An entity object determination method and apparatus, a server, and a storage medium

Abstract

Embodiments of the present application provide an entity object determination method and device, a server and a storage medium, wherein the method comprises: obtaining a target data object associated with a business scenario based on data objects under an object category associated with the business scenario; determining the relevance of an entity object and the business scenario according to historical interaction behavior data of the entity object and the target data object; determining the entity object as an entity object associated with the business scenario if the relevance of the entity object and the business scenario is not less than a preset relevance threshold, and the entity objects associated with the business scenario form an entity object group of the business scenario. Embodiments of the present application can determine the group of entity objects associated with the business scenario, and improve the comprehensiveness of group determination.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to a method, apparatus, server, and storage medium for determining entity objects. Background Technology

[0002] There are massive amounts of data objects on the Internet, and for different business scenarios, it is necessary to push data objects related to the business scenario to the entity objects.

[0003] In order to enable data objects related to business scenarios to be provided to the entities that need these data objects, it is particularly necessary to identify the group of entities associated with the business scenario. Therefore, how to provide an entity object identification scheme to identify the group of entities associated with the business scenario and improve the comprehensiveness of the group identification has become a technical problem that needs to be solved by those skilled in the art. Summary of the Invention

[0004] In view of this, embodiments of this application provide a method, apparatus, server, and storage medium for determining entity objects, so as to determine a group of entity objects associated with a business scenario and improve the comprehensiveness of group determination.

[0005] To achieve the above objectives, the embodiments of this application provide the following technical solutions:

[0006] A method for determining entity objects, comprising:

[0007] Based on the data objects under the object categories associated with the business scenario, obtain the target data objects associated with the business scenario;

[0008] Based on the historical interaction data between the entity object and the target data object, the relevance between the entity object and the business scenario is determined;

[0009] Entity objects whose relevance to the business scenario is not less than a preset relevance threshold are identified as entity objects associated with the business scenario, and the entity objects associated with the business scenario form the entity object group of the business scenario.

[0010] This application embodiment also provides an entity object determination device, including:

[0011] The target data object acquisition module is used to acquire the target data object associated with the business scenario based on the data object under the object category associated with the business scenario;

[0012] The relevance determination module is used to determine the relevance between the entity object and the business scenario based on the historical interaction behavior data between the entity object and the target data object.

[0013] The group determination module is used to determine entity objects whose relevance to the business scenario is not less than a preset relevance threshold as entity objects associated with the business scenario, and the entity objects associated with the business scenario form the entity object group of the business scenario.

[0014] This application also provides a server, including: at least one memory and at least one processor; the memory stores one or more computer-executable instructions, and the processor invokes the one or more computer-executable instructions to execute the entity object determination method as described above.

[0015] This application embodiment also provides a storage medium that stores one or more computer-executable instructions, which are used to execute the entity object determination method as described above.

[0016] The entity object determination method provided in this application embodiment can obtain target data objects associated with a business scenario based on data objects under object categories associated with that business scenario, avoiding the direct use of data objects associated with business scenarios described in cognitive graphs to determine the population of a business scenario. Therefore, based on an expanded set of target data objects associated with the business scenario, this application embodiment can determine the relevance of entity objects to the business scenario based on historical interaction data between entity objects and the target data objects. Entity objects with a relevance to the business scenario not less than a preset relevance threshold are then identified as entity objects associated with the business scenario. These entity objects form a group of entity objects associated with the business scenario, thus achieving the determination of the group of entity objects associated with the business scenario. Since this application embodiment determines the group of entity objects associated with the business scenario based on historical interaction data between entity objects and target data objects, it can, while ensuring the accuracy of entity object group determination, expand the group of entity objects associated with the business scenario as much as possible based on the expanded set of target data objects, thereby improving the comprehensiveness of group determination. As can be seen, the entity object determination method provided in this application embodiment can determine the group of entity objects associated with the business scenario and improve the comprehensiveness of the group determination. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0018] Figure 1 This is an example diagram showing the number of data objects associated with the overall business scenario in the cognitive graph;

[0019] Figure 2 A flowchart of the entity object determination method provided in the embodiments of this application;

[0020] Figure 3 A flowchart for determining the target data object associated with a business scenario, provided in an embodiment of this application;

[0021] Figure 4 Example diagrams for determining the matching sequence provided in embodiments of this application;

[0022] Figure 5 A flowchart for vectorizing business scenarios and data objects provided in this application embodiment;

[0023] Figure 6 This is an example diagram of the construction of the association graph provided in the embodiments of this application;

[0024] Figure 7 Example diagram of the sampling path provided in the embodiments of this application;

[0025] Figure 8 Example diagrams provided for training using the skip-gram algorithm in embodiments of this application;

[0026] Figure 9 Example diagram of the process of vectorizing business scenarios and data objects provided in the embodiments of this application;

[0027] Figure 10 A flowchart for determining the relevance of an entity object to a business scenario, provided as an embodiment of this application;

[0028] Figure 11 Example chart comparing the number of data objects associated with a business scenario;

[0029] Figure 12 Another comparative example chart showing the number of data objects associated with a business scenario;

[0030] Figure 13 A block diagram of the entity object determination device provided in the embodiments of this application;

[0031] Figure 14 Another block diagram of the entity object determination device provided in the embodiments of this application;

