Object feature construction method, device and computer readable storage medium

Abstract

Embodiments of the present application disclose an object feature construction method and device and a computer readable storage medium, which can be applied to various scenes such as cloud technology, artificial intelligence, intelligent transportation and auxiliary driving. The object label information of a to-be-constructed object in a target system is obtained. The label similarity between object labels is calculated, and the basic object label is determined in the object label information based on the label similarity and the label weight. Semantic propagation is performed in a preset label relationship graph based on the semantic information of the basic object label, and the associated object label and the label association path are obtained. The number of nodes of the label association path is counted, and the label weight of the basic object label corresponding to the associated object label is adjusted according to the number of nodes to obtain the associated label weight. The object feature corresponding to the to-be-constructed object is constructed based on the associated object label and the associated label weight. In this way, the accuracy of object feature construction is improved, and the efficiency of object feature construction is improved.

Description

Technical Field

[0001] This application relates to the field of Internet technology, specifically to a method, apparatus, and computer-readable storage medium for constructing object features. Background Technology

[0002] In recent years, with the rapid development of Internet technology, a large amount of information has been generated online, satisfying users' information needs in the information age. At the same time, the large amount of information also makes it difficult for users to quickly obtain the information they really need, reducing the efficiency of information acquisition. To address this, existing technologies use tagging of user interests and preferences, constructing object characteristics based on these tags, and then pushing content that users are interested in to users based on these object characteristics, thereby improving the accuracy of application business such as content push, business analysis, and operations.

[0003] Research and practice on existing technologies have revealed that in existing object feature construction methods, for inactive users or new users, the limited number of historical behaviors results in a smaller number of object feature labels, making it impossible to accurately characterize user preferences. Consequently, the accuracy of object feature construction is low, leading to low efficiency in object feature construction. Summary of the Invention

[0004] This application provides an object feature construction method, apparatus, and computer-readable storage medium, which can improve the accuracy of object feature construction and thus improve the efficiency of object feature construction.

[0005] This application provides an object feature construction method, including:

[0006] Obtain object tag information of the object to be constructed in the target system. The object tag information includes at least one object tag for an object attribute of the object to be constructed and a tag weight corresponding to the object tag. The tag weight is used to characterize the importance of the object attribute to the object to be constructed.

[0007] Calculate the tag similarity between the object tags, and determine the basic object tags in the object tag information based on the tag similarity and tag weights;

[0008] Extract the semantic information corresponding to the basic object label, and based on the semantic information, perform semantic propagation in the preset label relationship graph to obtain at least one associated object label and the label association path corresponding to the associated object label. The preset label relationship graph includes at least one node and node edges connected based on the semantic relationship between the nodes. Each node corresponds to a preset object label.

[0009] The number of nodes in the associated path of the tag is counted, and the tag weight of the base object tag corresponding to the associated object tag is adjusted according to the number of nodes to obtain the associated tag weight corresponding to the associated object tag.

[0010] The object tag information is updated based on the associated object tag and the associated tag weight corresponding to the associated object tag, and the object features corresponding to the object to be constructed are constructed based on the updated object tag information.

[0011] Accordingly, embodiments of this application provide an object feature construction apparatus, including:

[0012] The acquisition unit is used to acquire object tag information of the object to be constructed in the target system. The object tag information includes at least one object tag for an object attribute of the object to be constructed and a tag weight corresponding to the object tag. The tag weight is used to characterize the importance of the object attribute to the object to be constructed.

[0013] A determining unit is used to calculate the tag similarity between the object tags, and determine the basic object tags in the object tag information based on the tag similarity and tag weights;

[0014] The propagation unit is used to extract the semantic information corresponding to the basic object tag, and based on the semantic information, perform semantic propagation in the preset tag relationship graph to obtain at least one associated object tag and the tag association path corresponding to the associated object tag. The preset tag relationship graph includes at least one node and node edges connected based on the semantic relationship between the nodes. Each node corresponds to a preset object tag.

[0015] An adjustment unit is used to count the number of nodes in the associated path of the tag, and adjust the tag weight of the base object tag corresponding to the associated object tag according to the number of nodes, so as to obtain the associated tag weight corresponding to the associated object tag.

[0016] The construction unit is used to update the object tag information based on the associated object tag and the associated tag weight corresponding to the associated object tag, and to construct the object features corresponding to the object to be constructed based on the updated object tag information.

[0017] In one embodiment, the propagation unit includes:

[0018] The semantic propagation subunit is used to perform semantic propagation in a preset label relationship graph based on the semantic information to obtain at least one candidate associated object label and the candidate label association path corresponding to the candidate associated object label.

[0019] The correlation calculation subunit is used to count the number of tags of the candidate related object tags, and when the number of tags is greater than a preset tag number threshold, calculate the correlation between the candidate related object tags and the basic object tags;

[0020] The tag filtering subunit is used to filter at least one associated object tag from the candidate associated object tags based on the relevance, and to obtain the tag association path corresponding to the associated object tag.

[0021] In one embodiment, the determining unit includes:

[0022] An object tag group identification subunit is used to identify at least one object tag group in the object tags based on the tag similarity, wherein the object tag group includes at least two object tags;

[0023] The basic object tag filtering subunit is used to filter out at least one basic object tag from the object tag group based on the tag weight.

[0024] In one embodiment, the object tag group identification subunit includes:

[0025] The threshold acquisition module is used to acquire a first similarity threshold and a second similarity threshold, wherein the first similarity threshold is greater than the second similarity threshold;

[0026] The tag group filtering module is used to filter at least one first object tag group from the object tags based on the first similarity threshold and tag similarity, and to filter at least one second object tag group from the object tags based on the second similarity threshold and tag similarity.

[0027] The assignment module is used to treat the first object tag group and the second object tag group as object tag groups.

[0028] In one embodiment, the acquisition unit includes:

[0029] The object tag information quantity statistics subunit is used to count the number of target object tag information corresponding to the target object in the target system, and obtain the number of object tag information corresponding to the target object.

[0030] The object identification subunit is used to identify the object to be constructed in the target object based on the number of object tag information.

[0031] The object tag information acquisition subunit is used to acquire the object tag information of the object to be constructed in the target system.

[0032] In one embodiment, the object tag information quantity counting subunit includes:

[0033] The device attribute information acquisition module is used to acquire the device attribute information corresponding to the target object;

[0034] The device update object tag information determination module is used to identify at least one device update object in the target object based on the device attribute information, and determine the device update object tag information of the device update object according to the device attribute information;

[0035] The object tag information quantity statistics module is used to update the target object tag information based on the updated object tag information of the device, obtain the updated target object tag information, and count the number of updated target object tag information corresponding to the target object to obtain the object tag information quantity corresponding to the target object.

[0036] In one embodiment, the device update object tag information determination module includes:

[0037] The device tag information acquisition submodule is used to acquire the current device object tag information and historical device object tag information corresponding to the device update object based on the device attribute information;

[0038] The tag information fusion submodule is used to fuse the current device object tag information and the historical device object tag information to obtain the device update object tag information corresponding to the device update object.

[0039] In one embodiment, the object feature construction apparatus further includes:

[0040] The content to be pushed acquisition unit is used to acquire at least one piece of content to be pushed.

[0041] The target content recognition unit is used to identify at least one target content that matches the object features in the content to be pushed, based on the object features corresponding to the object to be constructed.

[0042] The content push unit is used to push the target content to the object to be constructed.

[0043] Furthermore, embodiments of this application also provide a computer-readable storage medium storing a plurality of instructions adapted for loading by a processor to execute steps in any of the object feature construction methods provided in embodiments of this application.

[0044] Furthermore, this application also provides a computer device, including a processor and a memory, wherein the memory stores an application program, and the processor is used to run the application program in the memory to implement the object feature construction method provided in this application.

[0045] This application also provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps in the object feature construction method provided in this application.

