

[0133] Based on the learning results of the unified graph model for multiple scenarios, the end-to-end generative model is constructed online for different scenarios using the graph model node representation as input, which can distribute different scenarios through a unified system, thereby greatly improving the efficiency of the system and the effectiveness of the method. The integration of search and recommendation is realized in the online system.

[0134] 3. Generate models to match satisfaction in capability space

[0135] The current mainstream search and recommendation algorithm system uses a semantic-level matching and distribution model for traditional content. This solution is more suitable for deterministic content such as videos, pictures and texts, but the intelligent agent is a digital person, and the description and setting of the intelligent agent cannot fully represent the capabilities of the intelligent agent. Intelligent agent distribution needs to model the demand matching degree from the perspective of capability matching. This embodiment uses a generative model to construct a large model for estimating the satisfaction of the capability matching space, achieves an in-depth understanding and matching modeling of user needs and intelligent agent capabilities, and conducts reinforcement learning alignment training through online user feedback to continuously improve the distribution effect of each scenario.

[0136] Further references Fig.10 and Fig.11 As an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of a device for constructing a generative heterogeneous graph model and an embodiment of an agent distribution device. The embodiment of the device for constructing a generative heterogeneous graph model and the embodiment of the agent distribution device are Figure 2 Corresponding to the embodiment of the method for constructing the generative heterogeneous graph model shown in FIG. Figure 6 The above device can be applied to various electronic devices.

[0137] like Fig.10As shown, the construction device 1000 of the generative heterogeneous graph model of this embodiment may include: a first interactive action chain extraction unit 1001, a second interactive action chain extraction unit 1002, and a generative heterogeneous graph model construction unit 1003. Among them, the first interactive action chain extraction unit 1001 is configured to extract a first interactive action chain containing an agent distribution behavior exceeding a preset satisfaction level from the historical search behavior and search behavior feedback of the first sample user; the second interactive action chain extraction unit 1002 is configured to extract a second interactive action chain containing an agent distribution behavior exceeding a preset satisfaction level from the historical recommended behavior and recommended behavior feedback of the second sample user; the generative heterogeneous graph model construction unit 1003 is configured to construct a generative heterogeneous graph model based on the first interactive action chain and the second interactive action chain, with users, behavior objects, and agents as different nodes and interactive actions as edges connecting corresponding nodes.

[0138] In this embodiment, in the apparatus 1000 for constructing a generative heterogeneous graph model, the specific processing of the first interactive action chain extraction unit 1001, the second interactive action chain extraction unit 1002, and the generative heterogeneous graph model construction unit 1003 and the technical effects thereof can be referred to respectively. Figure 2 The relevant descriptions of steps 201 - 203 in the corresponding embodiment are not repeated here.

[0139] In some optional implementations of this embodiment, the first interaction action chain extraction unit 1001 is further configured to:

[0140] Obtaining historical search behaviors and search behavior feedback of the first sample user;

[0141] From the historical search behaviors and search behavior feedback, filter the search interaction action chain containing the distribution behavior of any agent to obtain the target search interaction action chain;

[0142] Determine the actual satisfaction of the first sample user with respect to the distributed intelligent agent according to the subsequent interactive action of the first sample user with respect to the distributed intelligent agent in the target search interactive action chain;

[0143] The target search interaction action chain having an actual satisfaction level exceeding a preset satisfaction level is determined as the first interaction action chain.

[0144] In some optional implementations of this embodiment, the subsequent interaction action for determining the actual satisfaction level includes at least one of the following:

[0145] The duration of the first use of the distributed agent;

[0146] The total usage time, total usage times or average usage time of the distributed intelligent agents within a preset period of time;

[0147] Whether the distributed agent has been collected or forwarded;

[0148] The number of forwarding operations or forwarding channels for the distributed agents;

[0149] Whether the distributed agent has been evaluated and the sentiment of the evaluation content.

