Advertisement screening method and system based on artificial intelligence

Abstract

Disclosed is an artificial intelligence-based advertisement screening method and system, which relate to the field of advertisement service technology. The method uses artificial intelligence technology to perform correlation matching and fusion analysis on user behavior data and text description data of candidate advertisements to be placed, so as to intelligently judge the degree of adaptation between the candidate advertisements to be placed and the users, and realizes personalized recommendation of advertisements by accurately placing advertisements with high adaptation degrees, so that the advertisements can effectively reach the target audience.

Description

Technical Field

[0001] The present disclosure relates to the technical field of advertising services, and in particular to an advertising screening method and system based on artificial intelligence. Background Art

[0002] With the advent of the digital age, the advertising industry is undergoing unprecedented changes. Traditional advertising methods, such as radio, television, and newspapers, have gradually been replaced by Internet advertising. However, with the massive amount of information on Internet advertising and the growth of personalized user needs, how to accurately place advertisements has become a challenge facing advertisers.

[0003] In daily life, users are exposed to a huge amount of advertising information every day, which leads to the distraction of users' attention to advertisements. Advertisers need to stand out from the crowd and attract users' attention. But more importantly, each user's interests and preferences are unique. Advertisers need to provide personalized advertising content to attract audience users. Traditional advertising often adopts a one-size-fits-all approach, ignoring individual differences among users.

[0004] Moreover, with the development of artificial intelligence technology, it has gradually been applied to various industries and gradually matured. Artificial intelligence technology has also begun to be gradually integrated into the advertising industry; how to better use artificial intelligence technology to analyze, screen and process relevant advertising data, better provide users with personalized services, and improve the publicity effect of advertising is an important issue currently facing the advertising industry.

[0005] Therefore, an optimized solution is expected to better solve the above problems. Summary of the invention

[0006] The present disclosure is made in view of the above problems. One object of the present disclosure is to provide an advertisement screening method and system based on artificial intelligence.

[0007] The embodiment of the present disclosure provides an advertisement screening method based on artificial intelligence, which includes:

[0008] Get the time series of user behavior data;

[0009] Get the text description of the candidate advertisement to be delivered;

[0010] Performing contextual semantic analysis on the time series of the user behavior data to obtain a time series of contextual user behavior semantic encoding feature vectors;

[0011] Inputting the time series of the contextual user behavior semantic encoding feature vector into a self-supervised feature sequence significant fusion network to obtain a user behavior pattern self-supervised semantic representation vector;

[0012] Performing semantic coding on the text description of the candidate advertisement to be placed to obtain a semantic coding vector of characteristics of the advertisement to be placed;

[0013] Inputting the feature semantic encoding vector of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern into a balanced threshold feature vector adaptive fusion network to obtain a semantic fusion representation vector of the advertisement to be placed-user behavior;

[0014] Based on the to-be-delivered advertisement-user behavior semantic fusion representation vector, a screening result is determined.

[0015] For example, according to the artificial intelligence-based advertising screening method of an embodiment of the present disclosure, context semantic analysis is performed on the time series of the user behavior data to obtain a time series of context user behavior semantic encoding feature vectors, including:

[0016] Embedding and encoding each user behavior data in the time series of the user behavior data to obtain a time series of user behavior embedded coding vectors;

[0017] The time series of the user behavior embedding coding vector is input into the user behavior context encoder based on the Transformer layer to obtain the time series of the context user behavior semantic coding feature vector.

[0018] For example, according to an artificial intelligence-based advertising screening method of an embodiment of the present disclosure, the time series of the contextual user behavior semantic encoding feature vector is input into a self-supervised feature sequence significant fusion network to obtain a user behavior pattern self-supervised semantic representation vector, including:

[0019] Calculating the mean vector of the time series of the contextual user behavior semantic encoding feature vector as the clustering center of the self-supervised sequence distribution;

[0020] Calculating the user behavior semantic relevance between each context user behavior semantic encoding feature vector in the time series of the context user behavior semantic encoding feature vector and the cluster center of the self-supervised sequence distribution to obtain a sequence of user behavior semantic relevance;

[0021] Inputting the sequence of user behavior semantic relevance into a softmax function for soft maximum normalization processing to obtain a sequence of self-supervised context user behavior semantic relevance coefficients;

[0022] Taking the sequence of the self-supervised context user behavior semantic relevance coefficients as the weight sequence, the vector-by-vector weighted sum of the time series of the context user behavior semantic encoding feature vectors is calculated to obtain the user behavior pattern self-supervised semantic representation vector.

[0023] For example, according to the artificial intelligence-based advertising screening method of an embodiment of the present disclosure, the user behavior semantic relevance between each context user behavior semantic encoding feature vector in the time series of the context user behavior semantic encoding feature vector and the cluster center of the self-supervised sequence distribution is calculated to obtain a sequence of user behavior semantic relevance, including:

[0024] Encoding the context user behavior semantic encoding feature vector through a fully connected layer to obtain a context user behavior semantic fully connected encoding feature vector;

[0025] The vector product between the transposed vector of the self-supervised sequence distribution cluster center and the contextual user behavior semantic fully connected encoding feature vector is calculated to obtain the user behavior semantic relevance.

[0026] For example, according to the artificial intelligence-based advertisement screening method of an embodiment of the present disclosure, semantic encoding is performed on the text description of the candidate advertisement to be placed to obtain a semantic encoding vector of the characteristics of the advertisement to be placed, including:

[0027] Performing word segmentation processing on the text description of the candidate advertisement to be placed so as to convert the text description of the candidate advertisement to be placed into a word sequence consisting of a plurality of words;

[0028] Mapping each word in the word sequence to a word vector using a word embedding layer of a semantic encoder including a word embedding layer to obtain a sequence of word vectors;

[0029] Using the semantic encoder including the word embedding layer, the sequence of word vectors is encoded based on global context semantics to obtain a sequence of encoding vectors of advertisements to be delivered;

[0030] The sequence of the to-be-placed advertisement coding vectors is cascaded to obtain the to-be-placed advertisement characteristic semantic coding vector.