[0046] This application embodiment obtains object tag information of the object to be constructed in the target system. The object tag information includes at least one object tag for an object attribute of the object to be constructed and a tag weight corresponding to the object tag. The tag weight is used to characterize the importance of the object attribute to the object to be constructed. The tag similarity between the object tags is calculated, and based on the tag similarity and tag weight, a basic object tag is determined in the object tag information. Semantic information corresponding to the basic object tag is extracted, and based on the semantic information, semantic propagation is performed in a preset tag relationship graph to obtain at least one associated object tag and a tag association path corresponding to the associated object tag. The preset tag relationship graph includes at least one node and node edges connected based on the semantic relationship between nodes. Each node corresponds to a preset object tag. The number of nodes in the tag association path is counted, and the tag weight of the basic object tag corresponding to the associated object tag is adjusted according to the number of nodes to obtain the associated tag weight corresponding to the associated object tag. The object tag information is updated based on the associated object tag and the associated tag weight corresponding to the associated object tag, and the object features corresponding to the object to be constructed are constructed based on the updated object tag information. In this way, by determining the basic object tag in the object tag information and performing semantic propagation in the preset tag relationship graph based on the semantic information of the basic object tag, at least one associated object tag and the tag association path corresponding to the associated object tag are obtained. Then, the number of nodes in the tag association path is counted, and the tag weight of the basic object tag corresponding to the associated object tag is adjusted according to the number of nodes to obtain the associated tag weight corresponding to the associated object tag. Thus, based on the associated object tag and the associated tag weight corresponding to the associated object tag, the object tag information of the object to be recommended is expanded, enriching the number of object tags that the object to be built may be interested in. Then, based on the expanded object tag information, the object features corresponding to the object to be built are constructed, improving the accuracy of object feature construction and thus improving the efficiency of object feature construction. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a schematic diagram illustrating an implementation scenario of an object feature construction method provided in this application embodiment;

[0049] Figure 2 This is a flowchart illustrating an object feature construction method provided in an embodiment of this application;

[0050] Figure 3 This is a schematic diagram of the overall framework of an object feature construction method provided in an embodiment of this application;

[0051] Figure 4 This is a schematic diagram illustrating the process of obtaining device update object tag information in an object feature construction method provided in this application embodiment;

[0052] Figure 5a This is a schematic diagram of the feature extraction framework of an object feature construction method provided in an embodiment of this application;

[0053] Figure 5b This is a schematic diagram of the object tag expansion process of an object feature construction method provided in an embodiment of this application;

[0054] Figure 6 This is another flowchart illustrating an object feature construction method provided in an embodiment of this application;

[0055] Figure 7 This is a schematic diagram of the structure of the object feature construction device provided in the embodiments of this application;

[0056] Figure 8 This is a schematic diagram of the structure of the computer device provided in the embodiments of this application. Detailed Implementation

[0057] 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.

[0058] This application provides an object feature construction method, apparatus, and computer-readable storage medium. The object feature construction apparatus can be integrated into a computer device, which may be a server or a terminal, etc.

[0059] The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery network (CDN) acceleration services, and big data and artificial intelligence platforms. The terminal can include, but is not limited to, mobile phones, computers, smart voice interaction devices, smart home appliances, vehicle terminals, and aircraft. The terminal and server can be directly or indirectly connected via wired or wireless communication, which is not limited herein.

[0060] Please see Figure 1 Taking the integration of object feature construction devices into computer equipment as an example, Figure 1 This is a schematic diagram illustrating an implementation scenario of the object feature construction method provided in this application. The computer device can be a server or a terminal. The computer device can acquire object tag information of the object to be constructed in the target system. This object tag information includes at least one object tag for an object attribute of the object to be constructed and a tag weight corresponding to that object tag. The tag weight is used to characterize the importance of the object attribute to the object to be constructed. The device calculates the tag similarity between the object tags and determines basic object tags in the object tag information based on the tag similarity and tag weight. It extracts semantic information corresponding to the basic object tags and performs semantic propagation in a preset tag relationship graph based on the semantic information to obtain at least one associated object tag and a tag association path corresponding to the associated object tag. The preset tag relationship graph includes at least one node and node edges connected based on the semantic relationship between nodes. Each node corresponds to a preset object tag. The device counts the number of nodes in the tag association path and adjusts the tag weight of the basic object tag corresponding to the associated object tag according to the number of nodes to obtain the associated tag weight corresponding to the associated object tag. The device updates the object tag information based on the associated object tag and the associated tag weight corresponding to the associated object tag, and constructs the object features corresponding to the object to be constructed based on the updated object tag information.

[0061] It should be noted that the embodiments of the present invention can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, smart transportation, and assisted driving. Figure 1The illustrated scenario of the object feature construction method is merely an example. The implementation environment scenario of the object feature construction method described in this application is for the purpose of more clearly illustrating the technical solution of this application and does not constitute a limitation on the technical solution provided in this application. Those skilled in the art will understand that with the evolution of object feature construction and the emergence of new business scenarios, the technical solution provided in this application is also applicable to similar technical problems.

[0062] The solutions provided in this application are specifically illustrated through the following embodiments. It should be noted that the order of description of the following embodiments is not intended to limit the preferred order of the embodiments.

[0063] This embodiment will be described from the perspective of an object feature construction device, which can be integrated into a computer device, which can be a server, and this application does not limit it.

[0064] Please see Figure 2 , Figure 2 This is a flowchart illustrating the object feature construction method provided in an embodiment of this application. The object feature construction method includes:

[0065] In step 101, the object tag information of the object to be constructed in the target system is obtained.

[0066] The object to be constructed can be an object whose features are to be built. These features can be characteristic information representing the object, such as its interests, preferences, behaviors, and habits. The object can be a user, and the target system can be a system for constructing object tag information. Optionally, the object features can be tagged with information such as the object's interests and preferences, using a set of <tag, weight> pairs to characterize the object's interests. That is, object tag information can be used to construct object features. This object tag information can include at least one object tag for the object's attributes and the corresponding tag weight. The object attribute can be information representing the nature of the object and its relationship with other objects to be constructed, such as its interests, preferences, behaviors, and habits. The object tag can be information indicating the category or content of the object's attributes. For example, if the object to be constructed is a user, and the user has the object tag "crosstalk," it can indicate that the user is interested in crosstalk content. The label weight can be used to characterize the importance of an object attribute to the object to be constructed. For example, it can characterize the object's preference for a certain object label, or its interest in a certain object label. The label weight can be normalized to a number between 1 and 100 to characterize the importance of the object attribute to the object to be constructed, or it can be normalized to a number between 0 and 1. A larger number indicates greater importance. The specific setting can be determined according to actual needs, and this application does not impose any limitations. For example, suppose an object to be constructed has object label information <crosstalk, 40> and <ice cream, 80> in the target system. The label weight corresponding to the object label "crosstalk" is 40, and the label weight corresponding to the object label "ice cream" is 80, indicating that the object to be constructed is more interested in ice cream than crosstalk.

[0067] In existing object feature construction methods based on user viewing history and playback behavior, for inactive or new users, the limited number of historical behaviors results in a small number of object tags for constructed object features. This makes it difficult to accurately characterize user interests and preferences, leading to low accuracy and efficiency in object feature construction. To address this, objects with fewer object tags in the object set can have their tags expanded to obtain more tags for accurately characterizing user interests and preferences, thus improving the accuracy of object feature construction. In this case, the object to be constructed can be an object with fewer tags in the object set, which can be a whole consisting of at least one object. Optionally, there are various ways to obtain the object tag information of the object to be constructed in the target system. For example, the number of target object tags corresponding to the target object in the target system can be counted to obtain the number of object tags corresponding to the target object. Based on this number of object tags, the object to be constructed can be identified within the target object, and its object tag information in the target system can be obtained.

[0068] The target object can be an object that constructs object characteristics, such as all users of an application. The object to be constructed can be an object within the target object that requires object tag expansion, such as an object within the target object with a relatively small number of object tag information. The number of object tag information can be the number of object tag information present in each target object.

[0069] In one embodiment, object tag information corresponding to the same target object in different devices can be fused to increase the number of object tag information corresponding to the target object. There are various ways to count the number of target object tag information corresponding to the target object in the target system, for example, please refer to... Figure 3 , Figure 3 This is a schematic diagram of the overall framework of an object feature construction method provided in this application embodiment. It can obtain the device attribute information corresponding to the target object, identify at least one device update object in the target object based on the device attribute information, determine the device update object tag information of the device update object according to the device attribute information, update the target object tag information based on the device update object tag information, obtain the updated target object tag information, and count the number of updated target object tag information corresponding to the target object to obtain the number of object tag information corresponding to the target object.

[0070] The device attribute information can be the attribute information of the device corresponding to the target object. For example, it can be the attribute information of the device currently used by the target object and the devices used in the past. The device can be a computer device such as a mobile phone or computer. For example, it can be the mobile phone used by the user. The attribute information can include the activation time of the device, usage records, etc. The usage time of the device corresponding to the target object can be determined based on the activation time and usage records of the device corresponding to the target object. The device update object can be a target object whose current device usage time is within a preset time period. For example, it can be a target object that has recently replaced its device with a new device. The preset time period can be a time interval pre-set according to actual needs. When the usage time of the device currently used by the target object is within this time interval, it can be indicated that the device currently corresponding to the target object is a new device, and the target object can be identified as the device update object.