[0150] In some optional implementations of this embodiment, the second interaction action chain extraction unit 1002 is further configured to:

[0151] Obtain the historical recommended behaviors and recommended behavior feedback of the second sample user;

[0152] From the historical recommended behaviors and recommended behavior feedback, filter the recommended interactive action chain containing the distribution behavior of any agent to obtain the target recommended interactive action chain;

[0153] Determine the actual satisfaction of the second sample user with respect to the distributed intelligent agent according to the subsequent interactive action of the second sample user with respect to the distributed intelligent agent in the target recommendation interactive action chain;

[0154] The target recommended interactive action chain with an actual satisfaction level exceeding the preset satisfaction level is determined as the second interactive action chain.

[0155] In some optional implementations of this embodiment, the subsequent interaction action for determining the actual satisfaction level includes at least one of the following:

[0156] The viewing time of the information recommendation position where the distributed agent is located;

[0157] Whether the distributed agent is run;

[0158] The duration of the first run of the distributed agent and the time interval between the second run and the first run;

[0159] Whether a search operation related to the function of the distributed agent is actively performed after the distributed agent is run.

[0160] In some optional implementations of this embodiment, the behavior object used as a node may include at least one of the following:

[0161] The search words / search statements input by the user, the selected target search results, the recommended search words / search statements, and the information or objects recommended to be presented in the information recommendation position.

[0162] In some optional implementations of this embodiment, the generative heterogeneous graph model building unit 1003 is further configured to:

[0163] Get the generative graph model to be trained;

[0164] Extracting user entities, behavior object entities, agent entities, and interaction actions between different entities as edges connecting different nodes from the first interaction action chain and the second interaction action chain;

[0165] The generative graph model to be trained is trained by using the user entities, behavior object entities, intelligent agent entities as different nodes and the interaction actions between different entities as edges connecting different nodes to obtain a generative heterogeneous graph model.

[0166] like Fig.11 As shown, the agent distribution device 1100 of this embodiment may include: a user-related information acquisition unit 1101, a model use unit 1102, and an agent distribution unit 1103. The user-related information acquisition unit 1101 is configured to acquire user-related information of a target user; the model use unit 1102 is configured to input user-related information into a pre-trained generative heterogeneous graph model to obtain a target agent output by the generative heterogeneous graph model; the generative heterogeneous graph model is obtained according to the generative heterogeneous graph model construction device 1000; the agent distribution unit 1103 is configured to distribute the target agent to the target user.

[0167] In this embodiment, the specific processing of the user related information acquisition unit 1101, the model use unit 1102, and the agent distribution unit 1103 and the technical effects thereof can be referred to in Figure 6 The relevant descriptions of steps 601-603 in the corresponding embodiment are not repeated here.

[0168] In some optional implementations of this embodiment, the user related information acquisition unit 1101 is further configured to:

[0169] In response to the target user actively initiating a search behavior in the current session, a search demand is obtained from the search behavior actively initiated by the target user; wherein the user-related information includes the search demand.

[0170] In some optional implementations of this embodiment, the agent distribution unit 1103 is further configured to:

[0171] Distribute the target agent as the first candidate search result to the target user.

[0172] In some optional implementations of this embodiment, the user related information acquisition unit 1101 is further configured to:

[0173] In response to the target user not actively initiating a search behavior in the current session, user preference information is obtained from the target user's historical search session and / or historical recommendation session; wherein the user-related information includes the user preference information.

[0174] In some optional implementations of this embodiment, the agent distribution unit 1103 is further configured to:

[0175] Determine the target information recommendation position matching the target intelligent agent;

[0176] The target intelligent agent is sent to the target user in the form of being displayed in the target information recommendation position.

[0177] In some optional implementations of this embodiment, the display position of the target information recommendation position in the information flow presentation page is determined according to at least one of the following:

[0178] The total time the target user spends browsing all information during this browsing behavior, the duration of the user's browsing of any information during this browsing behavior, the total number of pieces of information browsed by the user during this browsing behavior, the preset position for presenting the waterfall-style information flow, and the end position of the body of the information being browsed.

