Recommendation reason generation method and device, storage medium and electronic equipment
By building knowledge graphs and metapaths, the problem of lack of data and heterogeneous dispersion in the game field is solved, and the efficient generation of personalized recommendation reasons is achieved, meeting the real-time performance requirements of game products.
Patent Information
- Application Number
- CN202411307921.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-18
- Publication Date
- 2024-12-06
AI Technical Summary
The lack of data in the gaming field and heterogeneous dispersion makes it difficult to generate personalized recommendation reasons and cannot meet the real-time performance requirements of frequent resource iterations.
By obtaining game heterogeneous data, building a knowledge graph, obtaining the metapath collection of target user entities, and using the metapath to generate personalized recommendation reasons to improve the efficiency and real-time generation of recommendation reasons.
It reduces the difficulty of generating recommendation reasons, improves efficiency, and meets the high-performance needs of game products due to frequent resource iterations, and generates personalized and convincing recommendation reasons.
Smart Images

Figure CN119089048A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of game technology, and in particular to a method for generating a recommendation reason, a device for generating a recommendation reason, a computer storage medium, and an electronic device. Background Art
[0002] Personalized recommendation is an activity that recommends objects of interest to users based on their interest characteristics. It is based on massive data mining. By analyzing and mining user behavior, users' personalized needs and interest characteristics are discovered, and information or products that users may be interested in are recommended to users, thereby achieving the technical effect of increasing user stickiness. In order to further enhance the service capabilities of the personalized recommendation system and improve the persuasiveness and surprise of the recommendation results to players, personalized recommendation reasons can usually be generated.
[0003] In relevant technical solutions, the focus is usually on generating personalized recommendation reasons in the fields of food, goods, etc. There are obvious opinion tendencies in the user data in this field, so the recommendation reasons can be generated by calculating the list or extracting selected comments.
[0004] However, the above-mentioned related technical solutions, on the one hand, due to the essential differences between the gaming field and the above-mentioned food, commodity and other fields, the data used to express opinion preferences in the gaming data is scarce and heterogeneous and scattered, making it difficult to extract selected opinions directly as recommendation reasons when making personalized recommendations in the gaming field, resulting in greater difficulty in generating recommendation reasons for the game. On the other hand, game products have the characteristics of frequent resource iteration, which puts forward higher requirements for the real-time generation of recommendation reasons, and the generation method of the above-mentioned related technical solutions is difficult to meet the performance requirements. Summary of the invention
[0005] The present disclosure provides a method for generating a recommendation reason, a device for generating a recommendation reason, a computer storage medium and an electronic device, thereby reducing the difficulty of generating a recommendation reason due to the problem of the scarcity and heterogeneous dispersion of data used to express opinion preferences in game data, and improving the efficiency of generating the recommendation reason, thereby meeting the high real-time performance requirements of game products due to the frequent resource iteration characteristics.
[0006] In a first aspect, an embodiment of the present disclosure provides a method for generating a recommendation reason, the method comprising: obtaining game heterogeneous data corresponding to a game application; obtaining a meta-path set corresponding to a target user entity in the game application based on a knowledge graph constructed based on the game heterogeneous data, the meta-path set comprising at least one meta-path consisting of a head node, a tail node and / or an intermediate node, and a connection relationship between each node; generating a target recommendation reason corresponding to the recommendation result based on the meta-path set and the recommendation result for the target user entity.
[0007] In a second aspect, an embodiment of the present disclosure provides a device for generating recommendation reasons, the device comprising: a data acquisition module, used to acquire game heterogeneous data corresponding to the game application; a meta-path determination module, used to determine a meta-path set corresponding to a target user entity in the game application based on a knowledge graph constructed based on the game heterogeneous data, the meta-path set containing at least one meta-path consisting of a head node, a tail node and / or an intermediate node, and a connection relationship between each node; a recommendation reason generation module, used to generate a target recommendation reason corresponding to the recommendation result based on the meta-path set and the recommendation result for the target user entity.
[0008] In a third aspect, an embodiment of the present disclosure provides a computer-readable storage medium having a computer program stored thereon, which implements the above method for generating recommendation reasons when executed by a processor.
[0009] In a fourth aspect, an embodiment of the present disclosure provides an electronic device, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the above method for generating recommendation reasons by executing the executable instructions.
[0010] In a fifth aspect, an embodiment of the present disclosure provides a computer program product, including a computer program, which is executed by a processor to implement the above method for generating recommendation reasons.
[0011] The technical solution disclosed in this disclosure has the following beneficial effects:
[0012] The above-mentioned method for generating recommendation reasons obtains the game heterogeneous data corresponding to the game application; obtains the meta-path set corresponding to the target user entity in the game application based on the knowledge graph constructed by the game heterogeneous data, and the meta-path set contains at least one meta-path composed of the head node, the tail node and / or the intermediate node and the connection relationship between each node; generates the target recommendation reason corresponding to the recommendation result according to the meta-path set and the recommendation result for the target user entity. Since there will be a rich MetaPath on the knowledge graph. MetaPath fully records all the connection relationships and intermediate concepts between the two specific concepts of the head and the tail, it is easy to use the semantic information on the concepts and relationships in MetaPath to generate MetaPath descriptions, and the specific content of its MetaPath description varies not only due to the MetaPath structure, but also due to the specific data attributes that conform to MetaPath. It can quickly generate personalized recommendation reasons, thereby solving the problem of the difficulty of generating recommendation reasons in the prior art due to the lack of data used to express opinion preferences in game data and the problem of heterogeneous dispersion, and achieving the technical effect of reducing the difficulty of generating recommendation reasons. In addition, the meta-path can be used to quickly obtain the correlation between heterogeneous game data, thereby improving the efficiency of generating recommendation reasons, and further meeting the high real-time performance requirements of game products due to the frequent resource iteration characteristics.