[0032] Figure 15 A block diagram of a server provided in an embodiment of this application. Detailed Implementation

[0033] E-commerce and other internet platforms can use cognitive graphs to identify groups of entity objects associated with business scenarios. In one example, an entity object might be a user, and a group of entity objects might be a user population. It's important to note that cognitive graphs explicitly express the needs of entity objects as nodes (called E-commerce Concepts) and connect these nodes with data objects (e.g., products), object categories, and external general domain knowledge, providing a unified data foundation for data object cognition, entity object cognition, and knowledge cognition. Specifically, the needs of entity objects in a cognitive graph can be defined as common-sense, semantically complete, and logically coherent phrases (e.g., preventing children from getting lost, gift-giving during the Mid-Autumn Festival). Business scenarios can be added to this cognitive graph to establish a connection between business scenarios and data objects. Furthermore, by utilizing the relationships between business scenarios and data objects in the cognitive graph, as well as the behavior of entity objects towards data objects over a historical period, the relationships between business scenarios and entity objects can be indirectly obtained, thus identifying the group of entity objects associated with a business scenario.

[0034] Specifically, based on the relationship between business scenarios and data objects in the cognitive graph, multiple data objects under the business scenario can be identified. Then, for a certain data object, the score of the entity object and the data object can be determined according to the historical behavior data of the entity object and the data object. Then, by combining the scores of the entity object and the data objects of the business scenario, the score of the entity object and the business scenario can be obtained. Finally, based on the score of the entity object and the business scenario, it can be determined whether the entity object belongs to the business scenario, thus realizing the identification of the entity object group of the business scenario.

[0035] However, the above method of determining the group of entity objects is based on the relationship between the business scenario and the data object described by the cognitive graph. Although the relationship between the business scenario and the data object described by the cognitive graph has high accuracy, the recall rate is insufficient. This means that if the number of data objects associated with the business scenario is small, the number of data objects that the entity object can act on in the business scenario is also small, resulting in a small group of entity objects in the determined business scenario and making it impossible to comprehensively determine the group of entity objects associated with the business scenario.

[0036] For example, such as Figure 1As shown, the number of data objects associated with different business scenarios in the cognitive graph varies, with some business scenarios having a very small number of associated data objects. One reason for this is that the concepts in the cognitive graph are mainly used for data object recommendation (e.g., for product recommendation), and the recall of data objects covered by the main concepts only needs to meet the data object recommendation requirements. However, a small number of data objects associated with a business scenario leads to a small group of identifiable entity objects in that business scenario. This small group of entity objects is clearly insufficient to meet the purposes of acquiring new entity objects, large-scale data object deployment, and recommendation in the business scenario. Therefore, while ensuring the accuracy of entity object group identification, how to expand the group of entity objects associated with the business scenario as much as possible and improve the comprehensiveness of entity object group identification has become an urgent problem to be solved.

[0037] Based on this, embodiments of this application provide a method, apparatus, server, and storage medium for determining entity objects, so as to determine a group of entity objects associated with a business scenario and improve the comprehensiveness of the determination of the entity object group.

[0038] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0039] In the optional implementation, Figure 2 This application illustrates an optional flow of the entity object determination method provided in an embodiment. This method can be applied to a server, such as an internet platform server, specifically an e-commerce platform server. Any business scenario set up by an e-commerce or other internet platform can be based on… Figure 2 The process shown identifies the group of entity objects associated with the business scenario (e.g., identifying the user group associated with the business scenario), such as... Figure 2 As shown, the method flow may include:

[0040] Step S100: Obtain the target data object associated with the business scenario based on the data object under the object category associated with the business scenario.

[0041] To maximize the number of entity objects associated with business scenarios and improve the comprehensiveness of entity object identification, this application embodiment can redetermine data objects associated with business scenarios. This expands the group of entity objects associated with business scenarios. To further expand the number of data objects associated with business scenarios, this application embodiment can obtain the data objects associated with business scenarios based on data objects under the object categories associated with business scenarios, avoiding the direct use of data objects associated with business scenarios as described in the cognitive graph. This expands the number of data objects associated with business scenarios. For ease of explanation, this application embodiment refers to the redetermined data objects associated with business scenarios as target data objects.

[0042] In a more specific optional implementation, embodiments of this application may determine the data objects under the object categories associated with the business scenarios as described in the cognitive graph (the categories of data objects) and the category attributes of the data objects, thereby determining the similarity between the business scenarios and the data objects (i.e., the data objects under the object categories associated with the business scenarios), and selecting the target data objects associated with the business scenarios from the data objects based on the similarity between the business scenarios and the data objects.

[0043] Step S110: Determine the relevance between the entity object and the business scenario based on the historical interaction behavior data between the entity object and the target data object.

[0044] Based on the target data objects associated with the business scenario, this application embodiment can obtain historical interaction behavior data between the entity object and the target data objects, thereby determining the relevance between the entity object and the business scenario based on the historical interaction behavior data between the entity object and the target data objects. For example, there can be multiple target data objects. This application embodiment can obtain the behavior data of the entity object towards each target data object within a historical time period. The behavior data can describe the behavior type, frequency, and other information of the entity object towards the target data object. Based on the behavior data of the entity object towards each target data object within a historical time period, this application embodiment can determine the relevance between the entity object and the business scenario. The relevance between the entity object and the business scenario can be represented by a score. The greater the relevance between the entity object and the business scenario, the higher the relevance between the entity object and the business scenario.