[0071] There are several ways to identify at least one device update object within the target object based on the device attribute information. For example, the device usage time corresponding to the target object can be extracted from the device attribute information, compared with a preset time period, and the target object whose device usage time falls within the preset time period is identified as the device update object based on the comparison result. The device usage time can be the usage time of the device corresponding to the target object, such as the time when the user replaced the device.

[0072] Optional, please refer to Figure 4 , Figure 4 This is a schematic diagram illustrating the process of obtaining device update object tag information in an object feature construction method provided in this application embodiment. It can acquire target objects that have logged in using a new device within the last d days, where d can be a pre-set value, such as 30, or more, depending on actual needs. Then, natural person identification can be used to determine whether a target object that has logged in using a new device within the last d days is a device update object. Natural person identification determines whether users using different devices correspond to the same natural person. For example, a relationship graph of "device-account-device" triplet relationships can be established based on the login relationship between accounts and devices. Strategies can be used to remove abnormal nodes and prune unreasonable edges. The weights of "device-device" edges can be scored and sorted. Filtering can be performed based on the weights of "device-device" edges to select those with high credibility. If a device exists in multiple "device-device" edges, the edge with the highest weight can be selected, thus identifying at least one device update object among the target objects.

[0073] After identifying at least one device update object within the target object based on the device attribute information, the device update object tag information can be determined according to the device attribute information. There are several ways to determine the device update object tag information based on the device attribute information. For example, the current device object tag information and historical device object tag information corresponding to the device update object can be obtained based on the device attribute information, and then the current device object tag information and historical device object tag information can be merged to obtain the device update object tag information corresponding to the device update object.

[0074] The current device object tag information can be the object tag information corresponding to the target object based on the currently used device. For example, it can be the object tag information built based on the user's currently used device. The historical device object tag information can be the object tag information corresponding to the target object based on the historically used devices. For example, it can be the object tag information built based on the historically used devices. The device update object tag information can be the object tag information obtained by fusing the current device object tag information and the historical device object tag information corresponding to the device update object.

[0075] There are several ways to fuse the current device object tag information and the historical device object tag information. For example, object tags can be extracted from both the current and historical device object tag information, and duplicate object tags can be removed. For instance, the <tag, weight> list after fusing the current and historical device object tag information can be deduplicated. If duplicate object tags exist in the fused <tag, weight> list, the weight values ​​corresponding to the duplicate object tags are summed, and the sum is used as the new weight for that object tag. For example, for object tag ti, the list might contain duplicate tags.<ti,wi1> and<ti,wi2> Where wi1 and wi2 are the label weights of object label ti in the current device object label information and the historical device object label information, respectively, then the label weights can be removed.<ti,wi1> and<ti,wi2> and added<ti,wi1+wi2> .

[0076] After counting the number of target object tags corresponding to the target object in the target system, and obtaining the number of object tags for that target object, the object to be constructed can be identified within the target object based on this number of object tags. There are several ways to identify the object to be constructed within the target object based on the number of object tags. For example, a quantity threshold can be obtained, and this threshold can be compared with the number of object tags for the target object. Based on the comparison result, target objects with fewer object tags than the quantity threshold are identified as objects to be constructed. This quantity threshold can be a pre-set critical value. When the number of object tags is not less than the critical value, it indicates that the number of object tags for the target object is sufficient, and the target object can be determined not to be a target object. When the number of object tags is less than the critical value, it indicates that the number of object tags for the target object is insufficient, and the target object can be determined as a target object to be constructed. The specific value of this quantity threshold can be selected according to the actual situation and is not limited here. For example, the quantity threshold can be 10, 20, etc.

[0077] In step 102, the label similarity between object labels is calculated, and the basic object labels are determined in the object label information based on the label similarity and label weight.

[0078] The label similarity can characterize the degree of similarity between object labels. The basic object label can be the object label determined in the object label information based on label similarity and label weight, and can be used to characterize the object label in the object label information.

[0079] There are several ways to calculate the label similarity between object labels. For example, you can extract features from the object labels to obtain the object label features corresponding to the object labels, and then calculate the similarity between the object label features to obtain the label similarity between the object labels.

[0080] The object label feature can be characteristic information representing the object label, and can be a word embedding vector. There are various ways to extract the object label features, such as using language models from the natural language processing field, including word2vec, glove, ELMo, and Bidirectional Encoder Representation from Transformers (BERT). For example, BERT can be used to obtain the object label features corresponding to the object label. Alternatively, please refer to [reference needed]. Figure 5a , Figure 5a This is a schematic diagram of the feature extraction framework for an object feature construction method provided in this application embodiment. Taking the object feature construction method provided in this application embodiment as an example in the field of video content recommendation, the pre-trained model BERT can be fine-tuned using video domain corpus to enable the fine-tuned BERT model to learn better embedding vector representations for object labels in the video domain. In this way, the object labels of the objects to be constructed can be input into the fine-tuned BERT model to obtain the corresponding object label features.

[0081] After extracting features from object labels to obtain their corresponding features, the similarity between these features can be calculated to obtain the label similarity between object labels. There are several ways to calculate the similarity between object label features, such as using Euclidean distance or cosine distance. For example, the similarity between object label features can be obtained by calculating the cosine distance, where the cosine value of the angle between two object label features Ei and Ej in the vector space is used as the magnitude of the difference between the two labels. The cosine value can range from [-1, 1]. A cosine value closer to 1 and an angle closer to 0 indicate greater similarity between the two object labels. The formula for calculating the cosine distance can be expressed as follows:

[0082]

[0083] After calculating the tag similarity between object tags, basic object tags can be determined from the object tag information based on tag similarity and tag weight. There are several ways to determine basic object tags from the object tag information based on tag similarity and tag weight. For example, at least one object tag group can be identified in the object tag information based on the tag similarity, and at least one basic object tag can be selected from the object tag group based on the tag weight.

[0084] The object tag group may include at least two object tags.

[0085] There are several ways to identify at least one object label group in the object label based on the label similarity. For example, a first similarity threshold and a second similarity threshold can be obtained. Based on the first similarity threshold and the label similarity, at least one first object label group can be selected in the object label. Based on the second similarity threshold and the label similarity, at least one second object label group can be selected in the object label. The first object label group and the second object label group are then used as the object label group.

[0086] To expand to more representative labels, object labels with low label weights can be removed from object label groups with high label similarity, while object labels with high label weights in the same object label groups can be retained. Simultaneously, each label in the object label groups with low label similarity can be retained to effectively expand the object labels. In this case, the first similarity threshold can be greater than the second similarity threshold. Both the first and second similarity thresholds can be pre-set thresholds. When the label similarity between object labels is greater than the first similarity threshold, object labels with similarity greater than the first similarity threshold can be combined into the first object label group; when the label similarity between object labels is less than the second similarity threshold, object labels with similarity less than the second similarity threshold can be combined into the second object label group. Thus, based on at least one first object label group and one second object label group, an object label group can be obtained.

[0087] After identifying at least one object tag group within the object tag based on the tag similarity, at least one basic object tag can be selected from the object tag group according to the tag weight. There are several ways to select at least one basic object tag from the tag group based on the tag weight. For example, the tag weights corresponding to the object tags in the first object tag group can be compared, and the target object tag can be selected from the first object tag pair based on the comparison result. The basic object tag is then determined based on the target object tag and the object tags in the second object tag group.

[0088] The target object label can be the object label with a higher label weight in the first object label group. Thus, the target object label with a higher label weight in the first object label group and the object label in the second object label group can be determined as the basic object label.

[0089] In step 103, semantic information corresponding to the basic object tags is extracted, and semantic propagation is performed in the preset tag relationship graph based on the semantic information to obtain at least one associated object tag and the tag association path corresponding to the associated object tag.

[0090] The semantic information can be information representing the semantics of the basic object label. Semantics can represent the interpretation of data symbols in the basic object label. The preset label relationship graph can be a node relationship graph pre-constructed based on preset object labels. The preset label relationship graph can include at least one node and node edges connected based on the semantic relationships between nodes. Each node can correspond to a preset object label, which can be a pre-acquired object label. The node edges can be edges connecting nodes corresponding to preset object labels based on the semantic relationships between preset object labels. Optionally, the preset label relationship graph can be a knowledge graph, for example, a knowledge graph based on preset object labels as entities. The associated object label can be at least one preset object label determined by semantic propagation of the semantic information of the basic object label in the preset label relationship graph and associated with the basic object label. The label association path can be the semantic propagation path corresponding to the associated object label in the preset label relationship graph. For example, please refer to... Figure 5b , Figure 5b This is a schematic diagram of the object tag extension process of an object feature construction method provided in this application embodiment. The associated object tag can be the preset object tag determined in the preset tag relationship graph based on the tag semantic propagation extension.