[0013] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] The drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and are used together with the specification to explain the principles of the present disclosure. Obviously, the drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0015] Figure 1 A schematic diagram showing an architecture of a system for generating a recommendation reason in this exemplary embodiment;
[0016] Figure 2 A flowchart schematically illustrates a method for generating a recommendation reason in this exemplary embodiment;
[0017] Figure 3a A schematic diagram schematically shows a game heterogeneous data in this exemplary embodiment;
[0018] Figure 3b A schematic diagram of a knowledge graph in this exemplary embodiment is schematically shown;
[0019] Figure 4 A flowchart of a method for generating recommendation reasons through a meta-path in this exemplary embodiment is schematically shown;
[0020] Figure 5 A schematic diagram of water wave network diffusion in this exemplary embodiment is schematically shown;
[0021] Figure 6 A schematic diagram of a knowledge structure of implicit association information in a knowledge graph in this exemplary embodiment is schematically shown;
[0022] Figure 7 A schematic diagram of a RippleNet network structure in this exemplary embodiment is schematically shown;
[0023] Figure 8 A schematic diagram of a distributed recommendation reason service decoupling structure in this exemplary embodiment is shown schematically;
[0024] Fig. 9 A schematic diagram of the structure of a device for generating a recommendation reason in this exemplary embodiment is schematically shown;
[0025] Fig.10 The structure diagram of an electronic device in this exemplary embodiment is schematically shown. DETAILED DESCRIPTION
[0026] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, exemplary embodiments can be implemented in a variety of forms and should not be construed as being limited to the examples set forth herein; on the contrary, these embodiments are provided so that the present disclosure will be more comprehensive and complete, and the concepts of the exemplary embodiments are fully conveyed to those skilled in the art. The described features, structures, or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced while omitting one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known technical solutions are not shown or described in detail to avoid obscuring various aspects of the present disclosure.
[0027] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.
[0028] The flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the steps. For example, some steps may be decomposed, while some steps may be combined or partially combined, so the actual execution order may change according to the actual situation.
[0029] In the relevant technical background, personalized recommendation is an activity that recommends objects of interest to users based on their interest characteristics. It is based on massive data mining. By analyzing and mining user behaviors, users' personalized needs and interest characteristics are discovered, and information or products that users may be interested in are recommended to users, thereby achieving the technical effect of increasing user stickiness. In order to further enhance the service capabilities of the personalized recommendation system and improve the persuasiveness and surprise of the recommendation results to users, personalized recommendation reasons can usually be generated to increase the certainty of users accepting personalized recommendation results.
[0030] In the relevant technical scheme, after determining the items to be recommended, it is usually possible to simply generate the recommendation reasons corresponding to the items to be recommended by calculating the list or refining the comments. For example, CN116597944A discloses a diet and exercise recommendation method based on a knowledge graph, which determines the basic recipes based on the user portrait and the pre-constructed food and nutrient elements, and energy knowledge graph, and determines the basic recommended exercise based on the user portrait and the pre-constructed disease and nutrient elements, and exercise type map, to achieve a two-way connection between diet and exercise recommendations, and control the energy intake and consumption, thereby controlling the energy balance of the human body. According to the energy status and exercise capacity information required by individual users, energy goals and exercise goals are planned, and recommended recipes and recommended exercises are obtained by combining knowledge graphs and machine learning algorithms. Finally, the reasons for recommending the recommended recipes are determined based on the recommended recipes and the pre-constructed disease and nutrition knowledge graphs.
[0031] CN117172876A discloses a method for recommending items. The method obtains the user's interaction data in the current session after providing an answer to a target user; determines whether it is suitable to recommend an item based on the interaction data in the current session; when it is determined that it is suitable to recommend an item, performs path analysis using the current knowledge graph to obtain at least one candidate path from a node indicating the target user to a node indicating a candidate item and a corresponding score; then determines an item to be recommended from at least one candidate item based on the obtained score, and generates a recommendation reason corresponding to the item to be recommended based on associated recommendation factors.
[0032] It needs to be explained that the knowledge graph recorded in the above-mentioned related schemes is called knowledge domain visualization or knowledge domain mapping map in the library and information industry. It is a series of various graphs that show the development process and structural relationship of knowledge. It uses visualization technology to describe knowledge resources and their carriers, and to mine, analyze, construct, draw and display knowledge and their interrelationships.
[0033] It can be seen from the above-mentioned related technical solutions that the existing methods for generating recommendation reasons are mostly focused on the fields of food, film, commodities, short videos, etc. The data in this application field is usually data in which users actively express their opinions. For example, users usually show that they like a type of food (such as sweet food), dislike a type of food (such as bitter food), etc., or when users see short videos they don’t like, they express their opinions by selecting "Don’t like this type of video, reduce recommendations", or when users see short videos they like, they express their opinions by selecting "Like this type of video, increase recommendations" or determine the user's opinions by detecting the length of time they watch such videos. Therefore, when generating recommendation reasons corresponding to the items to be recommended, specific opinion tendency features can be extracted from the above data, and the existing solutions are limited to generating corresponding recommendation reasons after finding related items for personalized recommendations from the perspective of path analysis of the knowledge graph.
[0034] However, the above-mentioned related technical solutions are difficult to apply to game products in the game field. The reasons and existing technical problems are as follows:
[0035] 1) Data in the gaming field is scarce and heterogeneous and dispersed.
[0036] In application fields such as food and product recommendations that are the focus of relevant technical solutions, users usually actively express their opinions, so the data has the characteristics of showing more opinions. When generating reasons for recommendations, the corresponding reasons for recommendations can be obtained by calculating scores (weights) or refining comments. For example, users can clearly express their likes or dislikes for a certain product. However, in the field of games, players usually rarely actively express their opinions, which makes the opinion preference data of players in the game extremely scarce and heterogeneous (heterogeneous dispersion refers to data with different structures, formats, semantics and sources), which makes it difficult to simply generate reasons for recommendations by calculating lists or refining comments, which makes the generation process of recommendation reasons in the game field more complicated.
[0037] 2) Business diversity makes it difficult to mass-produce the generated recommendation reasons.
[0038] Since the forms and implementation methods of personalized recommendation services within games are diverse, if the recommendation reason generation method of the above-mentioned related technical solution is used, it is necessary to develop a separate recommendation reason generation solution for each business form, which is time-consuming and labor-intensive, affects the efficiency of recommendation reason generation, and is extremely costly.