[0179] In some optional implementations of this embodiment, the agent distribution device 1100 further includes:

[0180] A feedback information acquisition unit, configured to acquire feedback information of a target user on the distributed target agent;

[0181] The new interaction action chain generation unit is configured to generate a new interaction action chain according to the target user and the feedback information of the distributed target intelligent agent, so as to update the generative heterogeneous graph model by using the new interaction action chain.

[0182] The above-mentioned embodiment exists as an apparatus embodiment corresponding to the above-mentioned method embodiment. The apparatus for constructing a generative heterogeneous graph model provided by the above-mentioned embodiment extracts a first interactive action chain including agent distribution behaviors that are considered satisfactory by users from historical search behaviors and search behavior feedback in search scenarios, and extracts a second interactive action chain including agent distribution behaviors that are considered satisfactory by users from historical recommended behaviors and recommendation behavior feedback in recommendation scenarios. Furthermore, different nodes played by users, behavior objects and agents and edges played by interactive actions can be extracted from each interactive action chain covering the search scenario and the recommendation scenario, and finally a generative heterogeneous graph model can be constructed based on these nodes and edges connecting different nodes, so that the constructed generative heterogeneous graph model integrates the association relationship between different users and different agents in the search scenario and the recommendation scenario in the form of a graph structure.

[0183] The agent distribution device provided in the above embodiment can, after obtaining user-related information of the target user, directly input the user-related information into the generative heterogeneous graph model constructed in the above manner, and quickly obtain the target agent that matches the target user without distinguishing the scene type. That is, the generative heterogeneous graph model can be used to distribute matching agents to users in all scenarios.

[0184] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can implement the generative heterogeneous graph model construction method and / or the intelligent agent distribution method described in any of the above embodiments when executing.

[0185] According to an embodiment of the present disclosure, the present disclosure also provides a readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to implement the method for constructing a generative heterogeneous graph model and / or the method for distributing an intelligent agent described in any of the above embodiments when executed.

[0186] An embodiment of the present disclosure provides a computer program product, which, when executed by a processor, can implement the steps of the method for constructing a generative heterogeneous graph model and / or the steps of the agent distribution method described in any of the above embodiments.

[0187] Fig.12 A schematic block diagram of an example electronic device 1200 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0188] like Fig.12As shown, the device 1200 includes a computing unit 1201, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1202 or a computer program loaded from a storage unit 1208 into a random access memory (RAM) 1203. In the RAM 1203, various programs and data required for the operation of the device 1200 can also be stored. The computing unit 1201, the ROM 1202, and the RAM 1203 are connected to each other via a bus 1204. An input / output (I / O) interface 1205 is also connected to the bus 1204.

[0189] A number of components in the device 1200 are connected to the I / O interface 1205, including: an input unit 1206, such as a keyboard, a mouse, etc.; an output unit 1207, such as various types of displays, speakers, etc.; a storage unit 1208, such as a disk, an optical disk, etc.; and a communication unit 1209, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1209 allows the device 1200 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0190] The computing unit 1201 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 1201 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 1201 performs the various methods and processes described above, such as the construction method of the generative heterogeneous graph model and / or the agent distribution method. For example, in some embodiments, the construction method of the generative heterogeneous graph model and / or the agent distribution method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 1208. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 1200 via the ROM 1202 and / or the communication unit 1209. When the computer program is loaded into RAM 1203 and executed by the computing unit 1201, one or more steps of the method for constructing a generative heterogeneous graph model and / or the method for distributing an agent described above may be performed. Alternatively, in other embodiments, the computing unit 1201 may be configured to execute the method for constructing a generative heterogeneous graph model and / or the method for distributing an agent by any other appropriate means (e.g., by means of firmware).

[0191] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0192] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0193] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0194] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types 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, voice input, or tactile input).

[0195] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0196] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and virtual private servers (VPS) services.