[0091] There are several ways to obtain at least one associated object label and its corresponding label association path by performing semantic propagation in a preset label relationship graph based on semantic information. For example, based on the semantic information, semantic propagation can be performed in the preset label relationship graph to obtain at least one candidate associated object label and its corresponding candidate label association path. The number of labels of the candidate associated object label can be counted, and when the number of labels is greater than a preset label number threshold, the correlation degree between the candidate associated object label and the basic object label can be calculated. Based on the correlation degree, at least one associated object label can be selected from the candidate associated object label, and the label association path corresponding to the associated object label can be obtained.

[0092] The candidate associated object label can be at least one preset object label associated with the base object label, obtained by semantic propagation of semantic information based on the base object label in a preset label relationship graph. The candidate label association path can be the semantic propagation path corresponding to the candidate associated object label in the preset label relationship graph. The label quantity can be the number of candidate associated object labels. The preset label quantity threshold can be a pre-defined threshold value. When the label quantity is not greater than the threshold value, it indicates that the number of candidate associated object labels is small and no further filtering is needed. When the label quantity is greater than the threshold value, it indicates that the number of candidate associated object labels is large, and further filtering can be performed to improve the accuracy of the candidate associated object labels and reduce unnecessary calculations. The association degree can be information characterizing the degree of association between the candidate associated object label and its corresponding base object label; for example, it can be the similarity between the candidate associated object label and its corresponding base object label.

[0093] There are several ways to obtain at least one candidate associated object label and the candidate label association path corresponding to the candidate associated object label by performing semantic propagation in the preset label relationship graph based on the semantic information. For example, at least one concept information and at least one category information corresponding to the basic object label can be extracted from the semantic information corresponding to the basic object label. Based on the relationship between the concept information and the category information, at least one candidate associated object label and the candidate label association path corresponding to the candidate associated object label can be identified in the preset label relationship graph.

[0094] The concept information can include information about the concept corresponding to the basic object label, and the category information can include information about the category corresponding to the basic object label. For example, assuming the basic object label is "Guo ××" (a crosstalk actor), then we can obtain the concept information "crosstalk" and the category information "language drama" corresponding to the basic object label.

[0095] There are several ways to identify at least one candidate associated object label and its corresponding candidate label association path in a preset label relationship graph based on the relationship between the concept information and the category information. For example, the preset label relationship graph can be a knowledge graph, which can be used to propagate the semantic information of the basic object label to expand at least one candidate associated object label based on the degree of semantic association between the semantic information corresponding to the candidate associated object label and the semantic information corresponding to the preset object label. Specifically, a knowledge graph with labels as entities can be constructed in advance based on the preset object label and its semantic information. For example, a knowledge graph with preset object labels as entities can be constructed based on the three-layer semantic relationship of label entity -> concept -> category. This knowledge graph contains three layers of semantic relationships, such as "Guo ×× -> crosstalk -> language comedy". Therefore, basic object tags can be input into the knowledge graph. Based on the semantic information of the basic object tags, they can be propagated within the knowledge graph. This allows for the selection of at least one candidate associated object tag and its corresponding candidate tag association path from a preset set of object tags based on the propagation path. For example, taking the object feature construction method provided in this application embodiment as an example in the field of video content recommendation, basic object tags can be input into the entity knowledge graph of the video domain. The propagation and expansion of object tags can be carried out using the correlation between conceptual and category information in the semantic information of the basic object tags. Assuming that an object to be constructed has an object tag of "Guo ××", it can be propagated and traversed in the knowledge graph through the correlation between conceptual and category information. For example, a propagation path "Guo ×× → crosstalk → language comedy → sketch → Song ×× (a sketch actor)" can be used to expand a new object tag "Song ××". Propagation through semantic information in the knowledge graph can result in multiple similar paths, thereby generating multiple candidate associated object tags that the object to be constructed may be interested in, thus increasing the number of object tags and improving the accuracy of object feature construction.

[0096] There are several ways to filter out at least one associated object label based on the correlation degree. For example, a correlation degree threshold can be obtained. When the correlation degree is greater than the correlation degree threshold, the candidate associated object label corresponding to the correlation degree can be determined as the associated object label. The correlation degree threshold can be a pre-set critical value of the correlation degree. The selection of the critical value can be determined according to actual needs.

[0097] In step 104, the number of nodes in the tag association path is counted, and the tag weight of the base object tag corresponding to the associated object tag is adjusted according to the number of nodes to obtain the associated tag weight corresponding to the associated object tag.

[0098] The number of nodes can be the number of nodes in the tag association path, and the associated tag weight can be the tag weight corresponding to the associated object tag.

[0099] There are several ways to adjust the tag weight of the base object tag corresponding to the associated object tag based on the number of nodes to obtain the associated tag weight of the associated object tag. For example, the weight coefficient of the associated object tag can be determined based on the number of nodes in the tag association path of the associated object tag, and the tag weight of the base object tag corresponding to the associated object tag can be weighted according to the weight coefficient to obtain the associated tag weight of the associated object tag.

[0100] The weighting coefficient can be a factor used to adjust the label weight. This coefficient can be a value such as 0.8. For example, when the number of nodes in the tag association path corresponding to an associated object label is large, it indicates a significant difference between the associated object label and its corresponding base object label. This means the object to be constructed is less likely to be interested in the associated object label. In this case, a smaller weighting coefficient can be set to reduce the influence of the associated object label on the object features of the object to be constructed. Conversely, when the number of nodes in the tag association path corresponding to an associated object label is small, it indicates a smaller difference between the associated object label and its corresponding base object label. This means the object to be constructed is more likely to be interested in the associated object label. In this case, a larger weighting coefficient can be set to increase the influence of the associated object label on the object features of the object to be constructed, thereby improving the accuracy of object feature construction while ensuring the expansion of interest in the object to be constructed. It should be noted that the specific value of this weighting coefficient can be set according to actual needs and is not limited here.

[0101] There are several ways to determine the weight coefficient corresponding to an associated object tag based on the number of nodes in the tag association path. For example, weight coefficients matching the number of nodes can be selected from a preset set of weight coefficients and used as the weight coefficients corresponding to the associated object tag. This preset set of weight coefficients can be a pre-defined set that includes at least one candidate weight coefficient, which can correspond to the number of nodes. Therefore, weight coefficients matching the number of nodes in the tag association path corresponding to the associated object tag can be selected from the preset set of weight coefficients.

[0102] In step 105, the object label information is updated based on the associated object label and the associated label weights corresponding to the associated object label, and the object features corresponding to the object to be constructed are constructed based on the updated object label information.

[0103] The updated object label information can be object label information updated based on associated object labels and their corresponding associated label weights. The object features can be feature information representing the object to be constructed, such as the object's interests, preferences, behaviors, and habits. In this way, semantic propagation can be used to find associated object labels that the object to be constructed may be interested in. The basic object labels and the newly generated associated object labels can be fused to obtain the object features corresponding to the object to be constructed, thereby enriching the number of object labels that the object to be constructed may be interested in, thus improving the accuracy of object feature construction and improving the efficiency of object feature construction.

[0104] There are several ways to update object tag information based on associated object tags and associated tag weights. For example, associated object tag information can be determined based on the associated object tag and associated tag weights, and the object tag information can be updated according to the associated object tag information.

[0105] Among them, the associated object tag information can be the object tag information corresponding to the associated object tag, that is, the object tag information extended by semantic propagation based on the basic object tag.

[0106] There are several ways to update the object's tag information based on the associated object's tag information. For example, please refer to [link / reference]. Figure 5b This allows you to integrate the associated object tag information into the object tag information of the object to be constructed, thereby obtaining the updated object tag information.

[0107] In one embodiment, after constructing the object features corresponding to the object to be constructed according to the object feature construction method provided in this application embodiment, content push processing can be performed on the object to be constructed based on the object features. There are various ways to perform content push processing on the object to be constructed based on its object features. For example, at least one piece of content to be pushed can be obtained, and based on the object features corresponding to the object to be constructed, at least one target content matching the object features can be identified in the content to be pushed, and the target content can be pushed to the object to be constructed.