[0039] 3) Failure to meet performance requirements.
[0040] Due to the frequent iteration of game product resources, higher requirements will be placed on the effectiveness of personalized recommendations, and thus higher requirements will be placed on the dynamics and real-time nature of recommendation reason generation. However, the offline recommendation reason generation method shown in the relevant technical solution is difficult to meet performance requirements.
[0041] 4) From the perspective of practical application, the access threshold is high.
[0042] As an innovative form of technology, every game that accesses recommendation reasons will face a series of underlying logical problems such as how to import data, how to generate recommendation reasons, and how to obtain recommendation reasons, which makes the access threshold of recommendation reasons high and difficult to generate recommendation reasons quickly.
[0043] In view of the above problems, the exemplary embodiment of the present disclosure proposes a method for generating a recommendation reason, which obtains heterogeneous game data corresponding to the game application; obtains a meta-path set corresponding to the target user entity in the game application based on the knowledge graph constructed by the heterogeneous game data, and the meta-path set contains at least one meta-path consisting of a head node, a tail node and / or an intermediate node and a connection relationship between each node; and generates a target recommendation reason corresponding to the recommendation result according to the meta-path set and the recommendation result for the target user entity. Since there will be a rich MetaPath on the knowledge graph. MetaPath fully records all the connection relationships and intermediate concepts between the two specific concepts of the head and the tail, it is easy to use the semantic information on the concepts and relationships in MetaPath to generate MetaPath descriptions, and the specific content of its MetaPath description varies not only due to the MetaPath structure, but also due to the specific data attributes that conform to MetaPath. It can quickly generate personalized recommendation reasons, thereby solving the difficulty of generating recommendation reasons due to the problem of the lack of data used to express opinion preferences in game data and heterogeneous dispersion in the prior art, and achieving the technical effect of reducing the difficulty of generating recommendation reasons. In addition, the meta-path can be used to quickly obtain the correlation between heterogeneous game data, thereby improving the efficiency of generating recommendation reasons, and further meeting the high real-time performance requirements of game products due to the frequent resource iteration characteristics.
[0044] The following will be combined Figure 1 , the system architecture of an exemplary application environment of a method for generating recommendation reasons and an apparatus for applying the same proposed in the present disclosure is described.
[0045] like Figure 1 The figure shows a schematic diagram of an application scenario of a method for generating a recommendation reason provided by an embodiment of the present disclosure. In the application scenario, a terminal device 101 and a server 102 may be included.
[0046] The terminal device 101 can be, for example, a mobile phone, a tablet computer (PAD), a laptop computer, a desktop computer, a smart TV, a smart car device, a smart wearable device, a smart TV, and an aircraft, and any other device involving personalized recommendations. The terminal device 101 can be installed with a target application, and the target application can have functions such as recommended items, presentation of recommended items, and text display of reasons for recommendation. For example, the target application can be an instant messaging application, a music application, a game application, a video application, a short video application, a news application, and a shopping application. The embodiments of the present disclosure are usually illustrated by taking a game application as an example. It should be noted that the application involved in the embodiments of the present application can be a software client, or a client such as a web page or a small program. The server 102 is a server corresponding to the software or a web page, a small program, etc., and does not limit the specific type of the client.
[0047] The server 102 may be a background server of the target application, used to provide corresponding background services for it, such as recommendation services and recommendation reason generation services, etc. It may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, i.e., content delivery networks (CDN), as well as big data and artificial intelligence platforms, but is not limited thereto.
[0048] It should be noted that the method for generating recommendation reasons in the embodiment of the present application can be executed by the terminal device 101 or the server 102 alone, or can be executed by the server 102 and the terminal device 101 together. The embodiment of the present disclosure does not impose any special limitation on this.
[0049] The server 102 and the terminal device 101 may each include one or more processors, memories, and interactive I / O interfaces, etc. The memories of the server 102 and the terminal device 101 may also store program instructions required to be executed in the method for generating recommendation reasons provided in the embodiment of the present disclosure, and these program instructions, when executed by the processor, can be used to implement the process for generating recommendation reasons provided in the embodiment of the present disclosure.
[0050] It should be noted that when the method for generating a recommendation reason provided in the embodiment of the present disclosure is executed by the server 102 or the terminal device 101 alone, the above-mentioned application scenario may also only include a single device of the server 102 or the terminal device 101, or the server 102 and the terminal device 101 may also be considered to be the same device. Of course, in actual application, when the method for generating a recommendation reason provided in the embodiment of the present disclosure is jointly executed by the server 102 and the terminal device 101, the server 102 and the terminal device 101 may also be the same device, that is, the server 102 and the terminal device 101 may be different functional modules of the same device, or virtual devices virtualized from the same physical device.
[0051] In the embodiment of the present disclosure, the terminal device 101 and the server 102 can be directly or indirectly connected to each other through one or more networks 103. The network 103 can be a wired network or a wireless network. For example, the wireless network can be a mobile cellular network or a wireless fidelity (Wireless-Fidelity, WIFI) network. Of course, it can also be other possible networks, which are not limited in the embodiment of the present disclosure. It should be noted that Figure 1The figure is only an example. In fact, the number of terminal devices and servers is not limited and is not specifically limited in the embodiments of the present disclosure.
[0052] Below, in combination with the application scenarios described above, the method provided by the exemplary embodiment of the present disclosure is described with reference to the accompanying drawings. It should be noted that the above application scenarios are only shown to facilitate understanding of the spirit and principle of the present disclosure, and the embodiments of the present disclosure are not limited in this respect. It should be noted that the following method can be executed by the above terminal device or server, or by the terminal device and the server together. Here, the execution of the terminal device or the server is specifically shown as an example.
[0053] See also Figure 2 , which is a flow chart of a method for generating a recommendation reason provided by an embodiment of the present disclosure, and which can be executed by a computer device / electronic device, and which can be Figure 1 In the terminal device 101 or the server 102 shown, the method for generating a recommendation reason provided by the embodiment of the present disclosure includes the following steps S201 to S203:
[0054] Step S201: Obtain heterogeneous game data corresponding to the game application.