[0031] For example, according to an artificial intelligence-based advertising screening method of an embodiment of the present disclosure, the feature semantic encoding vector of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern are input into an adaptive fusion network of a balanced threshold feature vector to obtain a semantic fusion representation vector of the advertisement to be placed-user behavior, including:

[0032] Determine a first balancing threshold dynamic fusion weight, a second balancing threshold dynamic fusion weight, and a third balancing threshold dynamic fusion weight based on the semantic encoding vector of the characteristic of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern;

[0033] Adaptively and dynamically integrate the feature semantic encoding vector of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern with the first balance threshold dynamic fusion weight, the second balance threshold dynamic fusion weight, and the third balance threshold dynamic fusion weight to obtain a feature semantic dynamic interaction fusion item of the advertisement to be placed and a user behavior semantic dynamic interaction fusion item;

[0034] The dynamic interaction fusion item of the semantics of the characteristics of the advertisement to be placed and the dynamic interaction fusion item of the semantics of the user behavior are fused element by element to obtain the semantic fusion representation vector of the advertisement to be placed-user behavior.

[0035] For example, according to the artificial intelligence-based advertising screening method of an embodiment of the present disclosure, the first balanced threshold dynamic fusion weight, the second balanced threshold dynamic fusion weight, and the third balanced threshold dynamic fusion weight are determined based on the semantic encoding vector of the feature of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern, including:

[0036] Cascading the semantic encoding vector of the feature of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern to obtain a semantic cascade vector of the advertisement to be placed-user behavior;

[0037] Calculate the position-by-position addition of the semantic encoding vector of the feature of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern to obtain an element-by-element addition vector of the semantics of the advertisement to be placed-user behavior;

[0038] Calculate the position-by-position element-by-element multiplication of the semantic encoding vector of the feature of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern to obtain an element-by-element multiplication vector of the semantics of the advertisement to be placed-user behavior;

[0039] The first transformation vector is multiplied by the advertisement to be placed-user behavior semantic cascade vector, and the resultant is added to the first bias parameter to obtain a cascade coefficient; the second transformation vector is multiplied by the advertisement to be placed-user behavior semantic cascade vector, and the resultant is added to the second bias parameter to obtain an addition coefficient; the third transformation vector is multiplied by the advertisement to be placed-user behavior semantic cascade vector, and the resultant is added to the third bias parameter to obtain a multiplication coefficient;

[0040] The cascade coefficient, the addition coefficient and the multiplication coefficient are respectively passed through a sigmoid activation function to obtain the first equalization threshold dynamic fusion weight, the second equalization threshold dynamic fusion weight and the third equalization threshold dynamic fusion weight.

[0041] For example, according to the artificial intelligence-based advertising screening method of an embodiment of the present disclosure, the first balanced threshold dynamic fusion weight, the second balanced threshold dynamic fusion weight, and the third balanced threshold dynamic fusion weight are used to adaptively and dynamically integrate the feature semantic encoding vector of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern to obtain the feature semantic dynamic interaction fusion item of the advertisement to be placed and the user behavior semantic dynamic interaction fusion item, including:

[0042] The average value of the first balancing threshold dynamic fusion weight, the second balancing threshold dynamic fusion weight and the third balancing threshold dynamic fusion weight is used as the user behavior semantic dynamic interaction fusion coefficient;

[0043] The value of 1 minus the user behavior semantic dynamic interaction fusion coefficient is used as the semantic dynamic interaction fusion coefficient of the feature of the advertisement to be placed;

[0044] Calculate the point multiplication of the semantic dynamic interaction fusion coefficient of the characteristic semantics of the advertisement to be placed and the semantic coding vector of the characteristic semantics of the advertisement to be placed to obtain the semantic dynamic interaction fusion item of the characteristic semantics of the advertisement to be placed;

[0045] The user behavior semantics dynamic interaction fusion term is obtained by calculating the dot multiplication of the user behavior semantics dynamic interaction fusion coefficient and the user behavior pattern self-supervised semantic representation vector.

[0046] For example, according to the artificial intelligence-based advertisement screening method of an embodiment of the present disclosure, the screening result is determined based on the to-be-delivered advertisement-user behavior semantic fusion representation vector, including:

[0047] The advertisement to be placed-user behavior semantic fusion representation vector is input into a classifier-based advertisement screening module to obtain the screening result, and the screening result is used to indicate whether to place the candidate advertisement to be placed.

[0048] The embodiment of the present disclosure also provides an artificial intelligence-based advertising screening system, which includes:

[0049] User behavior data acquisition module, used to obtain the time series of user behavior data;

[0050] A module for obtaining candidate advertisements to be placed, used for obtaining text descriptions of candidate advertisements to be placed;

[0051] A context semantic analysis module, used for performing context semantic analysis on the time series of the user behavior data to obtain a time series of context user behavior semantic encoding feature vectors;

[0052] A vector fusion module, used for inputting the time series of the contextual user behavior semantic encoding feature vector into a self-supervised feature sequence significant fusion network to obtain a user behavior pattern self-supervised semantic representation vector;

[0053] A semantic coding module, used for semantically coding the text description of the candidate advertisement to be placed to obtain a semantic coding vector of the characteristic of the advertisement to be placed;

[0054] A semantic fusion module, used for inputting the semantic encoding vector of the feature of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern into a balanced threshold feature vector adaptive fusion network to obtain a semantic fusion representation vector of the advertisement to be placed-user behavior;

[0055] The screening module is used to determine the screening result based on the to-be-delivered advertisement-user behavior semantic fusion representation vector.