[0108] The content to be pushed can be any content intended to be pushed, serving as a carrier of information. It can be presented to users for consumption, allowing them to obtain relevant information. For example, the content can include, but is not limited to, video, audio, text, and images, such as videos, news, and other similar content. The target content can be at least one piece of content in the content to be pushed that matches the object's characteristics, based on the object's corresponding features.

[0109] In this way, by expanding the object tag information of the object to be constructed, the number of object tag information of the object to be constructed is increased. Based on the expanded object tag information, object features are constructed. For the object to be constructed with few or no object tags of the original object features, more content that matches the user's interests can be recommended, thereby improving the accuracy of object feature construction. In turn, based on the constructed object features, the content push processing of the object to be constructed can be more accurate, thus improving the accuracy of content push.

[0110] As described above, this application embodiment obtains object tag information of the object to be constructed in the target system. This object tag information includes at least one object tag for the object attribute of the object to be constructed and the tag weight corresponding to the object tag. The tag weight is used to characterize the importance of the object attribute to the object to be constructed. The tag similarity between the object tags is calculated, and based on the tag similarity and tag weight, a basic object tag is determined in the object tag information. Semantic information corresponding to the basic object tag is extracted, and based on the semantic information, semantic propagation is performed in a preset tag relationship graph to obtain at least one associated object tag and the tag association path corresponding to the associated object tag. The preset tag relationship graph includes at least one node and node edges connected based on the semantic relationship between nodes. Each node corresponds to a preset object tag. The number of nodes in the tag association path is counted, and the tag weight of the basic object tag corresponding to the associated object tag is adjusted according to the number of nodes to obtain the associated tag weight corresponding to the associated object tag. The object tag information is updated based on the associated object tag and the associated tag weight corresponding to the associated object tag, and the object features corresponding to the object to be constructed are constructed based on the updated object tag information. In this way, by determining the basic object tag in the object tag information and performing semantic propagation in the preset tag relationship graph based on the semantic information of the basic object tag, at least one associated object tag and the tag association path corresponding to the associated object tag are obtained. Then, the number of nodes in the tag association path is counted, and the tag weight of the basic object tag corresponding to the associated object tag is adjusted according to the number of nodes to obtain the associated tag weight corresponding to the associated object tag. Thus, based on the associated object tag and the associated tag weight corresponding to the associated object tag, the object tag information of the object to be recommended is expanded, enriching the number of object tags that the object to be built may be interested in. Then, based on the expanded object tag information, the object features corresponding to the object to be built are constructed, improving the accuracy of object feature construction and thus improving the efficiency of object feature construction.

[0111] Based on the method described in the above embodiments, the following examples will provide further detailed explanations.

[0112] In this embodiment, the object feature construction device will be specifically integrated into a computer device as an example for explanation. The object feature construction method is executed by a server, and its application in the content push field will be specifically described. It is understood that in the specific embodiments of this application, user information and other related data are involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0113] For a better description of the embodiments of this application, please refer to Figure 6 , Figure 6 Another flowchart illustrating the object feature construction method provided in this application embodiment is shown below.

[0114] In step 201, the server obtains the device attribute information corresponding to the target object, identifies at least one device update object in the target object based on the device attribute information, obtains the current device object tag information and historical device object tag information corresponding to the device update object according to the device attribute information, and merges the current device object tag information and historical device object tag information to obtain the device update object tag information corresponding to the device update object.

[0115] There are several ways for the server to identify at least one device update object within the target object based on the device attribute information. For example, the server can extract the device usage time corresponding to the target object from the device attribute information, compare the device usage time corresponding to the target object with a preset time period, and determine the target object whose device usage time is within the preset time period as the device update object based on the comparison result. The device usage time can be the usage time of the device corresponding to the target object, such as the time when the user replaced the new device.

[0116] Optional, please continue to refer to Figure 4 The system can identify target objects that have logged in using new devices within the last d days. Here, d can be a pre-defined value, such as 30, and can be set according to actual needs. The server can then use natural person identification to determine whether these target objects are device update objects. Natural person identification determines whether users using different devices correspond to the same natural person. For example, the server can establish a relationship graph of "device-account-device" triplet relationships based on account and device login relationships. It can use strategies to remove abnormal nodes and prune unreasonable edges, score and sort the weights of "device-device" edges, and apply threshold filtering to select high-confidence "device-device" edges. If a device exists in multiple "device-device" edges, the edge with the highest weight can be selected, thus identifying at least one device update object among the target objects.

[0117] There are several ways for the server to merge the current device object tag information and the historical device object tag information. For example, the server can extract object tags from the current and historical device object tag information and remove duplicate object tags. For instance, it can perform deduplication on the <tag, weight> list after merging the current and historical device object tag information. If duplicate object tags exist in the merged <tag, weight> list, the weight values ​​corresponding to the duplicate object tags are added together, and the result is used as the new weight for that object tag. For example, for the object tag ti, the list might contain duplicate tags.<ti,wi1> and<ti,wi2> Where wi1 and wi2 are the label weights of object label ti in the current device object label information and the historical device object label information, respectively, then the label weights can be removed.<ti,wi1> and<ti,wi2> and added<ti,wi1+wi2> .

[0118] In step 202, the server updates the target object tag information based on the device update object tag information, obtains the updated target object tag information, counts the number of updated target object tag information corresponding to the target object, obtains the number of object tag information corresponding to the target object, and identifies the object to be constructed in the target object based on the number of object tag information.

[0119] The server can identify the object to be built from the target object based on the number of object tag information in several ways. For example, the server can obtain a quantity threshold and compare this threshold with the number of object tag information of the target object. Based on the comparison result, target objects with fewer object tag information than the quantity threshold are identified as objects to be built. This quantity threshold can be a pre-set cutoff value. When the number of object tag information is not less than the cutoff value, it indicates that the number of object tag information of the target object is sufficient, and the target object can be determined not to be built. When the number of object tag information is less than the cutoff value, it indicates that the number of object tag information of the target object is insufficient, and the target object can be determined as an object to be built. The specific value of this quantity threshold can be selected according to the actual situation and is not limited here. For example, the quantity threshold can be 10, 20, etc.

[0120] In step 203, the server obtains the object tag information of the object to be constructed in the target system, calculates the tag similarity between the object tags, and obtains the first similarity threshold and the second similarity threshold.

[0121] There are several ways for the server to calculate the label similarity between object labels. For example, it can extract features from the object labels to obtain the object label features corresponding to the object labels, calculate the similarity between the object label features, and obtain the label similarity between the object labels.

[0122] The object label feature can be characteristic information representing the object label, and can be a word embedding vector. There are various ways to extract the object label features, such as using language models like word2vec, glove, ELMo, and BERT. For example, BERT can be used to obtain the object label features corresponding to the object label. Optional methods are available for reference. Figure 5a Taking the object feature construction method provided in this application as an example in the field of video content recommendation, the pre-trained BERT model can be fine-tuned using video domain corpus. This allows the fine-tuned BERT model to learn better embedding vector representations for object labels in the video domain. In this way, the object labels of the objects to be constructed can be input into the fine-tuned BERT model to obtain the corresponding object label features.

[0123] After the server extracts features from object labels and obtains the corresponding object label features, it can calculate the similarity between these features, thus obtaining the label similarity between object labels. There are several ways the server can calculate the similarity between object label features, such as using Euclidean distance or cosine distance. For example, the similarity between object label features can be obtained by calculating the cosine distance, where the cosine value of the angle between two object label features Ei and Ej in the vector space is used as the magnitude of the difference between the two labels. The cosine value can range from [-1, 1]. A cosine value closer to 1 and an angle closer to 0 indicate greater similarity between the two object labels. The formula for calculating the cosine distance can be expressed as follows:

[0124]

[0125] In step 204, the server filters at least one first object label group from the object labels based on the first similarity threshold and label similarity, and filters at least one second object label group from the object labels based on the second similarity threshold and label similarity. The first object label group and the second object label group are used as object label groups, and at least one basic object label is filtered from the object label group according to the label weight.

[0126] There are several ways for the server to select at least one basic object tag from the tag group based on the tag weight. For example, the server can compare the tag weights corresponding to the object tags in the first object tag group, and select the target object tag from the first object tag pair based on the comparison result. Then, based on the target object tag and the object tags in the second object tag group, the basic object tag is determined. The target object tag can be the object tag with the higher tag weight in the first object tag group. Thus, the server can determine the target object tag with the higher tag weight in the first object tag group and the object tag in the second object tag group as the basic object tag.