[0055] Step S202: Determine a meta-path set corresponding to a target user entity in a game application based on a knowledge graph constructed based on heterogeneous game data. The meta-path set includes at least one meta-path consisting of a head node, a tail node and / or an intermediate node, and a connection relationship between each node.
[0056] Step S203: Generate a target recommendation reason corresponding to the recommendation result according to the meta-path set and the recommendation result for the target user entity.
[0057] In the technical solutions provided by some embodiments of the present disclosure, by obtaining the game heterogeneous data corresponding to the game application; obtaining the meta-path set corresponding to the target user entity in the game application based on the knowledge graph constructed by the game heterogeneous data, the meta-path set contains at least one meta-path consisting of the head node, the tail node and / or the intermediate node and the connection relationship between each node; according to the meta-path set and the recommendation result for the target user entity, the target recommendation reason corresponding to the recommendation result is generated. Since there will be a rich MetaPath on the knowledge graph. MetaPath fully records all the connection relationships and intermediate concepts between the two specific concepts of the head and the tail, it is easy to use the semantic information on the concepts and relationships in MetaPath to generate MetaPath descriptions, and the specific content of its MetaPath description varies not only due to the MetaPath structure, but also due to the specific data attributes that conform to MetaPath. It can quickly generate personalized recommendation reasons, thereby solving the problem of the difficulty of generating recommendation reasons in the prior art due to the lack of data used to express opinion preferences in game data and the problem of heterogeneous dispersion, and achieving the technical effect of reducing the difficulty of generating recommendation reasons. In addition, the meta-path can be used to quickly obtain the correlation between heterogeneous game data, thereby improving the efficiency of generating recommendation reasons, and further meeting the high real-time performance requirements of game products due to the frequent resource iteration characteristics.
[0058] The following will be combined with specific embodiments to Figure 2 The specific implementation methods of each step in the embodiment shown are described in detail:
[0059] In step S201, game heterogeneous data corresponding to the game application is obtained.
[0060] The game application can be any type of game, such as online games, stand-alone games, etc., and the embodiments of the present disclosure do not impose any special restrictions on this. The heterogeneous game data is scattered data, such as user basic data, gift package data, player relationship data, guild data, basic product data, etc.
[0061] In step S202, a meta-path set corresponding to a target user entity in a game application is determined based on a knowledge graph constructed based on heterogeneous game data, and the meta-path set includes at least one meta-path consisting of a head node, a tail node and / or an intermediate node, and a connection relationship between each node.
[0062] Among them, the meta-path set contains multiple meta-paths, which can be composed based on the connection relationship between each node in the knowledge graph, and the node can be a head node, a tail node, or a head node, a tail node and an intermediate node.
[0063] Among them, MetaPath is an abstract description of a specific structure in the knowledge graph. Any constructed knowledge graph will have rich MetaPath. MetaPath completely records all the connection relationships between the head node, the tail node and / or the intermediate nodes, and each node. Therefore, a MetaPath description can be directly generated based on the semantic information of all the connection relationships between each node and each node in MetaPath. The specific content of the MetaPath description varies not only due to the MetaPath structure, but also due to the specific data attributes that conform to MetaPath. Therefore, corresponding personalized recommendation reasons can be generated based on the metapath.
[0064] Exemplarily, after obtaining the game heterogeneous data corresponding to the game application, the meta-path set corresponding to the target user entity in the game application can be determined based on the knowledge graph constructed based on the game heterogeneous data. Prior to this, the knowledge graph needs to be pre-built.
[0065] In an optional embodiment of the present disclosure, each data entity in the game heterogeneous data is determined; each data entity in the game heterogeneous data is used as a node of the knowledge graph, and the relationship between each data entity is used as an edge of the knowledge graph, and a knowledge graph is constructed based on a preset knowledge graph structure; based on the knowledge graph constructed based on the game heterogeneous data, a meta-path set is obtained.
[0066] Exemplarily, when constructing a knowledge graph, it needs to be triggered from two aspects: the first aspect refers to the determination of the knowledge graph structure; the second aspect refers to the writing of data into the knowledge graph.
[0067] When determining the knowledge graph structure for the first aspect, the preset knowledge graph structure can be constructed when constructing the knowledge graph, or it can be a knowledge graph structure constructed using historical moments. Among them, the knowledge graph structure contains entities (also called concepts) and the relationship between entities, that is, entities (also called concepts) and the relationship between entities constitute the knowledge graph structure. In game applications, entities can be players (also called users), commodities, etc.
[0068] Regarding the data writing of the knowledge graph in the second aspect, the entities contained in the heterogeneous game data corresponding to the game application obtained in step S201 and the association relationships between the entities can be imported into the knowledge graph.
[0069] In other words, the above two aspects realize the use of each data entity in the game heterogeneous data as a node of the knowledge graph, and the relationship between each data entity as the edge of the knowledge graph, and construct a knowledge graph based on the preset knowledge graph structure. Figure 3a as well as Figure 3b An exemplary explanation of the constructed knowledge graph structure is given.
[0070] Figure 3a A schematic diagram schematically shows a game heterogeneous data in this exemplary embodiment, referring to Figure 3a Game applications contain a large amount of data, such as user basic data, gift package data, player relationship data, guild data, basic product data, etc. However, they have heterogeneous and scattered technical problems and do not contain the opinion tendency data actively expressed by players. This will make it difficult to extract and select opinions as recommendation reasons in game applications.
[0071] However, due to the complexity of the game system, there is a wealth of implicit information between various heterogeneous data such as between players, between players and products, and between products in the game application. These implicit information are interconnected through a complex network structure, and indirectly contain a large number of opinions that players have not actively expressed. As long as these implicit information are integrated and mined through a reasonable model structure, we can produce opinion arguments from them and form credible reasons for recommendations.