[0045] In an optional implementation, the embodiments of this application may define the following assumptions:

[0046] If an entity object performs an action on a data object in a certain business scenario, then the entity object is related to that business scenario. Different types of actions of the entity object represent different degrees of relatedness. For example, if entity object A purchases a data object in business scenario 1, while entity object B browses a data object in business scenario 1, then the relatedness of entity object A to business scenario 1 is higher than that of entity object B to business scenario 1.

[0047] The more times an entity interacts with a data object in a particular business scenario, the higher the relevance of that entity to that business scenario.

[0048] Based on the above assumptions, in the optional implementation of determining the relevance between an entity object and a business scenario, the embodiments of this application may first determine the relevance between the entity object and each target data object based on the entity object's behavior data for each target data object within a historical time period (which can be defined using scores), and then determine the relevance between the entity object and the business scenario based on the relevance between the entity object and each target data object.

[0049] Optionally, the historical time period refers to a period of time in the past, and the specific time period can be set according to the actual situation. This application embodiment does not impose any restrictions.

[0050] Step S120: Entity objects whose relevance to the business scenario is not less than a preset relevance threshold are identified as entity objects associated with the business scenario, and the entity objects associated with the business scenario form a group of entity objects of the business scenario.

[0051] This application embodiment can set or specify a relevance threshold. After determining the relevance between an entity object and the business scenario, this application embodiment can select entity objects whose relevance to the business scenario is not less than the relevance threshold through the relevance threshold, obtain entity objects associated with the business scenario, and then form a group of entity objects of the business scenario from the entity objects associated with the business scenario, thereby realizing the determination of the group of entity objects associated with the business scenario.

[0052] Optionally, the specific value of the correlation threshold can be set or specified according to the actual situation, and this application embodiment does not impose any limitations.

[0053] The entity object determination method provided in this application embodiment can obtain target data objects associated with a business scenario based on data objects under object categories associated with that business scenario, avoiding the direct use of data objects associated with business scenarios described in cognitive graphs to determine the population of a business scenario. Therefore, based on an expanded set of target data objects associated with the business scenario, this application embodiment can determine the relevance of entity objects to the business scenario based on historical interaction data between entity objects and the target data objects. Entity objects with a relevance to the business scenario not less than a preset relevance threshold are then identified as entity objects associated with the business scenario. These entity objects form a group of entity objects associated with the business scenario, thus achieving the determination of the group of entity objects associated with the business scenario. Since this application embodiment determines the group of entity objects associated with the business scenario based on historical interaction data between entity objects and target data objects, it can, while ensuring the accuracy of entity object group determination, expand the group of entity objects associated with the business scenario as much as possible based on the expanded set of target data objects, thereby improving the comprehensiveness of group determination. As can be seen, the entity object determination method provided in this application embodiment can determine the group of entity objects associated with the business scenario and improve the comprehensiveness of the group determination.

[0054] In the optional implementation, Figure 3 This application illustrates an optional process for determining target data objects associated with business scenarios, provided by an embodiment of the present application. This method can expand the scope of target data objects associated with business scenarios based on the object types associated with business scenarios as described in the cognitive graph, and the category attributes of the data objects, thereby solving the problem of small-scale data objects associated with some business scenarios in the cognitive graph. Optionally, this method can be applied to a server, such as the server of an internet platform, specifically an e-commerce platform server. Any business scenario set up by the internet platform can be based on... Figure 3 The process shown determines the target data object associated with the business scenario, such as... Figure 3 As shown, the method flow may include:

[0055] Step S200: Based on the object categories associated with business scenarios and the category attributes of data objects described in the cognitive graph, determine the candidate data objects under the object categories associated with business scenarios.

[0056] Optionally, in this application embodiment, based on the association between business scenarios and object categories in the cognitive graph, and the category attributes of data objects, all data objects whose category attributes match the object categories associated with the business scenario can be determined. All data objects can be used as candidate data objects under the object categories associated with the business scenario. In this application embodiment, the target data object can be selected from the candidate data objects. The candidate data objects under the object categories associated with the business scenario can constitute a set of candidate data objects for the business scenario. One business scenario can correspond to one set of candidate data objects. For example, taking an e-commerce platform as an example, the object categories associated with the business scenario of European and American makeup are such as cosmetics or perfumes. Then, all data objects under the object categories of cosmetics or perfumes can be uniformly determined as candidate data objects for the business scenario of European and American makeup, thus obtaining a set of candidate data objects for the business scenario of European and American makeup.

[0057] Step S210: Determine the similarity between the business scenario and the candidate data object.

[0058] For each candidate data object, the embodiments of this application can calculate the similarity between the business scenario and each candidate data object. In an optional implementation, the embodiments of this application can obtain the business scenario vector corresponding to the business scenario and the data object vector corresponding to each candidate data object, thereby calculating the similarity between the business scenario vector of the business scenario and the data object vector of each candidate data object, and obtaining the similarity between the business scenario and each candidate data object.

[0059] In a further optional implementation, embodiments of this application can determine the similarity between the business scenario vector and the candidate data object vector by calculating the cosine distance, Euclidean distance, or Pearson correlation coefficient between the business scenario vector and the data object vector.