[0197] The technical solution of the disclosed embodiment extracts a first interactive action chain containing agent distribution behaviors that are considered satisfactory by users from historical search behaviors and search behavior feedback in search scenarios, and extracts a second interactive action chain containing agent distribution behaviors that are considered satisfactory by users from historical recommended behaviors and recommendation behavior feedback in recommendation scenarios, and then can extract different nodes played by users, behavior objects and agents and edges played by interactive actions from each interactive action chain covering search scenarios and recommendation scenarios, and finally construct a generative heterogeneous graph model based on these nodes and edges connecting different nodes, so that the constructed generative heterogeneous graph model integrates the association relationship between different users and different agents in search scenarios and recommendation scenarios in the form of a graph structure. And after obtaining the user-related information of the target user, the target agent that matches the target user can be quickly obtained without distinguishing the scene type by directly inputting the user-related information into the generative heterogeneous graph model constructed in the above manner, that is, the matching agent can be distributed to the user in all scenarios by using the generative heterogeneous graph model.

[0198] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0199] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for constructing a generative heterogeneous graph model, comprising: Extracting a first interactive action chain including an agent distribution behavior exceeding a preset satisfaction level from the historical search behavior and search behavior feedback of the first sample user; Extracting a second interactive action chain including an agent distribution behavior exceeding the preset satisfaction level from the historical recommended behaviors and recommended behavior feedback of the second sample user; Based on the first interactive action chain and the second interactive action chain, a generative heterogeneous graph model is constructed by taking users, behavior objects, and intelligent agents as different nodes and taking interactive actions as edges connecting corresponding nodes.

2. The method according to claim 1, wherein: The step of extracting a first interactive action chain including an agent distribution behavior exceeding a preset satisfaction level from the historical search behavior and search behavior feedback of the first sample user includes: Obtaining historical search behaviors and search behavior feedback of the first sample user; Filtering the search interaction action chain including the distribution behavior of any agent from the historical search behavior and the search behavior feedback to obtain the target search interaction action chain; Determining actual satisfaction of the first sample user with respect to the distributed agent according to a subsequent interaction action of the first sample user with respect to the distributed agent in the target search interaction action chain; The target search interaction action chain having an actual satisfaction level exceeding the preset satisfaction level is determined as the first interaction action chain.

3. The method according to claim 2, wherein: The subsequent interactive action for determining the actual satisfaction level includes at least one of the following: The duration of the first use of the distributed agent; The total usage time, total usage times or average usage time of the distributed intelligent agents within a preset period of time; Whether the distributed agent has been collected or forwarded; The number of forwarding operations or forwarding channels for the distributed agents; Whether the distributed agent has been evaluated and the sentiment of the evaluation content.

4. The method according to claim 1, wherein: The step of extracting a second interactive action chain including an agent distribution behavior exceeding the preset satisfaction level from the historical recommended behaviors and recommended behavior feedback of the second sample user includes: Obtaining historical recommended behaviors and recommended behavior feedback of the second sample user; From the historical recommended behaviors and the recommended behavior feedback, filter the recommended interactive action chain containing the distribution behavior of any agent to obtain the target recommended interactive action chain; Determining actual satisfaction of the second sample user with respect to the distributed agent according to a subsequent interaction action of the second sample user with respect to the distributed agent in the target recommendation interaction action chain; The target recommended interactive action chain having an actual satisfaction level exceeding the preset satisfaction level is determined as the second interactive action chain.

5. The method according to claim 4, wherein: The subsequent interactive action for determining the actual satisfaction level includes at least one of the following: The viewing time of the information recommendation position where the distributed agent is located; Whether the distributed agent is run; The duration of the first run of the distributed agent and the time interval between the second run and the first run; Whether a search operation related to the function of the distributed agent is actively performed after the distributed agent is run.

6. The method according to claim 1, wherein: The behavior object used as the node may include at least one of the following: The search words / search statements input by the user, the selected target search results, the recommended search words / search statements, and the information or objects recommended to be presented in the information recommendation position.