[0072] The knowledge graph is the best tool for fitting complex network structures and integrating and mining related implicit information. Therefore, we introduce the knowledge graph as the underlying model structure of this solution, using entities such as players and products as nodes in the knowledge graph, and the rich relationships between entities as edges in the knowledge graph. This fundamentally connects the heterogeneous data scattered throughout the game, associates and strengthens the implicit information in the data, and solves the problem of data scarcity in generating personalized recommendation reasons in the game.
[0073] In step S203, a target recommendation reason corresponding to the recommendation result is generated according to the meta-path set and the recommendation result for the target user entity.
[0074] The recommendation reason corresponding to the recommendation result may be a recommendation reason text. Of course, it may also be a recommendation reason presented in other forms, such as voice. The embodiments of the present disclosure do not impose any special restrictions on this.
[0075] Figure 4 A flowchart schematically shows a method for generating a recommendation reason based on a meta-path in this exemplary embodiment, see Figure 4 In an optional embodiment of the present disclosure, the above step S202 of generating a target recommendation reason corresponding to the recommendation result according to the meta-path set and the recommendation result for the target user entity may include the following steps S401 to S407:
[0076] Step S401: Obtain a target user entity and a recommendation result corresponding to the target user entity as target path information corresponding to a head node and a tail node respectively.
[0077] Step S402: Determine whether the target path information satisfies the meta-path structure.
[0078] If the target path information satisfies the meta-path structure, step S403 is executed to obtain the target user entity, the recommendation result corresponding to the target user entity, and multiple attribute information of the target path information.
[0079] The attribute information may include attribute information of each user entity.
[0080] For example, the connection between user entity A and user entity B is "friends". Assuming that the connection carries the attribute information "days" and the attribute value is 100, we can consider that the "friends" connection between user entity A and user entity B carries the attribute information "you have been friends for 100 days". When generating a recommendation reason later, you can use a recommendation reason description like "you have been friends for more than 100 days".
[0081] For another example, the entities in the path can also carry attribute information. The connection relationship between user entity A and user entity B is "friend". User entity A has an attribute "location" with an attribute value of City A. User entity B also has an attribute "location" with an attribute value of City A. Since both user entity A and user entity B carry the attribute information "location is City A", when subsequently generating recommendation reasons, descriptions of recommendation reasons such as "You are both in City A" can be used.
[0082] Then, step S404 is executed to generate a first target recommendation reason for the target user entity and the recommendation result corresponding to the target user entity according to the multiple attribute information and the pre-configured generative text rule template.
[0083] In another optional embodiment of the present disclosure, if the target path information does not satisfy the meta-path structure, step S405 is executed to determine whether the target path information satisfies other meta-path structures.
[0084] If so, execute step S406 to generate a second target recommendation reason.
[0085] Exemplarily, when generating the second target recommendation reason, the second target recommendation reason for the target user entity and the recommendation result corresponding to the target user entity may also be generated according to multiple attribute information and a pre-configured generative text rule template.
[0086] On the contrary, if it is not satisfied, continue to search for other meta-path structures, or execute step S407 to generate a pre-configured third target recommendation reason.
[0087] In an optional embodiment of the present disclosure, when executing the above step S402 and determining whether the target path information satisfies the meta-path structure, the meta-path structure can be converted into a query statement template corresponding to the meta-path database; the target user entity and the recommendation result corresponding to the target user entity are added to the query statement template to obtain a target detection statement; and based on the detection result of the target detection statement, it is determined whether the target path information satisfies the meta-path structure.
[0088] The meta-path database contains at least one query statement template corresponding to the meta-path.
[0089] Exemplarily, the MetaPath structure is translated into a query statement template of the database, and the information related to the head and tail nodes in the template is left blank to be filled in; then, after receiving the specific recommended data (u, v), u and v are mapped to specific nodes in the knowledge graph, and the node information is added to the MetaPath query statement template to form a detection statement. According to the result returned by the detection statement, it can be judged whether u and v meet the MetaPath structure. If the recommended data (u, v) meets the MetaPath structure, the relevant attributes of u and v and the relevant attributes of each node and edge in the MetaPath path are directly read, and the attribute values are brought into the generative text rule template to generate personalized recommendation reasons.
[0090] Based on the above steps, the structure of the knowledge graph can usually be represented as KG = (C, R), where C = {c1, c2, ..., cn} is the concept set (entity set) in the knowledge graph, and R = {r1, r2, ..., rn} is the relationship set connecting two concepts in the knowledge graph. The concept set and the relationship set together constitute the concept layer of the knowledge graph. The metapath MetaPath = {c1, r1, c2, r2, ..., rn, cn}, as a substructure in the concept layer of the knowledge graph, alternately contains concepts and relationships, representing several jump connection relationships from one concept to another.
[0091] For example, for a target user entity u and a recommendation result v for player u, the MetaPath technology is used to generate a recommendation reason for the recommendation result pair (u, v) and combined with Figure 4 Steps: Check whether the metapath with u and v as the head node and tail node respectively satisfies the MetaPath structure. If so, extract u, v and specific attributes in the MetaPath path, bring them into the generative text rule template, and generate highly personalized recommendation reasons for u and v; if not, continue to check whether the path with u and v as the head and tail nodes respectively satisfies other MetaPath structures to obtain the second target recommendation reason, or finally give a pre-configured third target recommendation reason.
[0092] The recommendation reasons generated by the meta-path mentioned above can break the technical barriers between players and recommendation results, and thus produce highly convincing recommendation reason text.
[0093] On the basis of any of the above embodiments, in order to further improve the universality of the generated recommendation reasons and to break the technical problem that the related technical solutions need to develop a separate recommendation reason generation solution for each business form, which is time-consuming and labor-intensive, affects the efficiency of recommendation reason generation, and is extremely costly, the present application can also generate personalized recommendation reasons based on the RippleNet network.
[0094] If the preferences expressed by players are compared to a water drop, then the entire knowledge graph is like a pool. The ripples caused by each water drop in the pool intuitively express the reasons for the interest preference. By systematically performing the action of stirring up ripples on the knowledge graph to form a ripple network (RippleNet), we can efficiently and batch-exploit the reasons for the association between massive players and massive recommendation results.