[0056] Compared with the prior art, the AI-based advertising screening method and system of the embodiments of the present disclosure utilize AI deep learning technology to perform correlation matching and fusion analysis on user behavior data and text description data of candidate advertisements to be placed, so as to intelligently determine the degree of compatibility between the candidate advertisements to be placed and the user. In this way, advertisements with high degree of compatibility are accurately placed to achieve personalized recommendation of advertisements, so that advertisements can effectively reach the target audience and improve the publicity effect of advertisements. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following briefly introduces the drawings of the embodiments of the present disclosure. Obviously, the drawings described below only relate to some embodiments of the present disclosure, and are not intended to limit the present disclosure.

[0058] Figure 1 A schematic diagram of the application architecture of the artificial intelligence-based advertisement screening method in an embodiment of the present disclosure is shown;

[0059] Figure 2 A flowchart of an advertisement screening method based on artificial intelligence in an embodiment of the present disclosure is shown;

[0060] Figure 3 A flowchart of sub-step S530 of the artificial intelligence-based advertisement screening method in an embodiment of the present disclosure is shown;

[0061] Figure 4 A flowchart of sub-step S540 of the artificial intelligence-based advertisement screening method in an embodiment of the present disclosure is shown;

[0062] Figure 5 A flowchart of sub-step S550 of the artificial intelligence-based advertisement screening method in an embodiment of the present disclosure is shown;

[0063] Figure 6 A flowchart of sub-step S560 of the artificial intelligence-based advertisement screening method in an embodiment of the present disclosure is shown;

[0064] Figure 7 A schematic diagram of the structure of an advertisement screening system based on artificial intelligence in an embodiment of the present disclosure is shown;

[0065] Figure 8 An application scenario diagram of the artificial intelligence-based advertisement screening method in an embodiment of the present disclosure is shown;

[0066] Fig. 9 A schematic diagram of a storage medium according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0067] The technical solutions in the embodiments of the present disclosure will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present disclosure, rather than all embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present disclosure.

[0068] The terms used in this specification are those common terms currently widely used in the art in consideration of the functions of the present disclosure, but these terms may vary according to the intention of a person of ordinary skill in the art, precedents, or new technologies in the art. In addition, specific terms may be selected by the applicant, and in this case, their detailed meanings will be described in the detailed description of the present disclosure. Therefore, the terms used in the specification should not be understood as simple names, but rather as an overall description based on the meaning of the terms and the present disclosure.

[0069] Although the present disclosure makes various references to certain modules in the system according to embodiments of the present disclosure, any number of different modules can be used and run on a user terminal and / or server. The modules are illustrative only, and different aspects of the systems and methods can use different modules.

[0070] Flowcharts are used in this disclosure to illustrate the operations performed by the system according to the embodiments of the present disclosure. It should be understood that the preceding or following operations are not necessarily performed precisely in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. At the same time, other operations may also be added to these processes, or one or more operations may be removed from these processes.

[0071] Figure 1 A schematic diagram of the application architecture of an artificial intelligence-based advertisement screening method in an embodiment of the present disclosure is shown, including a server 100 and a terminal device 200.

[0072] The terminal device 200 and the server 100 can be connected via the Internet to achieve mutual communication. Optionally, the above-mentioned Internet uses standard communication technology and / or protocol. The Internet is usually the Internet, but it can also be any network, including but not limited to any combination of a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), a mobile, wired or wireless network, a private network or a virtual private network. In some embodiments, the data exchanged through the network can be represented by technologies and / or formats including Hyper Text Markup Language (HTML), Extensible Markup Language (XML), etc. In addition, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), Internet Protocol Security (IPsec), etc. can be used to encrypt all or some links. In other embodiments, customized and / or dedicated data communication technologies can also be used to replace or supplement the above-mentioned data communication technologies.

[0073] The server 100 can provide various network services for the terminal device 200, wherein the server 100 can be a single server, a server cluster consisting of several servers, or a cloud computing center. Specifically, the server 100 may include a processor 110 (Center Processing Unit, CPU), a memory 120, an input device 130, and an output device 140, etc. The input device 130 may include a keyboard, a mouse, a touch screen, etc., and the output device 140 may include a display device, such as a liquid crystal display (Liquid Crystal Display, LCD), a cathode ray tube (Cathode Ray Tube, CRT), etc.

[0074] The memory 120 may include a read-only memory (ROM) and a random access memory (RAM), and provides the processor 110 with program instructions and data stored in the memory 120. In the embodiment of the present disclosure, the memory 120 may be used to store the program of the artificial intelligence-based advertising screening method in the embodiment of the present disclosure.

[0075] The processor 110 calls the program instructions stored in the memory 120, and the processor 110 is used to execute the steps of any one of the artificial intelligence-based advertising screening methods in the embodiments of the present disclosure according to the obtained program instructions.

[0076] In addition, the application architecture diagram in the embodiment of the present disclosure is intended to more clearly illustrate the technical solution in the embodiment of the present disclosure, and does not constitute a limitation on the technical solution provided in the embodiment of the present disclosure. Of course, for other application architectures and business applications, the technical solution provided in the embodiment of the present disclosure is also applicable to similar problems.

[0077] The following is a non-restrictive description of the artificial intelligence-based advertising screening method provided according to at least one embodiment of the present disclosure through several examples or embodiments. As described below, different features in these specific examples or embodiments can be combined with each other without conflicting with each other to obtain new examples or embodiments, and these new examples or embodiments also fall within the scope of protection of the present disclosure.