[0060] Step S220: Based on the similarity between the business scenario and the candidate data object, determine the similarity threshold corresponding to the business scenario.

[0061] The similarity between a business scenario and candidate data objects can represent the relevance between the business scenario and the candidate data objects. In this application embodiment, a similarity threshold corresponding to the business scenario can be determined based on the similarity between the business scenario and each candidate data object, so that objects associated with the business scenario can be selected from the candidate data objects through the similarity threshold.

[0062] In an optional implementation, embodiments of this application may use a heuristic method to determine the similarity threshold corresponding to the business scenario; optionally, embodiments of this application may sort candidate data objects according to similarity, such as sorting candidate data objects in descending order of similarity, determining an order position that matches at least a first proportion of values ​​from the sorting, and using the similarity of the candidate data object corresponding to the order position as the similarity threshold corresponding to the business scenario.

[0063] Optionally, the first proportion can be matched with a first proportion of the number of data objects associated with the business scenarios described in the cognitive graph; for a clearer explanation, as shown below... Figure 4 As shown in the example, there are M candidate data objects in the business scenario. After sorting the candidate data objects in the business scenario according to the order of similarity from largest to smallest, the candidate data object corresponding to the order j is the j-th candidate data object in the sort. The value of j represents the j candidate data objects sorted according to the order of similarity from largest to smallest. Based on the association relationship between the business scenario and the data objects described in the cognitive graph, the embodiment of this application can determine the number of data objects associated with the business scenario described in the cognitive graph, and take a first proportion of the number of data objects to obtain the first proportion quantity. If the value of the order j matches at least the first proportion quantity (e.g., the order j is not less than the first proportion quantity, or the order j is the order closest to the first proportion quantity in the sort, etc.), then the embodiment of this application can use the similarity of the candidate data object corresponding to the order j as the similarity threshold corresponding to the business scenario.

[0064] Step S230: Based on the similarity threshold, determine the target data objects associated with the business scenario from the candidate data objects, whose similarity is not less than the similarity threshold.

[0065] After obtaining the similarity threshold of the business scenario, this application embodiment can select data objects with a similarity not less than the similarity threshold from the candidate data objects based on the similarity threshold, and use them as target data objects associated with the business scenario.

[0066] based on Figure 3 As shown in the process, this application embodiment can select target data objects associated with the business scenario from candidate data objects under the object category associated with the business scenario, thereby expanding the number of target data objects associated with the business scenario and providing a basis for expanding the group of entity objects associated with the business scenario.

[0067] Optional, in Figure 3In the process of determining the similarity between the business scenario and the candidate data object, the embodiments of this application can obtain the vectorized representations of the business scenario and the candidate data object. Then, by using the vectorized business scenario vector and the vectorized data object vector of the candidate data object, the similarity between the business scenario vector and the data object vector is calculated to obtain the similarity between the business scenario and the candidate data object. Based on this, in an optional implementation, Figure 5 This illustrates an optional process for vectorizing business scenarios with data objects, such as... Figure 5 As shown, the process may include:

[0068] Step S300: Based on the cognitive graph, the association between multiple business scenarios and data objects is represented, and each business scenario and each data object is set as a node to construct an association graph between business scenarios and data objects.

[0069] A cognitive graph can represent the relationship between multiple business scenarios and data objects. In other words, a cognitive graph can represent the data objects associated with multiple businesses. In this embodiment, the relationship between multiple business scenarios and data objects represented by the cognitive graph can be expressed as a relationship graph. That is, based on the relationship between multiple business scenarios and data objects represented by the cognitive graph, a relationship graph between business scenarios and data objects is constructed by setting each business scenario and each data object as a node.

[0070] In an optional implementation, embodiments of this application can use the business scenarios described in the cognitive graph as top nodes and the data objects described in the cognitive graph as child nodes. Based on the association relationship between the business scenarios and data objects described in the cognitive graph, the associated top nodes and child nodes are connected to obtain the association graph; for example, such as Figure 6 As shown, taking the cognitive graph to represent the data objects associated with business scenario 1 and business scenario 2 respectively as an example, where business scenario 1 is associated with data objects 1 and 2, and business scenario 2 is associated with data objects 2 and 3, then in this embodiment of the application, business scenario 1 and business scenario 2 can be the top nodes, and data objects 1, 2 and 3 can be the child nodes. Based on the association relationship between business scenario 1 and data objects 1 and 2, the top node of business scenario 1 is connected to the child nodes of data objects 1 and 2. Based on the association relationship between business scenario 2 and data objects 2 and 3, the top node of business scenario 2 is connected to the child nodes of data objects 2 and 3, thereby realizing the construction of the association graph.

[0071] Optionally, the association diagram can be in the form of a heterogeneous diagram, a bipartite diagram, etc., and the embodiments of this application are not limited thereto.

[0072] Step S310: Perform a random walk on the association graph starting from any node, sample the path of the random walk, and obtain sampled documents. Each sampled document includes a node passed by the path, and multiple sampled documents form a sampled document set.

[0073] After obtaining the association graph, for any node in the association graph, this embodiment of the application can perform a random walk with the node as the starting point of the path, and sample the random walk path to obtain a sampled document. A sampled document may include the nodes passed by a random walk path. Correspondingly, by sampling multiple random walk paths, multiple sampled documents can be obtained, and multiple sampled documents can form a sampled document set.