[0095] by Figure 5 For example, pet No. 2 - cat is recommended to the player, and the corresponding recommendation reason is obtained through multiple knowledge graph diffusions of the water wave network.
[0096] For example, the knowledge graph uses a ternary structure such as entity-relationship-entity to represent pieces of knowledge, denoted as (h, r, t), where h represents the entity head node, r represents the relationship link, and t represents the entity tail node. Figure 6 , the ternary relationship (gift package No. 60, subject, pet) means that the gift package No. 60 is mainly for pets. Usually, the products in the recommended product pool are connected to multiple entity nodes in the knowledge graph. For example, in addition to being connected to the "pet" entity node, the gift package No. 60 may also be connected to the "spiritual beast fairy drink" entity node.
[0097] Based on the above Figure 6 The knowledge graph structure shown in Figure 7 The RippleNet network uses the player's historical purchase item ID list as the player feature. The player feature and the predicted item ID feature form a user-item pair input network, and the network output is the probability of this player purchasing this item. The correlation degree (pi) between different entity node paths and predicted items in the knowledge graph obtained during the calculation process is used as the output recommendation reason.
[0098] First, for each player, we use the player's historical purchase id as the node seed of the knowledge graph. Through the structure of the knowledge graph, we perform k expansions according to the k-hop parameter to obtain the node set corresponding to the k-th hop in the knowledge graph, where the 0th hop is the player's historical purchase. We take all the triple combinations (h i ,r i ,t i ) is recorded as the RippleSet for the i-th hop i This ternary combination can be understood as: we expand the features of the previous player's historical purchase information according to the structure of the knowledge graph, and finally use RippleSet i To describe this player's preferences.
[0099] To distinguish RippleSet i The middle and tail node set set(t i ) obtained by different paths i , and the relevance of the final recommended product v, we combine the structure of the knowledge graph and introduce the weight probability p i , used to describe set(t i ) is the importance of different nodes in the i pass Get, which means that according to the head h i And the corresponding conversion relationship r i , through matrix multiplication, we can get the embedding vector representation of this path in the knowledge graph. The correlation between this path embedding vector and the recommended product embedding is used as the weight of this set of conversion relationships.
[0100] After obtaining the correlation between different paths of the knowledge graph and the predicted product, i After that, by and set(t i ) can be used to obtain the player feature vector describing the current jump. By adding the player feature vectors of hops 1 to k, we get the final player feature vector. We multiply the final player feature vector by the product embedding and use the sigmoid function to get the predicted values of the player and the recommended product:
[0101]
[0102] Through the above formula, the entire problem can be transformed into a classic binary classification problem. The model uses cross entropy as the loss function, adds the L2 norm, and is finally trained by the Adam algorithm.
[0103] Once RippleNet is constructed and trained, it can generate product recommendation results for players and use the RippleSet in the result generation process. i The similarity between the knowledge node embedding and the predicted item embedding p i , directly and synchronously generate personalized recommendation reasons.
[0104] Furthermore, after generating the first target recommendation reason according to MetaPath and obtaining the fourth target recommendation reason based on the RippleNet network as shown in the above embodiment, they can also be aggregated to obtain a new target recommendation reason.
[0105] In an optional embodiment of the present disclosure, a first probability of generating a first target recommendation reason is determined, and a second probability of generating a fourth target recommendation reason is determined; the first target recommendation reason and the fourth target recommendation reason are aggregated according to the first probability and the second probability to obtain a fifth target recommendation reason.
[0106] For example, the first target recommendation reason generated based on MetaPath is more controllable, while the fourth target recommendation reason obtained based on the RippleNet network is more flexible and novel. For this reason, in actual operation, the first target recommendation reason and the fourth target recommendation reason obtained by the above two methods are usually aggregated to obtain the fifth target recommendation reason.
[0107] For example, the recommendation reasons generated by MetaPath are displayed with a probability of 70%, making the generated recommendation reasons more controllable to ensure "business security"; the recommendation reasons generated based on the RippleNet network are displayed with a probability of 30%, making the generated recommendation reasons more flexible and novel. The above scheme can aggregate the advantages of the above two methods to generate recommendation reasons that take into account both reliability and novelty, thereby enhancing the player's gaming experience.
[0108] Furthermore, in order to overcome the technical problems of high access threshold and performance issues in the practical application of related technical solutions, a distributed recommendation service infrastructure was built, the structure of which is Figure 8 shown.
[0109] Among them, the knowledge graph data is distributedly stored in the graph database NebulaGraph in the form of shards. The excellent performance of NebulaGraph can support large-scale concurrent queries on knowledge graphs.
[0110] In this embodiment, the MetaPath recommendation reason generation technology is packaged into an interfaced distributed recommendation reason generation service. The service receives recommendation results generated in any way from any game from the interface, matches the MetaPath template and queries the knowledge graph in real time, and returns the hit and generated personalized recommendation reasons in real time. The RippleNet recommendation reason generation technology can also be packaged as a template job on the knowledge graph platform. It only needs to replace the knowledge graph data source in the template according to the different games, and the personalized recommendation reasons can be directly generated synchronously while the recommendation results are generated.
[0111] The development and access experience of recommendation reasons is the last threshold that affects its actual application and service to various game products. In order to reduce the complexity of recommendation reason development and access, shield technical details, and achieve the goal of lowering the threshold, the entire personalized recommendation reason generation solution can be systematically packaged into a platform to form a recommendation reason service platform.
[0112] In the recommendation reason service platform, when building the knowledge graph structure and writing data, users do not need to understand any underlying technical information of the knowledge graph structure. They can complete the knowledge graph structure design by simple clicks, dragging and other operations. And when importing data, users only need to select a pre-designed ontology, click to import the data source, and configure the mapping between the data field and the knowledge graph attribute to complete all the configurations of the knowledge graph data writing. There is no need to prepare data in a special format, understand the underlying information stored in the knowledge graph, or master the relevant statements for writing knowledge graph data. Click run to complete the knowledge graph data writing. And the data for building the graph comes from the original log data of the game in the hive library, and there is no additional data acquisition threshold.