[0078] In response to the problems in the above background technology, the technical concept of the present disclosure is: using artificial intelligence technology to perform correlation matching and fusion analysis on user behavior data and text description data of candidate ads to be placed, so as to intelligently determine the degree of compatibility between the candidate ads to be placed and the user. In this way, ads with high degree of compatibility are accurately placed to achieve personalized recommendation of ads, so that ads can effectively reach the target audience.

[0079] Based on this, Figure 2 The flowchart of the artificial intelligence-based advertising screening method in the embodiment of the present disclosure is shown. For example, the artificial intelligence-based advertising screening method can be executed by a server, which can be Figure 1 The server 100 shown in FIG. Figure 2As shown, according to the embodiment of the present disclosure, the artificial intelligence-based advertising screening method includes the following steps: S510, obtaining a time series of user behavior data; S520, obtaining a text description of an alternative advertisement to be placed; S530, performing contextual semantic analysis on the time series of the user behavior data to obtain a time series of contextual user behavior semantic encoding feature vectors; S540, inputting the time series of contextual user behavior semantic encoding feature vectors into a self-supervised feature sequence significant fusion network to obtain a user behavior pattern self-supervised semantic representation vector; S550, performing semantic encoding on the text description of the alternative advertisement to be placed to obtain a semantic encoding vector of characteristics of the advertisement to be placed; S560, inputting the semantic encoding vector of the characteristics of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern into an equalization threshold feature vector adaptive fusion network to obtain a semantic fusion representation vector of the advertisement to be placed-user behavior; S570, determining the screening result based on the semantic fusion representation vector of the advertisement to be placed-user behavior.

[0080] Specifically, in the technical solution of the present disclosure, first, the time series of user behavior data is obtained; and the text description of the candidate advertisements to be placed is obtained. Among them, the user behavior data can reflect the interests, preferences and needs of the user. By obtaining the time series of the user behavior data, the user's behavior pattern and interest bias can be identified, and the user's future needs can be predicted to a certain extent. The acquisition of this information is helpful for the precise delivery of advertisements. In actual application scenarios, the user behavior data can be various activity records of the user on the Internet, including but not limited to: browsing history, search history, purchase behavior and social media activities. It should be noted that the process of obtaining these data needs to protect the privacy of the user and is based on the user's authorization. In addition, the text description of the candidate advertisements to be placed provides semantic information of the advertisement content, which helps the system understand the theme, characteristics and value proposition of the advertisement. Through the acquisition of the text description, the system can use natural language processing technology to semantically encode and analyze it, so as to deeply understand the semantic content of the advertisement, and further realize the fit analysis between the user behavior pattern and the semantic content of the advertisement.

[0081] Then, each user behavior data in the time series of the user behavior data is embedded and encoded to obtain a time series of user behavior embedded coding vectors; and the time series of the user behavior embedded coding vectors is input into a user behavior context encoder based on the Transformer layer to obtain a time series of contextual user behavior semantic coding feature vectors. That is, each unstructured user behavior data is converted into a structured vector representation through embedding coding to simplify the complexity of the model; and then the user behavior context encoder is constructed using the Transformer layer to capture the long-distance dependencies in the time series of the user behavior embedded coding vectors through the self-attention mechanism, so as to understand the dynamic temporal pattern of user behavior.

[0082] Accordingly, in step S530, if Figure 3 As shown, contextual semantic analysis is performed on the time series of the user behavior data to obtain a time series of contextual user behavior semantic encoding feature vectors, including: S531, embedding encoding is performed on each user behavior data in the time series of the user behavior data to obtain a time series of user behavior embedded encoding vectors; S532, the time series of user behavior embedded encoding vectors is input into a user behavior context encoder based on a Transformer layer to obtain a time series of contextual user behavior semantic encoding feature vectors.

[0083] It should be understood that each of the contextual user behavior semantic encoding feature vectors expresses the contextual behavior pattern of the user in each local time span. However, in terms of the task of depicting and characterizing the comprehensive and global behavior pattern of the user, not all contextual user behavior semantic encoding feature vectors may have the same importance and contribution. In actual application scenarios, user behavior at certain time points may be more important than behavior at other time points. For example, user purchasing behavior during the holiday shopping season may be more important than usual; the degree of user involvement in certain behaviors, such as browsing time, purchase amount or interaction depth, may indicate that these behaviors are more important. Therefore, in order to capture the important features hidden in each of the contextual user behavior semantic encoding feature vectors, in the technical solution disclosed in the present invention, it is expected that the time series of the contextual user behavior semantic encoding feature vectors is input into the self-supervised feature sequence significant fusion network to obtain the user behavior pattern self-supervised semantic representation vector. That is, the self-supervised correlation significance in the time series of the contextual user behavior semantic encoding feature vector is captured by the self-supervised feature sequence significant fusion network, and the importance and contribution of each contextual user behavior semantic encoding feature vector are quantified by using this self-supervised correlation significance, thereby optimizing the comprehensive expression of user behavior.

[0084] In the embodiment of the present disclosure, in step S540, as Figure 4 As shown, the specific encoding process of inputting the time series of the context user behavior semantic encoding feature vector into the self-supervised feature sequence significant fusion network to obtain the user behavior pattern self-supervised semantic representation vector includes: S541, calculating the mean vector of the time series of the context user behavior semantic encoding feature vector as the self-supervised sequence distribution clustering center; S542, calculating the user behavior semantic relevance between each context user behavior semantic encoding feature vector in the time series of the context user behavior semantic encoding feature vector and the self-supervised sequence distribution clustering center to obtain a sequence of user behavior semantic relevance; S543, inputting the sequence of user behavior semantic relevance into the softmax function for soft maximum normalization processing to obtain a sequence of self-supervised context user behavior semantic relevance coefficients; S544, using the sequence of self-supervised context user behavior semantic relevance coefficients as the weight sequence, calculating the vector-by-vector weighted sum of the time series of the context user behavior semantic encoding feature vector to obtain the user behavior pattern self-supervised semantic representation vector.