[0074] For example, such as Figure 7 As shown, a random walk starting from a node of data object 1 can sample the following path: Data object 1 → Business scenario 1 → Data object 2. The sampled document for this path can include the nodes traversed by the path, i.e., nodes of data object 1, nodes of business scenario 1, and nodes of data object 2. A random walk starting from a node of business scenario 2 can sample the following path: Business scenario 2 → Data object 2 → Business scenario 1. The sampled document for this path can include the nodes traversed by the path, i.e., nodes of business scenario 2, nodes of data object 2, and nodes of business scenario 1. A random walk starting from a node of data object 3 can sample the following path: Data object 3 → Business scenario 2 → Data object 2. The sampled document for this path can include the nodes traversed by the path, i.e., nodes of data object 3, nodes of business scenario 2, and nodes of data object 2. Of course, the paths obtained by sampling above are only illustrative examples. Random walks starting from different nodes in the association graph can sample different paths, thus forming a set of sampled documents for different paths.

[0075] Step S320: Based on the sampled document set, perform vectorization training on each business scenario and each data object.

[0076] The sampled document set includes sampled documents corresponding to multiple paths of random walks, and the sampled documents include the nodes traversed by the paths. Since business scenarios and data objects are represented using nodes in the association graph, the order of the nodes included in the sampled documents can reflect the relationship between the business scenarios and data objects represented by the nodes. Based on this, the embodiments of this application can perform vectorized training on each node involved in the sampled document set formed by multiple sampled documents, so as to realize vectorized training on each business scenario and each data object.

[0077] In an optional implementation, based on the sampled document set, this embodiment of the application can use the skip-gram algorithm to perform vectorization training on each business scenario and each data object. It should be noted that the skip-gram algorithm is also known as the skip-word model algorithm. The text description of the business scenario and data object can be regarded as a word description. The skip-gram algorithm can generate words around a certain word in the text sequence, thereby realizing word vectorization. Therefore, the skip-gram algorithm can be used to perform vectorization training on each business scenario and each data object.

[0078] Example, combination Figure 7 and Figure 8 As shown, in Figure 7 Based on the above, the order of nodes in the sampled documents can reflect the relationship between the business scenario represented by the nodes and the data objects. For any node in the sampled document set, this embodiment can use the skip-gram algorithm to generate the surrounding nodes in the sampled document set, thereby realizing the vectorization of nodes (a node may represent a business scenario or a data object), such as... Figure 8 As shown, the nodes surrounding the node of business scenario 1 in the sampled document set are data object 1 and data object 2, and the nodes surrounding data object 2 in the sampled document set are business scenario 1.

[0079] Step S330: After training convergence, obtain the business scenario vector corresponding to each business scenario and the data object vector corresponding to each data object.

[0080] This application embodiment can define convergence conditions for vectorized training of business scenarios and vectorized training of data objects. After the vectorized training of business scenarios and data objects reaches the training convergence condition, this application embodiment can obtain the business scenario vector corresponding to each business scenario and the object vector corresponding to each data object.

[0081] In the optional implementation, Figure 5 The illustrated process can be based on unsupervised graph representation learning to vectorize business scenarios and data objects, thereby providing a foundation for data object expansion and recall within business scenarios. In one example, the process of vectorizing business scenarios and data objects can be as follows: Figure 9 As shown, firstly, the relationships between multiple business scenarios and data objects described in the cognitive graph are represented using a relationship graph. Then, random walks are performed on the nodes of the relationship graph, and the paths obtained from the walks are sampled to obtain a sampled document set. Subsequently, the skip-gram algorithm is used to train the vectorization of each node in the sampled document set. After the training converges, the learned vectorized representation of each node is obtained, that is, the business scenario vector corresponding to each business scenario and the object vector corresponding to each data object are obtained.

[0082] exist Figure 2 In the optional implementation of the process for determining the relevance between an entity object and a business scenario, this embodiment of the application may first determine the relevance between the entity object and each target data object associated with the business scenario, and then determine the relevance between the entity object and the business scenario based on the relevance between the entity object and each target data object; in an optional implementation, if there are multiple target data objects, Figure 10 This application illustrates an optional process for determining the relevance of an entity object to a business scenario, as provided in an embodiment of the present application. Figure 10 As shown, the process may include:

[0083] Step S400: Based on the behavioral data of entity objects towards each target data object within a historical time period, determine the correlation between entity objects and each target data object.

[0084] Optionally, for any target data object associated with a business scenario, this embodiment of the application can obtain behavioral parameters corresponding to various behavioral types of the entity object towards the target data object based on the entity object's behavioral data towards the target data object within a historical time period. It should be noted that an entity object may engage in multiple types of behavior towards a target data object within a historical time period. The behavioral parameter corresponding to a behavior type can represent the degree of the entity object's behavior towards the target data object in that behavior type. For example, behavior types may include browsing, purchasing, searching, etc. The behavioral parameters for browsing can be such as the number of browsing sessions and browsing time, the behavioral parameters for purchasing can be such as the number of purchase sessions and the purchase price, and the behavioral parameters for searching can be such as the number of searches and the search time. Therefore, for any target data object, this embodiment of the application can combine the behavioral parameters corresponding to various behavioral types of the entity object towards the target data object with the behavioral weights corresponding to the behavioral types, and then accumulate the combined results to obtain the relevance of the entity object towards the target data object.