[0127] In step 205, the server extracts the semantic information corresponding to the basic object tag, and performs semantic propagation in the preset tag relationship graph based on the semantic information to obtain at least one candidate associated object tag and the candidate tag association path corresponding to the candidate associated object tag.

[0128] In this process, the server can perform semantic propagation in a preset label relationship graph based on the semantic information to obtain at least one candidate associated object label and the candidate label association path corresponding to the candidate associated object label in various ways. For example, the server can extract at least one concept information and at least one category information corresponding to the basic object label from the semantic information corresponding to the basic object label, and identify at least one candidate associated object label and the candidate label association path corresponding to the candidate associated object label in the preset label relationship graph based on the association relationship between the concept information and the category information.

[0129] The server can identify at least one candidate associated object label and its corresponding candidate label association path within a pre-defined label relationship graph based on the relationship between the concept information and category information. This identification can be achieved in several ways. For example, the pre-defined label relationship graph can be a knowledge graph. The server can use the knowledge graph to propagate the semantic information of the basic object label, expanding the identification of at least one candidate associated object label based on the semantic relationship between the semantic information corresponding to the candidate associated object label and the semantic information corresponding to the pre-defined object label. Specifically, a knowledge graph with labels as entities can be pre-constructed based on the pre-defined object label and its semantic information. For example, a knowledge graph with pre-defined object labels as entities can be constructed based on a three-layer semantic relationship of label entity -> concept -> category. This knowledge graph contains three layers of semantic relationships, such as "Guo ×× -> crosstalk -> language-based comedy". Therefore, the server can input basic object tags into the knowledge graph, and propagate them within the knowledge graph based on the semantic information of the basic object tags. This allows for the selection of at least one candidate associated object tag and its corresponding candidate tag association path from preset object tags based on the propagation path. For example, taking the object feature construction method provided in this application embodiment as an example in the field of video content recommendation, basic object tags can be input into the entity knowledge graph of the video domain. The propagation and expansion of object tags can be achieved by utilizing the correlation between conceptual and category information in the semantic information of the basic object tags. Assuming an object to be constructed has an object tag "Guo ××", it can be propagated and traversed in the knowledge graph through the correlation between conceptual and category information. For example, a propagation path "Guo ×× → crosstalk → language comedy → sketch → Song ×× (a sketch actor)" can be used to expand a new object tag "Song ××". By propagating semantic information within a knowledge graph, multiple similar paths can be created, thereby generating multiple candidate related object labels that the object to be constructed may be interested in. This increases the number of object labels and, consequently, improves the accuracy of object feature construction.

[0130] In step 206, the server counts the number of tags of the candidate associated object tag, and when the number of tags is greater than a preset tag number threshold, calculates the correlation degree between the candidate associated object tag and the basic object tag. Based on the correlation degree, at least one associated object tag is selected from the candidate associated object tag, and the tag association path corresponding to the associated object tag is obtained.

[0131] There are several ways for the server to calculate the correlation between the candidate associated object tag and the base object tag. For example, it can calculate the similarity between the candidate associated object tag and the base object tag, and determine the correlation between the candidate associated object tag and the base object tag based on the similarity.

[0132] There are several ways for the server to filter out at least one associated object label based on the relevance. For example, the server can obtain a relevance threshold. When the relevance is greater than the relevance threshold, the candidate associated object label corresponding to the relevance can be determined as an associated object label. The relevance threshold can be a pre-set critical value of relevance. The selection of the critical value can be determined according to actual needs.

[0133] In step 207, the server counts the number of nodes in the tag association path and adjusts the tag weight of the base object tag corresponding to the associated object tag according to the number of nodes to obtain the associated tag weight corresponding to the associated object tag.

[0134] In this process, the server adjusts the tag weight of the base object tag corresponding to the associated object tag based on the number of nodes. There are several ways to obtain the associated tag weight corresponding to the associated object tag. For example, the server can determine the weight coefficient corresponding to the associated object tag based on the number of nodes in the tag association path corresponding to the associated object tag, and then weight the tag weight of the base object tag corresponding to the associated object tag based on the weight coefficient to obtain the associated tag weight corresponding to the associated object tag.

[0135] The weighting coefficient can be a factor used to adjust the label weight. This coefficient can be a value such as 0.8. For example, when the number of nodes in the tag association path corresponding to an associated object label is large, it indicates a significant difference between the associated object label and its corresponding base object label. This means the object to be constructed is less likely to be interested in the associated object label. In this case, a smaller weighting coefficient can be set to reduce the influence of the associated object label on the object features of the object to be constructed. Conversely, when the number of nodes in the tag association path corresponding to an associated object label is small, it indicates a smaller difference between the associated object label and its corresponding base object label. This means the object to be constructed is more likely to be interested in the associated object label. In this case, a larger weighting coefficient can be set to increase the influence of the associated object label on the object features of the object to be constructed, thereby improving the accuracy of object feature construction while ensuring the expansion of interest in the object to be constructed. It should be noted that the specific value of this weighting coefficient can be set according to actual needs and is not limited here.

[0136] There are several ways for the server to determine the weight coefficient corresponding to the associated object tag based on the number of nodes in the tag association path corresponding to the associated object tag. For example, the server can filter out the weight coefficient that matches the number of nodes from a preset set of weight coefficients and use it as the weight coefficient corresponding to the associated object tag.

[0137] In step 208, the server updates the object tag information based on the associated object tag and the associated tag weights corresponding to the associated object tag, and constructs the object features corresponding to the object to be constructed based on the updated object tag information.

[0138] There are several ways for the server to update the object tag information based on the associated object tag and the associated tag weight corresponding to the associated object tag. For example, the associated object tag information can be determined based on the associated object tag and the associated tag weight, and the object tag information can be updated according to the associated object tag information.

[0139] There are several ways for the server to determine the associated object label information based on the associated object label and the associated label weight. For example, assuming that the associated object label T and the associated label weight W corresponding to the associated object label T are obtained by expanding the basic object label of the object to be constructed, the associated object label information can be obtained as follows:<T,W> .

[0140] After determining the associated object tag information based on the associated object tag and the associated tag weight, the server can update the object tag information according to this associated object tag information. There are several ways the server can update the object tag information based on the associated object tag information; for example, please refer to [link to relevant documentation]. Figure 5b The server can integrate the associated object tag information into the object tag information of the object to be constructed, thereby obtaining the updated object tag information.

[0141] In step 209, the server obtains at least one piece of content to be pushed, identifies at least one target content that matches the object characteristics in the content to be pushed based on the object characteristics corresponding to the object to be built, and pushes the target content to the object to be built.

[0142] The server expands the object tag information of the object to be constructed, increasing the number of object tags. Based on this expanded object tag information, object features are constructed, enriching the object's characteristics. Therefore, for objects with few or no object tags in their original features, more content matching the user's interests can be recommended, improving the accuracy of object feature construction. When the server receives a content push request for an object to be constructed, it can filter the acquired push content based on the object's characteristics to select target content that matches the object—content that the object might be interested in—and then push this target content to the object. This allows for more accurate content push processing based on the constructed object features, improving the accuracy of content push.

[0143] As described above, this application embodiment obtains device attribute information corresponding to the target object through a server. Based on the device attribute information, at least one device update object is identified in the target object. According to the device attribute information, the current device object tag information and historical device object tag information corresponding to the device update object are obtained. The current device object tag information and historical device object tag information are fused to obtain the device update object tag information corresponding to the device update object. Based on the device update object tag information, the server updates the target object tag information to obtain the updated target object tag information, and counts the updated target object tag information corresponding to the target object. The server obtains the number of object tag information for the target object, and identifies the object to be constructed within the target object based on this number of object tag information. The server acquires the object tag information of the object to be constructed within the target system, calculates the tag similarity between the object tags, and obtains a first similarity threshold and a second similarity threshold. Based on the first similarity threshold and tag similarity, the server selects at least one first object tag group from the object tags, and based on the second similarity threshold and tag similarity, selects at least one second object tag group from the object tags. The first object tag group and the second object tag group are then used as a pair. For example, in a tag group, based on the tag weight, at least one basic object tag is selected from the object tag group. The server extracts the semantic information corresponding to the basic object tag, and based on the semantic information, performs semantic propagation in a preset tag relationship graph to obtain at least one candidate associated object tag and the candidate tag association path corresponding to the candidate associated object tag. The server counts the number of tags for the candidate associated object tag, and when the number of tags is greater than a preset tag number threshold, calculates the correlation degree between the candidate associated object tag and the basic object tag. Based on the correlation degree, at least one associated object tag is selected from the candidate associated object tag, and the tag association path corresponding to the associated object tag is obtained. The server counts the number of nodes in the tag association path, and adjusts the tag weight of the basic object tag corresponding to the associated object tag according to the number of nodes to obtain the associated tag weight corresponding to the associated object tag. The server updates the object tag information based on the associated object tag and the associated tag weight corresponding to the associated object tag, and constructs the object features corresponding to the object to be constructed based on the updated object tag information. The server obtains at least one piece of content to be pushed, and based on the object features corresponding to the object to be constructed, identifies at least one target content that matches the object features in the content to be pushed, and pushes the target content to the object to be constructed.Therefore, by determining basic object tags in object tag information and performing semantic propagation in a preset tag relationship graph based on the semantic information of these basic object tags, at least one associated object tag and its corresponding tag association path are obtained. The number of nodes in the tag association path is then counted, and the tag weight of the basic object tag corresponding to the associated object tag is adjusted based on the number of nodes to obtain the associated tag weight. This expands the object tag information of the object to be recommended based on the associated object tag and its corresponding associated tag weight, enriching the number of object tags that the object might be interested in. Then, based on the expanded object tag information, object features corresponding to the object to be built are constructed, improving the accuracy and efficiency of object feature construction. Furthermore, for objects to be built with few or no object tags in their original object features, more content matching user interests can be recommended, thus improving the accuracy of object feature construction. This allows for more accurate content push processing of the object to be built based on the constructed object features, improving the accuracy of content push.