[0113] It should be noted that in order to improve the efficiency of building knowledge graphs, the knowledge graph structures built in historical periods can be stored for one-click reuse. The knowledge graphs with the same structure between different products and different businesses can directly omit the structural design steps, saving the cost of building the knowledge graph. At the same time, for knowledge graphs with the same structure between different products and different businesses, the save as function of the graph can also realize one-click reuse of data writing configuration, omitting the data writing configuration steps and saving the cost of building the knowledge graph.
[0114] In order to implement the above-mentioned method for generating a recommendation reason, an apparatus for generating a recommendation reason is provided in one embodiment of the present disclosure. Fig. 9 The schematic architecture diagram of a device for generating a recommendation reason is schematically shown.
[0115] The recommendation reason generating device includes a data acquisition module 901 , a meta-path determination module 902 and a recommendation reason generating module 903 .
[0116] The data acquisition module 901 is used to acquire heterogeneous game data corresponding to the game application;
[0117] A meta-path determination module 902 is used to determine a meta-path set corresponding to a target user entity in a game application based on a knowledge graph constructed from heterogeneous game data, wherein the meta-path set includes at least one meta-path consisting of a head node, a tail node and / or an intermediate node, and a connection relationship between each node;
[0118] The recommendation reason generating module 903 is used to generate a target recommendation reason corresponding to the recommendation result according to the meta-path set and the recommendation result for the target user entity.
[0119] In an optional embodiment of the present disclosure, the meta-path determination module 902 is specifically used to determine each data entity in the game heterogeneous data; use each data entity in the game heterogeneous data as a node of the knowledge graph, and use the relationship between each data entity as an edge of the knowledge graph, and construct a knowledge graph based on a preset knowledge graph structure; obtain a meta-path set based on the knowledge graph constructed based on the game heterogeneous data.
[0120] In an optional embodiment of the present disclosure, a recommendation reason generation module 903 is used to obtain a target user entity and a recommendation result corresponding to the target user entity as target path information corresponding to a head node and a tail node respectively, and determine whether the target path information satisfies a meta-path structure; if the target path information satisfies the meta-path structure, obtain the target user entity, the recommendation result corresponding to the target user entity, and multiple attribute information of the target path information; and generate a first target recommendation reason for the target user entity and the recommendation result corresponding to the target user entity according to the multiple attribute information and a pre-configured generative text rule template.
[0121] In an optional embodiment of the present disclosure, the recommendation reason generation module 903 can also be used to determine whether the target path information satisfies other meta-path structures if the target path information does not satisfy the meta-path structure, and if so, generate a second target recommendation reason; or, if the target path information does not satisfy the meta-path structure, generate a pre-configured third target recommendation reason.
[0122] In an optional embodiment of the present disclosure, the recommendation reason generation module 903 can also be used to convert the meta-path structure into a query statement template corresponding to the meta-path database, where the meta-path database contains at least one query statement template corresponding to the meta-path; add the target user entity and the recommendation result corresponding to the target user entity to the query statement template to obtain a target detection statement; and determine whether the target path information satisfies the meta-path structure based on the detection result of the target detection statement.
[0123] In an optional embodiment of the present disclosure, the device may also include a network construction module, which is used to construct a RippleNet network based on a preset knowledge graph result composed of the association relationship between the target user entity and other commodity entities; the diffusion module is used to determine the target other commodity entities related to the target user entity based on the game heterogeneous data, and perform multiple diffusions to obtain a diffusion node set corresponding to each diffusion; a recommendation reason generation module 903 is used to determine the weight probability of each diffusion node set in the meta-path set, and generate a fourth target recommendation reason based on the weight probability.
[0124] In an optional embodiment of the present disclosure, a recommendation reason generation module 903 is used to determine a first probability of generating the first target recommendation reason, and to determine a second probability of generating the fourth target recommendation reason; the first target recommendation reason and the fourth target recommendation reason are aggregated according to the first probability and the second probability to obtain the fifth target recommendation reason.
[0125] In an optional embodiment of the present disclosure, a storage module is used to store the knowledge graph in a graph database in the form of shards; and connect the graph database in the form of a function interface so as to generate a target recommendation reason corresponding to the recommendation result by calling the function interface.
[0126] The device 900 for generating reason for recommendation provided in the embodiment of the present disclosure can execute the technical solution of the method for generating reason for recommendation in any of the above-mentioned embodiments, and its implementation principle and beneficial effects are similar to those of the method for generating reason for recommendation, and reference can be made to the implementation principle and beneficial effects of the method for generating reason for recommendation, which will not be repeated here.
[0127] In an exemplary embodiment of the present disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the above method of the present specification is stored. In some possible implementations, various aspects of the present invention may also be implemented in the form of a program product, which includes a program code, and when the program product is run on a terminal device, the program code is used to enable the terminal device to execute the steps according to various exemplary embodiments of the present invention described in the above "Exemplary Method" section of the present specification.
[0128] The program product for implementing the above method according to an embodiment of the present invention may adopt a portable compact disc read-only memory (CD-ROM) and include program code, and may be run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited thereto, and in this document, a readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, an apparatus or a device.
[0129] The program product may adopt any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable 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 above.
[0130] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.
[0131] The program code contained on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, radio frequency (RF), etc., or any suitable combination of the foregoing.
[0132] Program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages such as Java, C++, etc., and conventional procedural programming languages such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., via the Internet using an Internet Service Provider).
[0133] In an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above method is also provided. It will be appreciated by those skilled in the art that various aspects of the present invention may be implemented as a system, method or program product. Therefore, various aspects of the present invention may be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or an implementation combining hardware and software aspects, which may be collectively referred to herein as a "circuit", "module" or "system".
[0134] Refer to the following Fig.10 The electronic device 1000 according to this embodiment of the present invention will be described. Fig.10 The electronic device 1000 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.
