A knowledge graph construction method for advertising recommendation

Abstract

The present invention discloses a knowledge graph construction method for advertisement recommendation, comprising the following steps: S10, obtaining historical advertisement delivery data to form a data set, wherein the data set contains a variety of advertisement features that can reflect advertisement delivery information; S20, ontology construction, defining the link relationship between advertisement features; S30, distinguishing fixed-length features, variable-length features and mixed features; a fixed-length feature is that the advertisement feature can only be linked to another unique advertisement feature under a link relationship; a variable-length feature is that the advertisement feature can be linked to multiple different advertisement features under a link relationship; a mixed feature is that there is a constraint relationship between the advertisement feature and other advertisement features; S40, knowledge is extracted in different ways according to the characteristics of the fixed-length feature, the variable-length feature and the mixed feature; S50, each triple is constructed to form a knowledge graph. The present invention realizes the reasonable extraction of advertisement data into a knowledge graph.

Description

Technical Field

[0001] The present invention relates to the technical field of data management and analysis, and in particular to a method for constructing a knowledge graph for advertisement recommendation. Background Art

[0002] In today's digital business environment, advertising recommendation systems play a vital role, providing companies with an effective way to promote products and services. With the rapid development of the Internet, users' activities on digital platforms have generated a large amount of data, including information such as ad clicks, search history, and user preferences. In order to better understand user needs and improve advertising effectiveness, advertising recommendation systems have gradually evolved into a data-driven intelligent tool. In order to promote advertising in a targeted manner, it is crucial to collect historical advertising data and analyze the correlation between data.

[0003] Knowledge graph is a tool that can present data in a structured form. Using knowledge graph can more conveniently sort out the relationship between data. The problem of how to reasonably extract advertising data so that the advertising data can be reasonably presented through knowledge graph needs to be solved. Summary of the invention

[0004] The purpose of the present invention is to provide a method for constructing a knowledge graph for advertisement recommendation, and the technical problem to be solved is how to reasonably extract advertisement data into the knowledge graph.

[0005] To achieve the above object, the solution of the present invention is: a knowledge graph construction method for advertising recommendation, comprising the following steps:

[0006] S10, data acquisition, acquiring historical advertising data to form a data set, the data set containing a variety of advertising features that can reflect advertising information;

[0007] S20, ontology construction, defining the link relationship between advertising features;

[0008] S30, feature classification, distinguishing fixed-length features, variable-length features and mixed features;

[0009] The fixed-length feature is that the advertisement feature can only be linked to another advertisement feature under a link relationship;

[0010] The variable length feature is that the advertisement feature can be linked to multiple different advertisement features under one link relationship;

[0011] The mixed feature is that there is a constraint relationship between this advertising feature and other advertising features;

[0012] S40, knowledge extraction, performing knowledge extraction according to different types to form triples;

[0013] For each fixed-length feature, the triplet formed is:

[0014] ;

[0015] in, represents the i-th triple, The identifier representing the i-th advertisement feature, The identifier of the j-th link relationship, Indicates the advertising feature associated with the i-th advertising feature through the j-th link relationship, and ⊕ is a placeholder;

[0016] For each variable-length feature, the triplet formed is:

[0017] ;

[0018] in, represents the kth triplet of the i-th advertisement feature, The identifier representing the i-th advertisement feature, The identifier of the j-th link relationship, Indicates the kth advertising feature associated with the ith advertising feature under the jth link relationship, and ⊕ is a placeholder;

[0019] For each mixed feature, the triplet formed is:

[0020] ;

[0021] in, represents the kth triplet of the i-th advertisement feature, The identifier representing the i-th advertisement feature, The identifier of the cth constraint relationship. The identifier of the constrained relationship corresponding to the constraint relationship. represents the jth advertising feature to which the ith advertising feature is linked under the constraint relationship, represents the kth advertising feature to which the i-th advertising feature is linked under the constraint relationship;

[0022] S50, construct each triple to form a knowledge graph.

[0023] Furthermore, the data set includes advertising data, and the advertising features contained in the advertising data include at least product information of the products involved in each advertisement, geographical location information of each advertisement, operator information of each advertisement, promotional keyword information used in each advertisement, and placement scene information of each advertisement.

[0024] Furthermore, the product information at least includes the product name and product category.

[0025] Furthermore, the geographic location information includes at least the province and / or city where the advertisement is placed.