[0085] Among them, in step S542, the user behavior semantic correlation between each context user behavior semantic encoding feature vector in the time series of the context user behavior semantic encoding feature vector and the cluster center of the self-supervised sequence distribution is calculated to obtain a sequence of user behavior semantic correlations, including: encoding the context user behavior semantic encoding feature vector through a fully connected layer to obtain a context user behavior semantic fully connected encoding feature vector; calculating the vector product between the transposed vector of the self-supervised sequence distribution cluster center and the context user behavior semantic fully connected encoding feature vector to obtain the user behavior semantic correlation.

[0086] In a specific example, the time series of the context user behavior semantic encoding feature vector is input into the self-supervised feature sequence significant fusion network to obtain the user behavior pattern self-supervised semantic representation vector, including: inputting the time series of the context user behavior semantic encoding feature vector into the self-supervised feature sequence significant fusion network to process it with the following fusion formula to obtain the user behavior pattern self-supervised semantic representation vector; wherein the fusion formula is:

[0087]

[0088]

[0089]

[0090] in, Represents the total number of semantic encoding feature vectors of contextual user behavior, is the first in the time series of the semantic encoding feature vector of the contextual user behavior contextual user behavior semantic encoding feature vector, is the cluster center of the self-supervised sequence distribution, is the weight matrix of the fully connected layer, is the bias vector of the fully connected layer, represents the sigmoid activation function, represents transpose, It is The semantic relevance of user behavior, represents the softmax activation function, It is the self-supervised semantic representation vector of user behavior pattern.

[0091] More specifically, the process of calculating the user behavior semantic relevance (self-supervised correlation significance for the entire sequence) is to learn the text semantic features of each context user behavior semantic encoding feature vector through a fully connected layer, that is, to automatically learn this text semantic feature through a (trained) weight matrix and a (trained) bias vector, and then fuse the transposed vector of the self-supervised sequence distribution cluster center with the text semantic features of each context user behavior semantic encoding feature vector to obtain the sequence of user behavior semantic relevance, so as to numerically represent the correlation between the text semantic features of each context user behavior semantic encoding feature vector and the overall user behavior semantic features. In essence, this is a weighted probability distribution mechanism, that is, assigning greater weight to important content and reducing weight to other content. Such a mechanism is more focused on finding useful information in the input data that is significantly related to the current task.

[0092] At the same time, semantic coding is performed on the text description of the candidate advertisement to be placed to capture the semantic content and meaning of the candidate advertisement to be placed, thereby obtaining a semantic coding vector of the characteristic of the candidate advertisement to be placed.

[0093] Accordingly, in step S550, if Figure 5 As shown, semantic encoding is performed on the text description of the alternative advertisement to be placed to obtain a semantic encoding vector of the characteristics of the advertisement to be placed, including: S551, word segmentation processing is performed on the text description of the alternative advertisement to be placed to convert the text description of the alternative advertisement to be placed into a word sequence composed of multiple words; S552, using a word embedding layer of a semantic encoder including a word embedding layer to map each word in the word sequence to a word vector to obtain a sequence of word vectors; S553, using the semantic encoder including the word embedding layer to perform global context-based semantic encoding on the sequence of word vectors to obtain a sequence of encoding vectors of the advertisement to be placed; S554, cascading the sequence of encoding vectors of the advertisement to be placed to obtain the semantic encoding vector of the characteristics of the advertisement to be placed.

[0094] It should be understood that by constructing the correlation between the dynamic temporal pattern of user behavior expressed by the self-supervised semantic representation vector of the user behavior pattern and the semantic features of the advertisement to be placed expressed by the semantic coding vector of the characteristics of the advertisement to be placed, the degree of adaptation between the two can be quantified, thereby guiding the personalized placement of advertisements. Therefore, in the technical solution disclosed in the present invention, further, the semantic coding vector of the characteristics of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern are input into the balanced threshold feature vector adaptive fusion network to obtain the advertisement to be placed-user behavior semantic fusion representation vector. Among them, in the process of feature interactive fusion, the balanced threshold feature vector adaptive fusion network adaptively balances the importance of the fusion information from the self-supervised semantic representation vector of the user behavior pattern and the semantic coding vector of the characteristics of the advertisement to be placed through the balanced threshold dynamic fusion weight, so that the model can flexibly adjust the importance between the two, thereby obtaining the final interactive representation.

[0095] Accordingly, in step S560, if Figure 6 As shown, the semantic encoding vector of the characteristic of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern are input into the balanced threshold feature vector adaptive fusion network to obtain the semantic fusion representation vector of the advertisement to be placed-user behavior, including: S561, determining the first balanced threshold dynamic fusion weight, the second balanced threshold dynamic fusion weight and the third balanced threshold dynamic fusion weight based on the semantic encoding vector of the characteristic of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern; S562, adaptively and dynamically integrating the semantic encoding vector of the characteristic of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern with the first balanced threshold dynamic fusion weight, the second balanced threshold dynamic fusion weight and the third balanced threshold dynamic fusion weight to obtain the semantic dynamic interaction fusion item of the characteristic of the advertisement to be placed and the semantic dynamic interaction fusion item of the user behavior; S563, element-by-element fusion of the semantic dynamic interaction fusion item of the characteristic of the advertisement to be placed and the semantic dynamic interaction fusion item of the user behavior to obtain the semantic fusion representation vector of the advertisement to be placed-user behavior.