[0144] To better implement the above methods, embodiments of the present invention also provide an object feature construction apparatus, which can be integrated into a computer device, which can be a server.

[0145] For example, such as Figure 7 The diagram shown is a structural schematic of an object feature construction device provided in an embodiment of this application. The object feature construction device may include an acquisition unit 301, a determination unit 302, a propagation unit 303, an adjustment unit 304, and a construction unit 305, as follows:

[0146] The acquisition unit 301 is used to acquire object tag information of the object to be constructed in the target system. The object tag information includes at least one object tag for the object attribute of the object to be constructed and the tag weight corresponding to the object tag. The tag weight is used to characterize the importance of the object attribute to the object to be constructed.

[0147] The determining unit 302 is used to calculate the label similarity between the object labels and determine the basic object labels in the object label information based on the label similarity and label weight.

[0148] The propagation unit 303 is used to extract the semantic information corresponding to the basic object label, and based on the semantic information, perform semantic propagation in the preset label relationship graph to obtain at least one associated object label and the label association path corresponding to the associated object label. The preset label relationship graph includes at least one node and node edges connected based on the semantic relationship between the nodes. Each node corresponds to a preset object label.

[0149] The adjustment unit 304 is used to count the number of nodes in the associated path of the tag, and adjust the tag weight of the base object tag corresponding to the associated object tag according to the number of nodes, so as to obtain the associated tag weight corresponding to the associated object tag.

[0150] The construction unit 305 is used to update the object label information based on the associated object label and the associated label weight corresponding to the associated object label, and to construct the object features corresponding to the object to be constructed based on the updated object label information.

[0151] In one embodiment, the propagation unit 303 includes:

[0152] The semantic propagation subunit is used to perform semantic propagation in a preset label relationship graph based on the semantic information, so as to obtain at least one candidate associated object label and the candidate label association path corresponding to the candidate associated object label.

[0153] The correlation calculation subunit is used to count the number of tags of the candidate related object tag, and when the number of tags is greater than a preset tag number threshold, calculate the correlation between the candidate related object tag and the base object tag;

[0154] The tag filtering subunit is used to filter at least one related object tag from the candidate related object tags based on the relevance, and to obtain the tag association path corresponding to the related object tag.

[0155] In one embodiment, the determining unit 302 includes:

[0156] The object tag group identification subunit is used to identify at least one object tag group in the object tag based on the tag similarity, wherein the object tag group includes at least two object tags;

[0157] The basic object label filtering subunit is used to filter out at least one basic object label from the object label group based on the label weight.

[0158] In one embodiment, the object tag group identification subunit includes:

[0159] The threshold acquisition module is used to acquire a first similarity threshold and a second similarity threshold, wherein the first similarity threshold is greater than the second similarity threshold;

[0160] The tag group filtering module is used to filter at least one first object tag group from the object tags based on the first similarity threshold and tag similarity, and to filter at least one second object tag group from the object tags based on the second similarity threshold and tag similarity.

[0161] The assignment module is used to treat the first object tag group and the second object tag group as object tag groups.

[0162] In one embodiment, the acquisition unit 301 includes:

[0163] The object tag information quantity statistics subunit is used to count the number of target object tag information corresponding to the target object in the target system, and obtain the number of object tag information corresponding to the target object.

[0164] The object identification subunit is used to identify the object to be constructed in the target object based on the number of the object's label information.

[0165] The object tag information acquisition subunit is used to acquire the object tag information of the object to be built in the target system.

[0166] In one embodiment, the object tag information quantity counting subunit includes:

[0167] The device attribute information acquisition module is used to acquire the device attribute information corresponding to the target object;

[0168] The device update object tag information determination module is used to identify at least one device update object in the target object based on the device attribute information, and determine the device update object tag information of the device update object according to the device attribute information.

[0169] The object tag information quantity statistics module is used to update the object tag information based on the device, update the target object tag information, obtain the updated target object tag information, and count the number of updated target object tag information corresponding to the target object, thus obtaining the object tag information quantity corresponding to the target object.

[0170] In one embodiment, the device updates the object tag information determination module, including:

[0171] The device tag information acquisition submodule is used to obtain the current device object tag information and historical device object tag information corresponding to the updated device object based on the device attribute information;

[0172] The tag information fusion submodule is used to fuse the tag information of the current device object and the tag information of the historical device object to obtain the tag information of the device update object corresponding to the device update object.

[0173] In one embodiment, the object feature construction apparatus further includes:

[0174] The content to be pushed acquisition unit is used to acquire at least one piece of content to be pushed.

[0175] The target content recognition unit is used to identify at least one target content that matches the object feature in the content to be pushed, based on the object feature corresponding to the object to be constructed.

[0176] The content push unit is used to push the target content to the object to be built.

[0177] In practice, each of the above units can be implemented as an independent entity or can be arbitrarily combined to be implemented as the same or several entities. For the specific implementation of each of the above units, please refer to the previous method embodiments, which will not be repeated here.

[0178] As can be seen from the above, in this embodiment of the application, the acquisition unit 301 acquires object tag information of the object to be constructed in the target system. The object tag information includes at least one object tag for the object attribute of the object to be constructed and the tag weight corresponding to the object tag. The tag weight is used to characterize the importance of the object attribute to the object to be constructed. The determination unit 302 calculates the tag similarity between the object tags and determines the basic object tags in the object tag information based on the tag similarity and tag weight. The propagation unit 303 extracts the semantic information corresponding to the basic object tags and performs semantic propagation in the preset tag relationship graph based on the semantic information to obtain at least one The preset label relationship graph includes at least one node and node edges connected based on the semantic relationship between nodes, with each node corresponding to a preset object label. The adjustment unit 304 counts the number of nodes in the label relationship path and adjusts the label weight of the basic object label corresponding to the associated object label according to the number of nodes to obtain the associated label weight corresponding to the associated object label. The construction unit 305 updates the object label information based on the associated object label and the associated label weight corresponding to the associated object label, and constructs the object features corresponding to the object to be constructed based on the updated object label information. In this way, by determining the basic object tag in the object tag information and performing semantic propagation in the preset tag relationship graph based on the semantic information of the basic object tag, at least one associated object tag and the tag association path corresponding to the associated object tag are obtained. Then, the number of nodes in the tag association path is counted, and the tag weight of the basic object tag corresponding to the associated object tag is adjusted according to the number of nodes to obtain the associated tag weight corresponding to the associated object tag. Thus, based on the associated object tag and the associated tag weight corresponding to the associated object tag, the object tag information of the object to be recommended is expanded, enriching the number of object tags that the object to be built may be interested in. Then, based on the expanded object tag information, the object features corresponding to the object to be built are constructed, improving the accuracy of object feature construction and thus improving the efficiency of object feature construction.

[0179] This application also provides a computer device, such as... Figure 8 As shown, it illustrates a structural diagram of a computer device involved in an embodiment of this application. This computer device may be a server, specifically:

[0180] The computer device may include components such as a processor 401 with one or more processing cores, a memory 402 with one or more computer-readable storage media, a power supply 403, and an input unit 404. Those skilled in the art will understand that... Figure 8 The computer device structure shown does not constitute a limitation on the computer device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:

[0181] Processor 401 is the control center of the computer device, connecting various parts of the computer device through various interfaces and lines. It performs various functions and processes data by running or executing software programs and / or modules stored in memory 402, and by calling data stored in memory 402. Optionally, processor 401 may include one or more processing cores; preferably, processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into processor 401.