[0026] Furthermore, the data set also includes user data, and the advertising features contained in the user data include at least the user's province, the user's city, the device used, the operator and the data source.

[0027] Furthermore, the link relationship includes at least the correspondence between each advertisement and the geographical location of delivery, the correspondence between each advertisement and the operator, the correspondence between each advertisement and the delivery device, and the correspondence between each advertisement and the delivery scene position.

[0028] Furthermore, the advertisement features of the header entity of each triplet refer to the code of the advertisement, and different advertisements are distinguished by different codes.

[0029] After adopting the above scheme, the beneficial effects of the present invention are: by distinguishing fixed-length features, variable-length features and mixed features before extracting triples, and then extracting and constructing triples for different types respectively; for fixed-length features, it is achieved to ensure data standardization and reduce data redundancy, thereby effectively solving the high data similarity problem when the subsequent knowledge graph is applied to advertising recommendation, and the unified processing of fixed-length features can ensure the consistency of information of the same type and avoid the influence of homogeneous information; for variable-length features, it is achieved to capture the diverse information in the text and avoid information convergence, so that the uniqueness of each advertising information is retained; for mixed features, by incorporating constraint relationships into triples, the triples of mixed features bring a deeper semantic understanding to the knowledge graph and fully reflect the interaction between features; thereby achieving a more reasonable extraction of advertising data into the knowledge graph. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic diagram of a process of the present invention. DETAILED DESCRIPTION

[0031] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0032] The following will fully describe embodiments of the present invention with reference to the accompanying drawings. It should be noted that the present invention can be implemented in different forms and is not limited to the embodiments illustrated herein. These embodiments are provided to make the present disclosure thorough and complete and to fully convey the scope of the present invention to those skilled in the art.

[0033] The present invention provides a method for constructing a knowledge graph for advertisement recommendation, comprising the following steps:

[0034] S10, data acquisition, acquiring historical advertisement delivery data to form a data set, the data set includes a variety of advertisement features that can reflect advertisement delivery information, the specific types of advertisement features are not limited, and can be acquired according to actual needs; in a preferred embodiment, it is divided into advertisement data and user data; the advertisement data also includes product information of the products involved in each advertisement, geographical location information of each advertisement delivery, operator information of each advertisement delivery, promotional keyword information used in each advertisement, and delivery scene and position information of each advertisement, etc. More specifically, the product information at least includes the product name and product classification, the geographical location information at least includes the province and / or city where the advertisement is delivered, the operator information refers to the network operator of the network used for delivery of the advertisement, and the promotional keyword information can be the efficacy, purpose and usage scenario of the product, etc.; the user data also includes the province where the user is located, the city where the user is located, the device used, the operator and the data source, etc.; preferably, after acquiring the data, the data is also cleaned and filtered to remove data that does not need further analysis;

[0035] S20, ontology construction, defining the link relationship between the advertising features. The specific link relationship can be defined in combination with actual needs and is not specifically limited; the preferred link relationship includes the link relationship between each advertisement and the delivery APP, the link relationship between each advertisement and the delivery geographical location, the link relationship between each advertisement and the delivery device, the link relationship between each advertisement and the operator that delivers the advertisement, the link relationship between each advertisement and the data source in the advertisement, the link relationship between each advertisement and the cause of the data in the advertisement, and the link relationship between each advertisement and the delivery scene position;

[0036] S30, feature classification, distinguishing fixed-length features, variable-length features and mixed features; fixed-length features are that the advertisement feature can only be linked to another unique advertisement feature under a link relationship; variable-length features are that the advertisement feature can be linked to multiple different advertisement features under a link relationship; mixed features are that there is a constraint relationship between the advertisement feature and other advertisement features, for example;

[0037] Table 1 Feature classification diagram

[0038]

[0039] S40, knowledge extraction, performing knowledge extraction according to different types to form triples; specifically in this embodiment, the advertisement feature of the head entity of each triple refers to the code of the advertisement, and different advertisements are distinguished by different codes, but this is not limited to this;

[0040] For each fixed-length feature, the triplet formed is:

[0041] ;

[0042] in, represents the i-th triple, The identifier representing the i-th advertisement feature, The identifier of the j-th link relationship, Indicates the advertising feature associated with the i-th advertising feature through the j-th link relationship, and ⊕ is a placeholder;