7. The method according to any one of claims 1 to 6, wherein: The method of constructing a generative heterogeneous graph model based on the first interactive action chain and the second interactive action chain by taking users, behavior objects, and intelligent agents as different nodes and taking interactive actions as edges connecting corresponding nodes includes: Get the generative graph model to be trained; Extracting user entities, behavior object entities, agent entities, and interaction actions between different entities as edges connecting different nodes from the first interaction action chain and the second interaction action chain; The generative graph model to be trained is trained by utilizing the interaction actions between user entities, behavior object entities, intelligent agent entities as different nodes and different entities as edges connecting different nodes to obtain the generative heterogeneous graph model.

8. An agent distribution method, comprising: Obtain user-related information of the target user; Inputting the user-related information into a pre-trained generative heterogeneous graph model to obtain a target agent output by the generative heterogeneous graph model; wherein the generative heterogeneous graph model is obtained according to the method for constructing a generative heterogeneous graph model according to any one of claims 1 to 7; Distribute the target agent to the target user.

9. The method according to claim 8, wherein: The obtaining of user-related information of the target user includes: In response to the target user actively initiating a search behavior in the current session, a search demand is obtained from the search behavior actively initiated by the target user; wherein the user-related information includes the search demand.

10. The method according to claim 9, wherein: The step of distributing the target agent to the target user comprises: The target agent is distributed to the target user as the first candidate search result.

11. The method according to claim 8, wherein: The obtaining of user-related information of the target user includes: In response to the target user not actively initiating a search behavior in the current session, user preference information is obtained from the target user's historical search session and / or historical recommendation session; wherein the user-related information includes the user preference information.

12. The method according to claim 11, wherein: The step of distributing the target agent to the target user comprises: Determine a target information recommendation position matching the target intelligent agent; The target intelligent agent is sent to the target user in a manner of being displayed in the target information recommendation position.

13. The method according to claim 12, wherein: The display position of the target information recommendation position in the information flow presentation page is determined according to at least one of the following: The total time the target user spends browsing all information during this information browsing behavior, the duration of the user's browsing of any information during this information browsing behavior, the total number of pieces of information browsed by the user during this information browsing behavior, the preset position for presenting the waterfall-style information flow, and the end position of the information body of the browsed information.

14. The method according to any one of claims 8 to 13, further comprising: Obtaining feedback information from the target user on the distributed target agent; A new interaction action chain is generated according to the target user and feedback information to the distributed target agent, so as to update the generative heterogeneous graph model by using the new interaction action chain.

15. A device for constructing a generative heterogeneous graph model, comprising: A first interactive action chain extraction unit is configured to extract a first interactive action chain including an agent distribution behavior exceeding a preset satisfaction level from the historical search behavior and search behavior feedback of the first sample user; A second interactive action chain extraction unit is configured to extract a second interactive action chain including an agent distribution behavior exceeding the preset satisfaction level from the historical recommended behaviors and recommended behavior feedback of the second sample user; The generative heterogeneous graph model construction unit is configured to construct a generative heterogeneous graph model based on the first interaction action chain and the second interaction action chain, with users, behavior objects, and intelligent agents as different nodes and interaction actions as edges connecting corresponding nodes.

16. The device according to claim 15, wherein: The first interactive action chain extraction unit is further configured to: Obtaining historical search behaviors and search behavior feedback of the first sample user; Filtering the search interaction action chain including the distribution behavior of any agent from the historical search behavior and the search behavior feedback to obtain the target search interaction action chain; Determining actual satisfaction of the first sample user with respect to the distributed agent according to a subsequent interaction action of the first sample user with respect to the distributed agent in the target search interaction action chain; The target search interaction action chain having an actual satisfaction level exceeding the preset satisfaction level is determined as the first interaction action chain.

17. The device according to claim 16, wherein: The subsequent interactive action for determining the actual satisfaction level includes at least one of the following: The duration of the first use of the distributed agent; The total usage time, total usage times or average usage time of the distributed intelligent agents within a preset period of time; Whether the distributed agent has been collected or forwarded; The number of forwarding operations or forwarding channels for the distributed agents; Whether the distributed agent has been evaluated and the sentiment of the evaluation content.