[0182] The memory 402 can be used to store software programs and modules. The processor 401 executes various functional applications and object feature constructions by running the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 402 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 with access to the memory 402.

[0183] The computer device also includes a power supply 403 that supplies power to the various components. Preferably, the power supply 403 can be logically connected to the processor 401 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 403 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0184] The computer device may also include an input unit 404, which can be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0185] Although not shown, the computer device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 401 in the computer device loads the executable files corresponding to the processes of one or more applications into the memory 402 according to the following instructions, and the processor 401 runs the applications stored in the memory 402 to realize various functions, as follows:

[0186] Obtain object label information of the object to be constructed in the target system. This object label information includes at least one object label for an object attribute of the object to be constructed and the corresponding label weight. The label weight is used to characterize the importance of the object attribute to the object to be constructed. Calculate the label similarity between the object labels and determine the basic object labels in the object label information based on the label similarity and label weight. Extract the semantic information corresponding to the basic object labels and perform semantic propagation in a preset label relationship graph based on the semantic information to obtain at least one associated object label and the label association path corresponding to the associated object label. The preset label relationship graph includes at least one node and node edges connected based on the semantic relationship between nodes. Each node corresponds to a preset object label. Count the number of nodes in the label association path and adjust the label weight of the basic object label corresponding to the associated object label according to the number of nodes to obtain the associated label weight corresponding to the associated object label. Update the object label information based on the associated object label and the associated label weight corresponding to the associated object label, and construct the object features corresponding to the object to be constructed based on the updated object label information.

[0187] The specific implementation of each of the above operations can be found in the preceding embodiments, and will not be repeated here. It should be noted that the computer device provided in this application embodiment and the object feature construction method in the above embodiments belong to the same concept, and its specific implementation process can be found in the above method embodiments, and will not be repeated here.

[0188] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0189] Therefore, embodiments of this application provide a computer-readable storage medium storing a plurality of instructions that can be loaded by a processor to execute steps in any of the object feature construction methods provided in embodiments of this application. For example, the instructions can execute the following steps:

[0190] Obtain object label information of the object to be constructed in the target system. This object label information includes at least one object label for an object attribute of the object to be constructed and the corresponding label weight. The label weight is used to characterize the importance of the object attribute to the object to be constructed. Calculate the label similarity between the object labels and determine the basic object labels in the object label information based on the label similarity and label weight. Extract the semantic information corresponding to the basic object labels and perform semantic propagation in a preset label relationship graph based on the semantic information to obtain at least one associated object label and the label association path corresponding to the associated object label. The preset label relationship graph includes at least one node and node edges connected based on the semantic relationship between nodes. Each node corresponds to a preset object label. Count the number of nodes in the label association path and adjust the label weight of the basic object label corresponding to the associated object label according to the number of nodes to obtain the associated label weight corresponding to the associated object label. Update the object label information based on the associated object label and the associated label weight corresponding to the associated object label, and construct the object features corresponding to the object to be constructed based on the updated object label information.

[0191] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0192] Since the instructions stored in the computer-readable storage medium can execute the steps in any of the object feature construction methods provided in the embodiments of this application, the beneficial effects that any of the object feature construction methods provided in the embodiments of this application can achieve can be realized. For details, please refer to the previous embodiments, which will not be repeated here.

[0193] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various optional implementations of the above embodiments.

[0194] The foregoing has provided a detailed description of an object feature construction method, apparatus, and computer-readable storage medium provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. An object feature construction method characterized by comprising: The method comprises the following steps: acquiring object label information of an object to be constructed in a target system, the object label information comprising at least one object label of an object attribute of the object to be constructed and a label weight corresponding to the object label, the label weight being used to represent the importance of the object attribute to the object to be constructed; calculating label similarity between the object labels and determining a basic object label in the object label information based on the label similarity and the label weight; extracting semantic information corresponding to the basic object label and performing semantic propagation in a preset label relationship graph based on the semantic information to obtain at least one associated object label and a label association path corresponding to the associated object label, the preset label relationship graph comprising at least one node and a node edge connected based on a semantic relationship between the nodes, each node corresponding to a preset object label; counting the number of nodes of the label association path and adjusting the label weight of the basic object label corresponding to the associated object label according to the number of nodes to obtain an associated label weight corresponding to the associated object label; updating the object label information based on the associated object label and the associated label weight corresponding to the associated object label, and constructing an object feature corresponding to the object to be constructed based on the updated object label information.

2. The object feature construction method of claim 1, wherein The method comprises the following steps: based on the semantic information, performing semantic propagation in a preset label relationship graph to obtain at least one candidate associated object label and a candidate label association path corresponding to the candidate associated object label; counting the number of labels of the candidate associated object label and calculating the association degree between the candidate associated object label and the basic object label when the number of labels is greater than a preset label number threshold; based on the association degree, screening at least one associated object label from the candidate associated object label and obtaining a label association path corresponding to the associated object label.

3. The object feature construction method of claim 1, wherein The method comprises the following steps: based on the label similarity, identifying at least one object label group from the object labels, the object label group comprising at least two object labels; based on the label weight, screening at least one basic object label from the object label group.

4. The object feature construction method of claim 3, wherein The method comprises the following steps: acquiring a first similarity threshold and a second similarity threshold, the first similarity threshold being greater than the second similarity threshold; based on the first similarity threshold and the label similarity, screening at least one first object label group from the object labels, and based on the second similarity threshold and the label similarity, screening at least one second object label group from the object labels; the first object label group and the second object label group are used as the object label group.

5. The object feature construction method according to any one of claims 1 to 4, characterized by, The object label information of the to-be-constructed object in the target system is acquired, and the object label information includes: The number of target object label information corresponding to the target object in the target system is counted to obtain the number of object label information corresponding to the target object; The to-be-constructed object is identified in the target object based on the number of object label information; The object label information of the to-be-constructed object in the target system is acquired.

6. The object feature construction method of claim 5, wherein The number of target object label information corresponding to the target object in the target system is counted to obtain the number of object label information corresponding to the target object, including: The device attribute information corresponding to the target object is acquired; At least one device update object is identified in the target object based on the device attribute information, and the device update object label information of the device update object is determined according to the device attribute information; The target object label information is updated based on the device update object label information to obtain updated target object label information, and the number of updated target object label information corresponding to the target object is counted to obtain the number of object label information corresponding to the target object.

7. The object feature construction method of claim 6, wherein The device update object label information of the device update object is determined according to the device attribute information, including: The current device object label information and the historical device object label information corresponding to the device update object are acquired according to the device attribute information; The current device object label information and the historical device object label information are fused to obtain the device update object label information corresponding to the device update object.

8. The object feature construction method of claim 1, wherein, After the object feature corresponding to the to-be-constructed object is constructed based on the updated object label information, the method further includes: At least one to-be-pushed content is acquired; At least one target content matching the object feature is identified in the to-be-pushed content based on the object feature corresponding to the to-be-constructed object; The target content is pushed to the to-be-constructed object.

9. An object feature construction apparatus characterized by comprising: It includes: An acquisition unit is configured to acquire object label information of a to-be-constructed object in a target system, the object label information including at least one object label of an object attribute of the to-be-constructed object and a label weight corresponding to the object label, the label weight being used to represent the importance of the object attribute to the to-be-constructed object; A determination unit is configured to calculate label similarity between the object labels and determine a basic object label in the object label information based on the label similarity and the label weight; A propagation unit is configured to extract semantic information corresponding to the basic object label and perform semantic propagation in a preset label relationship graph based on the semantic information to obtain at least one associated object label and a label association path corresponding to the associated object label, the preset label relationship graph including at least one node and a node edge connected based on a semantic relationship between the nodes, each node corresponding to a preset object label; An adjustment unit is configured to count the number of nodes of the label association path and adjust the label weight of the basic object label corresponding to the associated object label according to the number of nodes to obtain an associated label weight corresponding to the associated object label. The constructing unit is configured to update the object label information based on the association object label and the association label weight corresponding to the association object label, and construct the object feature corresponding to the to-be-constructed object based on the updated object label information.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a plurality of instructions, which are adapted to be loaded by the processor to perform the steps in the object feature construction method of any one of claims 1 to 8.

Images