[0043] For each variable-length feature, the triplet formed is:

[0044] ;

[0045] in, represents the kth triplet of the i-th advertisement feature, The identifier representing the i-th advertisement feature, The identifier of the j-th link relationship, Indicates the kth advertising feature associated with the ith advertising feature under the jth link relationship, and ⊕ is a placeholder;

[0046] For each mixed feature, the triplet formed is:

[0047] ;

[0048] in, represents the kth triplet of the i-th advertisement feature, The identifier representing the i-th advertisement feature, The identifier of the cth constraint relationship. The identifier of the constrained relationship corresponding to the constraint relationship. represents the jth advertising feature to which the ith advertising feature is linked under the constraint relationship, represents the kth advertising feature to which the i-th advertising feature is linked under the constraint relationship;

[0049] S50. Construct each triple into a knowledge graph. Constructing a knowledge graph using the extracted triples is a conventional technical means in this field and will not be described in detail. Finally, the Neo4j graph database is used to store the knowledge graph, so that the correlation between the advertising data is clearer, so as to facilitate subsequent reference and analysis when making advertising recommendations.

[0050] The above description is only a preferred embodiment of the present invention and is not a limitation on the design of this case. Any equivalent changes made based on the design key of this case shall fall within the protection scope of this case.

Claims

1. A knowledge graph construction method for advertising recommendation, characterized in that: The following steps are involved: S10, data acquisition, acquiring historical advertising data to form a data set, the data set containing a variety of advertising features that can reflect advertising information; S20, ontology construction, defining the link relationship between advertising features; S30, feature classification, distinguishing fixed-length features, variable-length features and mixed features; The fixed-length feature is that the advertisement feature can only be linked to another advertisement feature under a link relationship; The variable length feature is that the advertisement feature can be linked to multiple different advertisement features under one link relationship; The mixed feature is that there is a constraint relationship between this advertising feature and other advertising features; S40, knowledge extraction, performing knowledge extraction according to different types to form triples; For each fixed-length feature, the triplet formed is: ； in, represents the i-th triple, The identifier representing the i-th advertisement feature, The identifier of the j-th link relationship, Indicates the advertising feature associated with the i-th advertising feature through the j-th link relationship, and ⊕ is a placeholder; For each variable-length feature, the triplet formed is: ； in, represents the kth triplet of the i-th advertisement feature, The identifier representing the i-th advertisement feature, The identifier of the j-th link relationship, Indicates the kth advertising feature associated with the ith advertising feature under the jth link relationship, and ⊕ is a placeholder; For each mixed feature, the triplet formed is: ； in, represents the kth triplet of the i-th advertisement feature, The identifier representing the i-th advertisement feature, The identifier of the cth constraint relationship. The identifier of the constrained relationship corresponding to the constraint relationship. represents the jth advertising feature to which the ith advertising feature is linked under the constraint relationship, represents the kth advertising feature to which the i-th advertising feature is linked under the constraint relationship; S50, construct each triple to form a knowledge graph.

2. A knowledge graph construction method for advertising recommendation as claimed in claim 1, characterized in that: The data set includes advertising data, and the advertising features contained in the advertising data include at least product information of the products involved in each advertisement, geographical location information of each advertisement, operator information of each advertisement, promotional keyword information used in each advertisement, and placement scene information of each advertisement.

3. A knowledge graph construction method for advertising recommendation as claimed in claim 2, characterized in that: Product information, including at least product name and product category.

4. A knowledge graph construction method for advertising recommendation as claimed in claim 2, characterized in that: Geographic location information, including at least the province and / or city where the advertisement is served.

5. A knowledge graph construction method for advertising recommendation as claimed in claim 2, characterized in that: The data set also includes user data, and the advertising features contained in the user data include at least the user's province, user's city, device used, operator and data source.

6. A knowledge graph construction method for advertising recommendation as claimed in claim 5, characterized in that: The link relationship at least includes the correspondence between each advertisement and the geographical location of delivery, the correspondence between each advertisement and the operator, the correspondence between each advertisement and the delivery device, and the correspondence between each advertisement and the delivery scene position.

7. A knowledge graph construction method for advertising recommendation according to claim 1, characterized in that: The advertisement features of the header entity of each triplet refer to the code of the advertisement, and different advertisements are distinguished by different codes.

Images