[0135] like Fig.10 As shown, the electronic device 1000 is in the form of a general computing device. The components of the electronic device 1000 may include but are not limited to: the at least one processing unit 1010, the at least one storage unit 1020, a bus 1030 connecting different system components (including the storage unit 1020 and the processing unit 1010), and a display unit 1040.
[0136] The storage unit stores program codes, which can be executed by the processing unit 1010, so that the processing unit 1010 performs the steps according to various exemplary embodiments of the present invention described in the above “Exemplary Method” section of this specification. For example, the processing unit 1010 can perform the following steps: Figure 2 Steps S201 to S203 shown in FIG.
[0137] The storage unit 1020 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 10201 and / or a cache storage unit 10202 , and may further include a read-only storage unit (ROM) 10203 .
[0138] The storage unit 1020 may also include a program / utility 10204 having a set (at least one) of program modules 10205, such program modules 10205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.
[0139] Bus 1030 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.
[0140] The electronic device 1000 may also communicate with one or more external devices 1500 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 1000, and / or communicate with any device that enables the electronic device 1000 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 1050. In addition, the electronic device 1000 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 1060. As shown, the network adapter 1060 communicates with other modules of the electronic device 1000 via a bus 1030. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 1000, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, disk arrays (Redundant Arrays of Independent Disks, RAID) systems, tape drives, and data backup storage systems.
[0141] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the implementation of the present disclosure.
[0142] In addition, the above-mentioned figures are only schematic illustrations of the processes included in the method according to an exemplary embodiment of the present invention, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.
[0143] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.
[0144] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary technical means in the art that are not disclosed in the present disclosure. The specification and examples are to be considered as exemplary only, and the true scope and spirit of the present disclosure are indicated by the claims.
[0145] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.
Claims
1. A method for generating a recommendation reason, characterized in that: include: Obtain heterogeneous game data corresponding to the game application; Determine a meta-path set corresponding to a target user entity in the game application based on a knowledge graph constructed based on the game heterogeneous data, wherein the meta-path set includes at least one meta-path consisting of a head node, a tail node and / or an intermediate node, and a connection relationship between each node; According to the meta-path set and the recommendation result for the target user entity, a target recommendation reason corresponding to the recommendation result is generated.
2. The method according to claim 1, characterized in that The step of determining the meta-path set corresponding to the target user entity in the game application based on the knowledge graph constructed by the game heterogeneous data includes: Determine each data entity in the heterogeneous game data; Taking each data entity in the heterogeneous game data as a node of the knowledge graph, and taking the relationship between the data entities as an edge of the knowledge graph, and constructing the knowledge graph based on a preset knowledge graph structure; The meta-path set is obtained based on the knowledge graph constructed based on the heterogeneous game data.
3. The method according to claim 1, characterized in that The step of generating a target recommendation reason corresponding to the recommendation result according to the meta-path set and the recommendation result for the target user entity comprises: Obtaining a target user entity and a recommendation result corresponding to the target user entity as target path information corresponding to a head node and a tail node respectively, and determining whether the target path information satisfies a meta-path structure; If the target path information satisfies the meta-path structure, acquiring the target user entity, the recommendation result corresponding to the target user entity, and multiple attribute information of the target path information; A first target recommendation reason for the target user entity and the recommendation result corresponding to the target user entity is generated according to the multiple attribute information and a pre-configured generative text rule template.
4. The method according to claim 3, characterized in that The method further comprises: If the target path information does not satisfy the meta-path structure, determining whether the target path information satisfies other meta-path structures, and generating a second target recommendation reason if the target path information satisfies other meta-path structures; Alternatively, if the target path information does not satisfy the meta-path structure, a pre-configured third target recommendation reason is generated.
5. The method according to claim 3, characterized in that: The step of determining whether the target path information satisfies the meta-path structure includes: Converting the meta-path structure into a query statement template corresponding to a meta-path database, wherein the meta-path database contains a query statement template corresponding to at least one meta-path; Adding the target user entity and the recommendation result corresponding to the target user entity to the query statement template to obtain a target detection statement; According to the detection result of the target detection statement, it is determined whether the target path information satisfies the meta-path structure.
6. The method according to claim 3, characterized in that The method further comprises: Build the RippleNet network based on the preset knowledge graph results composed of the association relationship between the target user entity and other product entities; Based on the RippleNet network, multiple diffusions are performed on the target other commodity entities related to the target user entity in the heterogeneous data of the game to obtain a diffusion node set corresponding to each diffusion; The weight probability of each diffusion node set in the meta-path set is determined, and a fourth target recommendation reason is generated based on the weight probability.
7. The method according to claim 6, characterized in that The method further comprises: Determining a first probability of generating the first target recommendation reason, and determining a second probability of generating the fourth target recommendation reason; The first target recommendation reason and the fourth target recommendation reason are aggregated according to the first probability and the second probability to obtain a fifth target recommendation reason.
8. The method according to claim 1, characterized in that The method further comprises: Storing the knowledge graph in a graph database in the form of shards; The graph database is connected in the form of a function interface so as to generate a target recommendation reason corresponding to the recommendation result by calling the function interface.
9. A device for generating a recommendation reason, characterized in that: The device comprises: A data acquisition module is used to acquire heterogeneous game data corresponding to the game application; A meta-path determination module is used to determine a meta-path set corresponding to a target user entity in a game application based on a knowledge graph constructed based on heterogeneous game data, wherein the meta-path set includes at least one meta-path consisting of a head node, a tail node and / or an intermediate node, and a connection relationship between each node; The recommendation reason generation module is used to generate a target recommendation reason corresponding to the recommendation result based on the meta-path set and the recommendation result for the target user entity.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for generating a recommendation reason according to any one of claims 1 to 8 is implemented.
11. An electronic device, characterized in that: include: processor; as well as A memory, configured to store executable instructions of the processor; The processor is configured to execute the method for generating a recommendation reason as described in any one of claims 1 to 8 by executing the executable instructions.
Citation Information
Patent Citations
Diet exercise recommendation method and device based on knowledge graph, equipment and medium
CN116597944A
Method and device for recommending articles
CN117172876A
Cited By
Using one or more neural networks to identify service feedback
WO2026123152A1