[0085] Optionally, the formula for calculating the relevance between an entity object and a target data object can be as follows:

[0086]

[0087] Where u represents entity object u, i represents target data object i, score(u, i) represents the relevance of entity object u to target data object i, a represents the a-th behavior type, A represents all behavior types, and typa (a)(u，i) The weight represents the behavior parameter of entity object u for the a-th behavior type of target data object i. aThis represents the behavior weight of the a-th behavior type. Therefore, when determining the relevance of an entity object to any target data object, this embodiment of the application can multiply the behavior parameter of one behavior type of the entity object to the target data object with the behavior weight of that behavior type to obtain the multiplication result of one behavior type of the entity object to the target data object. Then, the multiplication results of various behavior types of the entity object to the target data object are accumulated to obtain the relevance of the entity object to the target data object.

[0088] Step S410: Accumulate the correlation between the entity object and each target data object to obtain the correlation between the entity object and the business scenario.

[0089] After obtaining the correlation between the entity object and each target data object associated with the business scenario, the embodiments of this application can accumulate the correlation between the entity object and each target data object associated with the business scenario to obtain the correlation between the entity object and the business scenario.

[0090] Optionally, the relevance between entity objects and business scenarios can be obtained using the following formula:

[0091] Where s represents business scenario s.

[0092] This application embodiment does not directly use the data objects associated with business scenarios described by cognitive graphs. Instead, it obtains the target data objects associated with the business scenarios based on the data objects under the object categories associated with the business scenarios, thereby redefining the data objects associated with the business scenarios. This can greatly increase the number of target data objects associated with the business scenarios. Based on expanding the target data objects associated with the business scenarios, this application embodiment determines the group of entity objects associated with the business scenarios, which can ensure the scale of the group of entity objects associated with the business scenarios, solve the problem of insufficient scale of the group of entity objects associated with the business scenarios, and improve the comprehensiveness of the group determination.

[0093] Through experimental comparison Figure 11 This paper presents a comparative example of the number of data objects associated with a business scenario after redefining the data objects associated with that scenario. It can be seen that the number of data objects redefining the business scenario in this embodiment is nearly double the number of data objects associated with the business scenario described in the cognitive graph. Further experimental comparisons... Figure 12 This shows that the number of data objects across all business scenarios has increased, and the overall distribution has shifted to the right.

[0094] In a further experimental comparison, taking users as the entity object, Table 1 below shows the comparison between the number of data objects originally described in the cognitive graph and the number of data objects re-determined in the embodiments of this application under the two business scenarios of "Spring Festival Wine" and "Winter Warming Wine", as well as the comparison of the determined user group size; it can be seen that both the number of data objects associated with the business scenarios determined in the embodiments of this application and the user group size associated with the business scenarios have been increased by about 100%.

[0095]

[0096] Table 1

[0097] It is evident that the embodiments of this application have a significant effect on expanding the scale of the entity object group associated with business scenarios and improving the comprehensiveness of group identification.

[0098] The foregoing describes multiple embodiment schemes provided by the embodiments of this application. The optional methods described in each embodiment scheme can be combined and cross-referenced with each other without conflict, thereby extending to a variety of possible embodiment schemes. These can all be considered as the embodiment schemes disclosed and published by the embodiments of this application.

[0099] The entity object determination apparatus provided in the embodiments of this application is described below. The entity object determination apparatus described below can be considered as a server, which is a functional module required to implement the entity object determination method provided in the embodiments of this application. The content of the entity object determination apparatus described below can be referred to in correspondence with the content of the entity object determination method described above.

[0100] In the optional implementation, Figure 13 An optional block diagram of the entity object determination device provided in an embodiment of this application is shown, such as... Figure 13 As shown, the device may include:

[0101] The target data object acquisition module 100 is used to acquire the target data object associated with the business scenario based on the data object under the object category associated with the business scenario;

[0102] The relevance determination module 110 is used to determine the relevance between the entity object and the business scenario based on the historical interaction behavior data between the entity object and the target data object.

[0103] The group determination module 120 is used to determine entity objects whose relevance to the business scenario is not less than a preset relevance threshold as entity objects associated with the business scenario, and the entity objects associated with the business scenario form a group of entity objects of the business scenario.

[0104] Optionally, there are multiple target data objects; the relevance determination module 110 is used to determine the relevance between the entity object and the business scenario based on the historical interaction behavior data between the entity object and the target data object, including:

[0105] Based on the behavioral data of entity objects towards each target data object within a historical time period, determine the correlation between entity objects and each target data object;

[0106] The relevance of the entity object to the business scenario is determined based on the relevance between the entity object and each target data object.

[0107] Optionally, the relevance determination module 110 is used to determine the relevance between an entity object and each target data object based on the entity object's behavioral data towards each target data object within a historical time period, including:

[0108] For any target data object, based on the behavior data of entity objects towards the target data object within a historical time period, obtain the behavior parameters corresponding to various behavior types of entity objects towards the target data object; combine the behavior parameters corresponding to various behavior types of entity objects towards the target data object with the behavior weights corresponding to the behavior types respectively, and sum the combined results to obtain the relevance of entity objects towards the target data object.