[0096] Wherein, in step S561, the first balance threshold dynamic fusion weight, the second balance threshold dynamic fusion weight and the third balance threshold dynamic fusion weight are determined based on the semantic coding vector of the characteristic of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern, including: cascading the semantic coding vector of the characteristic of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern to obtain a cascade vector of the semantic coding vector of the characteristic of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern to obtain an element-by-element addition vector of the semantic coding vector of the characteristic of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern to obtain a cascade ... The quantities are multiplied element by element by position to obtain the element-by-element multiplication vector of the advertisement to be placed-user behavior semantics; the first transformation vector is multiplied by the advertisement to be placed-user behavior semantics cascade vector and then added to the first bias parameter to obtain the cascade coefficient, the second transformation vector is multiplied by the advertisement to be placed-user behavior semantics cascade vector and then added to the second bias parameter to obtain the addition coefficient, the third transformation vector is multiplied by the advertisement to be placed-user behavior semantics cascade vector and then added to the third bias parameter to obtain the multiplication coefficient; the cascade coefficient, the addition coefficient and the multiplication coefficient are respectively passed through the sigmoid activation function to obtain the first balance threshold dynamic fusion weight, the second balance threshold dynamic fusion weight and the third balance threshold dynamic fusion weight.

[0097] Among them, in step S562, the semantic encoding vector of the characteristic of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern are adaptively and dynamically integrated with the first balance threshold dynamic fusion weight, the second balance threshold dynamic fusion weight and the third balance threshold dynamic fusion weight to obtain the semantic dynamic interaction fusion item of the characteristic of the advertisement to be placed and the user behavior semantic dynamic interaction fusion item, including: taking the average value of the first balance threshold dynamic fusion weight, the second balance threshold dynamic fusion weight and the third balance threshold dynamic fusion weight as the user behavior semantic dynamic interaction fusion coefficient; taking the value obtained by subtracting the user behavior semantic dynamic interaction fusion coefficient from 1 as the semantic dynamic interaction fusion coefficient of the characteristic of the advertisement to be placed; calculating the dot product of the semantic dynamic interaction fusion coefficient of the characteristic of the advertisement to be placed and the semantic encoding vector of the characteristic of the advertisement to be placed to obtain the semantic dynamic interaction fusion item of the characteristic of the advertisement to be placed; calculating the dot product of the user behavior semantic dynamic interaction fusion coefficient and the self-supervised semantic representation vector of the user behavior pattern to obtain the user behavior semantic dynamic interaction fusion item.

[0098] In a specific example, the feature semantic encoding vector of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern are input into the balanced threshold feature vector adaptive fusion network to obtain the advertisement to be placed-user behavior semantic fusion representation vector, including: inputting the feature semantic encoding vector of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern into the balanced threshold feature vector adaptive fusion network to process them with the following adaptive fusion formula to obtain the advertisement to be placed-user behavior semantic fusion representation vector; wherein the adaptive fusion formula is:

[0099]

[0100] in, and are the semantic encoding vector of the feature of the advertisement to be placed and the self-supervisory semantic representation vector of the user behavior pattern, Indicates cascade, represents vector multiplication, It means adding elements by position. represents element-by-element multiplication by position, , and are the first transformation vector, the second transformation vector and the third transformation vector respectively, , and are the first bias parameter, the second bias parameter and the third bias parameter respectively, represents the sigmoid activation function, , and They are the first balance threshold dynamic fusion weight, the second balance threshold dynamic fusion weight and the third balance threshold dynamic fusion weight, , It is the semantic fusion representation vector of the advertisement to be delivered and the user behavior.

[0101] Then, the advertisement to be placed-user behavior semantic fusion representation vector is input into a classifier-based advertisement screening module to obtain a screening result, and the screening result is used to indicate whether to place the candidate advertisement to be placed.

[0102] Accordingly, in step S570, based on the to-be-delivered advertisement-user behavior semantic fusion representation vector, a screening result is determined, including: inputting the to-be-delivered advertisement-user behavior semantic fusion representation vector into a classifier-based advertisement screening module to obtain the screening result, and the screening result is used to indicate whether to deliver the to-be-delivered alternative advertisement.

[0103] It should be understood that the role of the classifier is to use the given categories and known training data to learn classification rules and classifiers, and then classify (or predict) unknown data. Logistic regression, SVM, etc. are often used to solve binary classification problems. For multi-class classification problems, logistic regression or SVM can also be used, but multiple binary classifications are required to form a multi-classification, but this is prone to errors and inefficient. Commonly used multi-classification methods include the Softmax classification function.

[0104] In the technical solution disclosed in the present invention, the user behavior pattern self-supervised semantic representation vector is used to represent the semantic features of the full-time-domain behavior pattern of the user behavior data obtained by significantly fusing the time series of the user behavior data with the semantic coding and the self-supervised feature sequence. The semantic coding vector of the characteristics of the advertisement to be placed is the text semantic feature of the text description of the candidate advertisement to be placed. In particular, the modal differences and dimensional differences between the time series of the user behavior data and the text description of the candidate advertisement to be placed will cause the feature distribution to be inconsistent between the self-supervised semantic representation vector of the user behavior pattern and the feature distribution of the semantic coding vector of the characteristics of the advertisement to be placed, and the advertisement to be placed-user behavior semantic fusion representation vector obtained by adaptively fusion of the balanced threshold feature vector also has a knowledge offset relative to the source domain data distribution, thereby causing uncertainty in its understanding of the class probability relative to the classifier, reducing the speed of classification training and the accuracy of classification results.