18. The device according to claim 15, wherein: The second interactive action chain extraction unit is further configured to: Obtaining historical recommended behaviors and recommended behavior feedback of the second sample user; From the historical recommended behaviors and the recommended behavior feedback, filter the recommended interactive action chain containing the distribution behavior of any agent to obtain the target recommended interactive action chain; Determining actual satisfaction of the second sample user with respect to the distributed agent according to a subsequent interaction action of the second sample user with respect to the distributed agent in the target recommendation interaction action chain; The target recommended interactive action chain having an actual satisfaction level exceeding the preset satisfaction level is determined as the second interactive action chain.

19. The device according to claim 18, wherein: The subsequent interactive action for determining the actual satisfaction level includes at least one of the following: The viewing time of the information recommendation position where the distributed agent is located; Whether the distributed agent is run; The duration of the first run of the distributed agent and the time interval between the second run and the first run; Whether a search operation related to the function of the distributed agent is actively performed after the distributed agent is run.

20. The device according to claim 15, wherein: The behavior object used as the node may include at least one of the following: The search words / search statements input by the user, the selected target search results, the recommended search words / search statements, and the information or objects recommended to be presented in the information recommendation position.

21. The device according to any one of claims 15 to 20, wherein: The generative heterogeneous graph model building unit is further configured to: Get the generative graph model to be trained; Extracting user entities, behavior object entities, agent entities, and interaction actions between different entities as edges connecting different nodes from the first interaction action chain and the second interaction action chain; The generative graph model to be trained is trained by utilizing the interaction actions between user entities, behavior object entities, intelligent agent entities as different nodes and different entities as edges connecting different nodes to obtain the generative heterogeneous graph model.

22. An intelligent agent distribution device, comprising: A user related information acquisition unit, configured to acquire user related information of a target user; A model using unit, configured to input the user-related information into a pre-trained generative heterogeneous graph model to obtain a target agent output by the generative heterogeneous graph model; wherein the generative heterogeneous graph model is obtained according to the construction device of the generative heterogeneous graph model according to any one of claims 15 to 21; The agent distribution unit is configured to distribute the target agent to the target user.

23. The device according to claim 22, wherein: The user related information acquisition unit is further configured to: In response to the target user actively initiating a search behavior in the current session, a search demand is obtained from the search behavior actively initiated by the target user; wherein the user-related information includes the search demand.

24. The device according to claim 23, wherein: The agent distribution unit is further configured to: The target agent is distributed to the target user as the first candidate search result.

25. The device according to claim 22, wherein: The user related information acquisition unit is further configured to: In response to the target user not actively initiating a search behavior in the current session, user preference information is obtained from the target user's historical search session and / or historical recommendation session; wherein the user-related information includes the user preference information.

26. The device according to claim 25, wherein: The agent distribution unit is further configured to: Determine a target information recommendation position matching the target intelligent agent; The target intelligent agent is sent to the target user in a manner of being displayed in the target information recommendation position.

27. The device according to claim 26, wherein: The display position of the target information recommendation position in the information flow presentation page is determined according to at least one of the following: The total time the target user spends browsing all information during this information browsing behavior, the duration of the user's browsing of any information during this information browsing behavior, the total number of pieces of information browsed by the user during this information browsing behavior, the preset position for presenting the waterfall-style information flow, and the end position of the information body of the browsed information.

28. The apparatus according to any one of claims 22 to 27, further comprising: A feedback information acquisition unit, configured to acquire feedback information of the target user on the distributed target agent; The new interaction action chain generating unit is configured to generate a new interaction action chain according to the target user and the feedback information of the distributed target intelligent agent, so as to update the generative heterogeneous graph model by using the new interaction action chain.

29. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for constructing a generative heterogeneous graph model described in any one of claims 1-7 and / or the agent distribution method described in any one of claims 8-14.

30. A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the method for constructing a generative heterogeneous graph model as described in any one of claims 1-7 and / or the agent distribution method as described in any one of claims 8-14.

31. A computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the method for constructing a generative heterogeneous graph model according to any one of claims 1 to 7 and / or the steps of the agent distribution method according to any one of claims 8 to 14.

Images