[0109] Optionally, the relevance determination module 110 is used to determine the relevance between the entity object and the business scenario based on the relevance between the entity object and each target data object, including:

[0110] The correlation between the entity object and each target data object is accumulated to obtain the correlation between the entity object and the business scenario.

[0111] Optionally, the target data object acquisition module 100 is used to acquire the target data object associated with the business scenario based on data objects under the object category associated with the business scenario, including:

[0112] Based on the object categories and data object category attributes associated with the business scenarios described in the cognitive graph, candidate data objects under the object categories associated with the business scenarios are determined.

[0113] Determine the similarity between the business scenario and the candidate data object;

[0114] Based on the similarity between the business scenario and the candidate data objects, a target data object associated with the business scenario is selected from the candidate data objects.

[0115] Optionally, the target data object acquisition module 100 is used to determine the similarity between the business scenario and the candidate data object, including:

[0116] Obtain the business scenario vector corresponding to the business scenario and the data object vector corresponding to the candidate data object;

[0117] Calculate the similarity between the business scenario vector and the data object vector to obtain the similarity between the business scenario and the candidate data object.

[0118] Optionally, the target data object acquisition module 100 is used to select target data objects associated with the business scenario from the candidate data objects based on the similarity between the business scenario and the candidate data objects, including:

[0119] Based on the similarity between the business scenario and the candidate data object, a similarity threshold corresponding to the business scenario is determined;

[0120] Based on the similarity threshold, target data objects associated with the business scenario are determined from the candidate data objects, whose similarity is not less than the similarity threshold.

[0121] Optionally, the target data object acquisition module 100 is used to determine a similarity threshold corresponding to the business scenario based on the similarity between the business scenario and the candidate data object, including:

[0122] Candidate data objects are sorted according to similarity. From the sorted data, a position is determined that matches at least a first proportion of values. The similarity of the candidate data objects corresponding to the position is used as the similarity threshold corresponding to the business scenario.

[0123] Optionally, the first proportion is matched with a first proportion of the number of data objects associated with the business scenario described in the cognitive graph.

[0124] Optional, further, Figure 14 This illustration shows another optional block diagram of the entity object determination apparatus provided in the embodiments of this application, in conjunction with... Figure 13 and Figure 14 As shown, the device may further include:

[0125] The vectorization module 130 is used to represent the association between multiple business scenarios and data objects based on the cognitive graph. Each business scenario and each data object is set as a node to construct an association graph between business scenarios and data objects. A random walk is performed on the association graph starting from any node, and the paths of the random walk are sampled to obtain sampled documents. Each sampled document includes a node traversed by the path, and multiple sampled documents form a sampled document set. Based on the sampled document set, vectorization training is performed on each business scenario and each data object. After training convergence, the business scenario vector corresponding to each business scenario and the data object vector corresponding to each data object are obtained.

[0126] Optionally, the vectorization module 130 is used to represent the relationships between multiple business scenarios and data objects based on the cognitive graph. Each business scenario and each data object is set as a node to construct a relationship graph between business scenarios and data objects, including:

[0127] Using the business scenarios described in the cognitive graph as top nodes and the data objects described in the cognitive graph as child nodes, the related top nodes and child nodes are connected based on the association between the business scenarios and data objects described in the cognitive graph to obtain the association graph.

[0128] Optionally, the vectorization module 130 is used to perform vectorization training on each business scenario and each data object based on the sampled document set, including:

[0129] Based on the sampled document set, the skip character model algorithm is used to perform vectorized training on each business scenario and each data object.

[0130] This application embodiment also provides a server that can be loaded with the entity object determination device described above to implement the entity object determination method provided in this application embodiment. In an optional implementation, Figure 15 An optional block diagram of the server is shown, such as... Figure 15 As shown, the server may include: at least one processor 1, at least one communication interface 2, at least one memory 3, and at least one communication bus 4;

[0131] In this embodiment of the application, the number of processor 1, communication interface 2, memory 3, and communication bus 4 is at least one, and processor 1, communication interface 2, and memory 3 communicate with each other through communication bus 4;

[0132] Optionally, communication interface 2 can be an interface for a communication module used for network communication;

[0133] Optionally, processor 1 may be a CPU (Central Processing Unit), GPU (Graphics Processing Unit), NPU (Embedded Neural Network Processor), FPGA (Field Programmable Gate Array), TPU (Tensor Processing Unit), AI chip, ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of this application.

[0134] Memory 3 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0135] The memory 3 stores one or more computer-executable instructions, and the processor 1 invokes the one or more computer-executable instructions to execute the entity object determination method provided in the embodiments of this application.

[0136] This application also provides a storage medium that stores one or more computer-executable instructions for executing the entity object determination method provided in this application.

[0137] While the embodiments disclosed above are described in this application, this application is not limited thereto. Any person skilled in the art can make various modifications and alterations without departing from the spirit and scope of this application; therefore, the scope of protection of this application should be determined by the scope defined in the claims.