[0105] In a preferred embodiment, the semantic fusion representation vector of the advertisement to be placed and the user behavior is input into the classifier-based advertisement screening module to obtain the screening result, including: determining the negative class probability value corresponding to the non-placement of the candidate advertisement to be placed obtained by the semantic fusion representation vector of the advertisement to be placed through the classifier-based advertisement screening module, and subtracting the negative class probability value from 1 to obtain the approval class probability value; respectively calculating the power function of each eigenvalue of the semantic fusion representation vector of the advertisement to be placed and the user behavior to be placed with the negative class probability value as an exponent to obtain the negative class vector of the semantic fusion of the advertisement to be placed; respectively calculating the power function of each eigenvalue of the semantic fusion representation vector of the advertisement to be placed and the approval class value as an exponent to obtain the approval class vector of the semantic fusion of the advertisement to be placed and the approval class value as an exponent to obtain the negative ... The classification probability value is dot-multiplied with the consent class vector of the advertisement to be placed-user behavior semantic fusion to obtain a first intermediate class vector of the advertisement to be placed-user behavior semantic fusion, and the consent class probability value is dot-multiplied with the negative class vector of the advertisement to be placed-user behavior semantic fusion to obtain a second intermediate class vector of the advertisement to be placed-user behavior semantic fusion; after dot-multiplying the first intermediate class vector of the advertisement to be placed-user behavior semantic fusion with the second intermediate class vector of the advertisement to be placed-user behavior semantic fusion, further dot-add the dot product result of the negative class probability value and the negative class vector of the advertisement to be placed-user behavior semantic fusion to obtain an optimized representation vector of the advertisement to be placed-user behavior semantic fusion; the optimized representation vector of the advertisement to be placed-user behavior semantic fusion is input into the classifier-based advertisement screening module to obtain the screening result.

[0106] Based on this, when the applicant of the present disclosure inputs the semantic fusion representation vector of the advertisement to be placed and the user behavior into the advertisement screening module based on the classifier for classification, in order to achieve unsupervised domain adaptation from the feature distribution domain of the semantic fusion representation vector of the advertisement to be placed and the probability distribution domain of the classification probability, the negative class probability value and the consent class probability value obtained by the semantic fusion representation vector of the advertisement to be placed through the advertisement screening module based on the classifier are used as domain proxies, and the moving average of the probability distribution of the semantic fusion representation vector of the advertisement to be placed and the knowledge transfer from the unlabeled classification feature to the labeled probability distribution is achieved by superimposing the feature domain probability distribution knowledge of the semantic fusion representation vector of the advertisement to be placed and the user behavior, so as to promote the classification operation of the semantic fusion representation vector of the advertisement to be placed and the speed of classification training and the accuracy of classification results are improved.

[0107] Based on the above embodiments, see Figure 7, which is a structural diagram of an artificial intelligence-based advertising screening system 800 in an embodiment of the present disclosure. The artificial intelligence-based advertisement screening system 800 includes: a user behavior data acquisition module 810, which is used to acquire the time series of user behavior data; a module 820 for acquiring alternative advertisements to be placed, which is used to acquire text descriptions of alternative advertisements to be placed; a context semantic analysis module 830, which is used to perform context semantic analysis on the time series of the user behavior data to obtain a time series of context user behavior semantic encoding feature vectors; a vector fusion module 840, which is used to input the time series of the context user behavior semantic encoding feature vectors into a self-supervised feature sequence significant fusion network to obtain a user behavior pattern self-supervised semantic representation vector; a semantic encoding module 850, which is used to perform semantic encoding on the text description of the alternative advertisements to be placed to obtain a semantic encoding vector of the characteristics of the advertisements to be placed; a semantic fusion module 860, which is used to input the semantic encoding vector of the characteristics of the advertisements to be placed and the self-supervised semantic representation vector of the user behavior pattern into an equalization threshold feature vector adaptive fusion network to obtain a semantic fusion representation vector of the advertisements to be placed-user behavior; and a screening module 870, which is used to determine a screening result based on the semantic fusion representation vector of the advertisements to be placed-user behavior.

[0108] Those skilled in the art will appreciate that the specific functions and operations of each module in the above-mentioned artificial intelligence-based advertising screening system 800 have been described in detail above. Figures 2 to 6 The description of the artificial intelligence-based advertising screening method has been introduced in detail, and therefore its repeated description will be omitted.

[0109] Figure 8 FIG. 1 is an application scenario diagram of the artificial intelligence-based advertising screening method according to an embodiment of the present disclosure. Figure 8 As shown, in this application scenario, first, the time series of user behavior data is obtained (for example, Figure 8 D1 as shown in the figure) and a text description of the candidate ad to be served (for example, Figure 8 Then, the time series of the user behavior data and the text description of the candidate advertisement to be delivered are input into a server (for example, Figure 8 In S) as shown in , the server is able to use the artificial intelligence-based advertising screening algorithm to process the time series of the user behavior data and the text description of the alternative advertisements to be placed to obtain a screening result indicating whether to place the alternative advertisements to be placed.

[0110] Based on the above embodiments, another exemplary embodiment of an electronic device is also provided in the embodiments of the present disclosure. In some possible implementations, the electronic device in the embodiments of the present disclosure may include a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the steps of the artificial intelligence-based advertising screening method in the above embodiments can be implemented when the processor executes the program.

[0111] For example, the electronic device is disclosed in the present invention Figure 1 Taking the server 100 in the example as an example, the processor in the electronic device is the processor 110 in the server 100, and the memory in the electronic device is the memory 120 in the server 100.