Claims

1. A method of determining a physical object, wherein, The method comprises the following steps: Based on the data objects under the object category associated with the business scenario, the target data objects associated with the business scenario are obtained; According to the historical interaction behavior data of the entity object and the target data objects, the relevance of the entity object to the business scenario is determined; The entity objects with a relevance to the business scenario not less than a preset relevance threshold are determined as the entity objects associated with the business scenario, and the entity objects associated with the business scenario form the entity object group of the business scenario; Wherein, the target data objects are multiple; the relevance of the entity object to the business scenario is determined according to the historical interaction behavior data of the entity object and the target data objects, which comprises: Based on the behavior data of the entity object to each target data object in the historical time period, the relevance of the entity object to each target data object is determined; According to the relevance of the entity object to each target data object, the relevance of the entity object to the business scenario is determined.

2. The physical object determination method according to claim 1, wherein, The relevance of the entity object to each target data object is determined based on the behavior data of the entity object to each target data object in the historical time period, which comprises: For any target data object, based on the behavior data of the entity object to the target data object in the historical time period, the behavior parameters corresponding to various behavior types of the entity object to the target data object are obtained; the behavior parameters corresponding to various behavior types of the entity object to the target data object are combined with the behavior weights corresponding to the behavior types respectively, and the combined results are accumulated to obtain the relevance of the entity object to the target data object.

3. The physical object determination method according to claim 1, wherein, The relevance of the entity object to the business scenario is determined according to the relevance of the entity object to each target data object, which comprises: The relevance of the entity object to each target data object is accumulated to obtain the relevance of the entity object to the business scenario.

4. The physical object determination method according to claim 1, wherein, The target data objects associated with the business scenario are obtained based on the data objects under the object category associated with the business scenario, which comprises: Based on the object category associated with the business scenario and the category attribute of the data object represented in the cognitive graph, the candidate data objects under the object category associated with the business scenario are determined; The similarity between the business scenario and the candidate data object is determined; According to the similarity between the business scenario and the candidate data object, the target data object associated with the business scenario is selected from the candidate data objects.

5. The physical object determination method according to claim 4, wherein, The similarity between the business scenario and the candidate data object is determined, which comprises: The business scenario vector corresponding to the business scenario and the data object vector corresponding to the candidate data object are obtained; The similarity between the business scenario vector and the data object vector is calculated to obtain the similarity between the business scenario and the candidate data object.

6. The solid object determination method according to claim 5, wherein, According to the similarity between the business scenario and the candidate data object, the target data object associated with the business scenario is selected from the candidate data objects, which comprises: Based on the similarity between the business scenario and the candidate data object, the similarity threshold value corresponding to the business scenario is determined; determine, based on the similarity threshold, a target data object associated with the business scenario from the candidate data objects, wherein a similarity of the target data object to the business scenario is not less than the similarity threshold.

7. The solid object determination method according to claim 6, wherein, The determining of the similarity threshold corresponding to the business scenario based on the similarity between the business scenario and the candidate data objects comprises: ranking the candidate data objects according to the similarity, determining a sequence position matching at least a first proportion of the values from the ranking, and taking the similarity of the candidate data object corresponding to the sequence position as the similarity threshold corresponding to the business scenario.

8. The physical object determination method according to claim 7, wherein, The first proportion matches a first proportion of the number of data objects associated with the business scenario in the cognitive graph.

9. The solid object determination method according to claim 5, wherein, Further comprising: setting each business scenario and each data object as a node based on the association relationship between the plurality of business scenarios and the data objects expressed by the cognitive graph to construct an association graph of the business scenarios and the data objects; performing random walk on the association graph with any node as a starting point, sampling the path of the random walk to obtain a sampling document, wherein one sampling document includes a node passed by a path, and a plurality of sampling documents form a sampling document set; performing vectorization training on each business scenario and each data object based on the sampling document set; after the training converges, obtaining a business scenario vector corresponding to each business scenario and a data object vector corresponding to each data object.

10. The solid object determination method according to claim 9, wherein, The setting of each business scenario and each data object as a node based on the association relationship between the plurality of business scenarios and the data objects expressed by the cognitive graph to construct an association graph of the business scenarios and the data objects comprises: connecting the associated top nodes and sub-nodes based on the association relationship between the business scenarios and the data objects expressed by the cognitive graph to obtain the association graph.

11. The solid object determination method of claim 9, wherein, The vectorization training on each business scenario and each data object based on the sampling document set comprises: performing vectorization training on each business scenario and each data object based on the sampling document set using a skip-gram model algorithm.

12. An entity object determination apparatus, wherein, Comprising: a target data object acquisition module configured to acquire a target data object associated with a business scenario based on data objects under an object category associated with the business scenario; a relevance determination module configured to determine the relevance of an entity object to the business scenario according to historical interaction behavior data of the entity object and the target data object, wherein the target data object is a plurality of target data objects, and the target data object acquisition module is configured to determine the relevance of the entity object to each target data object based on behavior data of the entity object with respect to each target data object within a historical time period, and determine the relevance of the entity object to the business scenario according to the relevance of the entity object to each target data object A group determination module is configured to determine, as an entity object associated with the business scenario, an entity object having a relevance to the business scenario that is not less than a preset relevance threshold, and the entity objects associated with the business scenario form an entity object group of the business scenario.

13. A server, wherein, Comprise: at least one memory and at least one processor; The memory stores one or more computer-executable instructions, and the processor invokes the one or more computer-executable instructions to perform the entity object determination method according to any one of claims 1-11.

14. A storage medium, wherein, The storage medium stores one or more computer-executable instructions, and the one or more computer-executable instructions are used to perform the entity object determination method according to any one of claims 1-11.

Images