[0112] An embodiment of the present disclosure also provides a computer-readable storage medium. Fig. 9 FIG. 1 is a schematic diagram of a computer-readable storage medium 1000 according to an embodiment of the present disclosure. Fig. 9 As shown, the computer-readable storage medium 1000 stores computer-executable instructions 1001. When the computer-executable instructions 1001 are executed by the processor, the artificial intelligence-based advertising screening method according to the embodiment of the present disclosure described with reference to the above figures can be executed. The computer-readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory may include, for example, a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may include, for example, a read-only memory (ROM), a hard disk, a flash memory, etc.

[0113] The embodiments of the present disclosure also provide a computer program product or a computer program, which includes computer executable instructions, and the computer executable instructions are stored in a computer readable storage medium. The processor of the computer device reads the computer executable instructions from the computer readable storage medium, and the processor executes the computer executable instructions, so that the computer device executes the artificial intelligence-based advertising screening method according to the embodiments of the present disclosure.

[0114] Those skilled in the art will appreciate that the contents disclosed in this disclosure may be subject to various modifications and improvements. For example, the various devices or components described above may be implemented by hardware, or by software, firmware, or a combination of some or all of the three.

[0115] In addition, although the present disclosure makes various references to certain units in the system according to embodiments of the present disclosure, any number of different units may be used and run on the client and / or server. The units are illustrative only, and different aspects of the systems and methods may use different units.

[0116] Those skilled in the art will appreciate that all or part of the steps in the above method can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk. Optionally, all or part of the steps in the above embodiment can also be implemented using one or more integrated circuits. Accordingly, each module / unit in the above embodiment can be implemented in the form of hardware or in the form of a software functional module. The present disclosure is not limited to any particular form of combination of hardware and software.

[0117] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present disclosure belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology and should not be interpreted in an idealized or highly formal sense, unless explicitly defined as such herein.

[0118] The above is an explanation of the present disclosure and should not be considered as a limitation thereof. Although several exemplary embodiments of the present disclosure have been described, it will be readily understood by those skilled in the art that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present disclosure. Therefore, all such modifications are intended to be included within the scope of the present disclosure as defined in the claims. It should be understood that the above is an explanation of the present disclosure and should not be considered as being limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present disclosure is defined by the claims and their equivalents.

Claims

1. An advertisement screening method based on artificial intelligence, characterized in that: include: Get the time series of user behavior data; Get the text description of the candidate advertisement to be delivered; Performing contextual semantic analysis on the time series of the user behavior data to obtain a time series of contextual user behavior semantic encoding feature vectors, which includes: Embedding and encoding each user behavior data in the time series of the user behavior data to obtain a time series of user behavior embedded coding vectors; Inputting the time series of the user behavior embedding coding vector into a user behavior context encoder based on a Transformer layer to obtain the time series of the context user behavior semantic coding feature vector; Inputting the time series of the contextual user behavior semantic encoding feature vector into the self-supervised feature sequence significant fusion network to obtain the user behavior pattern self-supervised semantic representation vector, which includes: Calculating the mean vector of the time series of the contextual user behavior semantic encoding feature vector as the clustering center of the self-supervised sequence distribution; Calculating the user behavior semantic relevance between each context user behavior semantic encoding feature vector in the time series of the context user behavior semantic encoding feature vector and the cluster center of the self-supervised sequence distribution to obtain a sequence of user behavior semantic relevance; Inputting the sequence of user behavior semantic relevance into a softmax function for soft maximum normalization processing to obtain a sequence of self-supervised context user behavior semantic relevance coefficients; Taking the sequence of the self-supervised context user behavior semantic relevance coefficients as a weight sequence, calculating the vector-by-vector weighted sum of the time series of the context user behavior semantic encoding feature vectors to obtain the user behavior pattern self-supervised semantic representation vector; Semantically encoding the text description of the candidate advertisement to be placed to obtain a semantic encoding vector of the characteristic of the advertisement to be placed, which includes: Performing word segmentation processing on the text description of the candidate advertisement to be placed so as to convert the text description of the candidate advertisement to be placed into a word sequence consisting of a plurality of words; Mapping each word in the word sequence to a word vector using a word embedding layer of a semantic encoder including a word embedding layer to obtain a sequence of word vectors; Using the semantic encoder including the word embedding layer, the sequence of word vectors is encoded based on global context semantics to obtain a sequence of encoding vectors of advertisements to be delivered; Cascading the sequence of the to-be-delivered advertisement coding vectors to obtain the to-be-delivered advertisement characteristic semantic coding vector; Inputting the feature semantic encoding vector of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern into the balanced threshold feature vector adaptive fusion network to obtain the advertisement to be placed-user behavior semantic fusion representation vector, which includes: Determining a first balanced threshold dynamic fusion weight, a second balanced threshold dynamic fusion weight, and a third balanced threshold dynamic fusion weight based on the semantic encoding vector of the feature of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern includes: Cascading the semantic encoding vector of the feature of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern to obtain a semantic cascade vector of the advertisement to be placed-user behavior; Calculate the position-by-position addition of the semantic encoding vector of the feature of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern to obtain an element-by-element addition vector of the semantics of the advertisement to be placed-user behavior; Calculate the position-by-position element-by-element multiplication of the semantic encoding vector of the feature of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern to obtain an element-by-element multiplication vector of the semantics of the advertisement to be placed-user behavior; The first transformation vector is multiplied by the advertisement to be placed-user behavior semantic cascade vector, and the resultant is added to the first bias parameter to obtain a cascade coefficient; the second transformation vector is multiplied by the advertisement to be placed-user behavior semantic cascade vector, and the resultant is added to the second bias parameter to obtain an addition coefficient; the third transformation vector is multiplied by the advertisement to be placed-user behavior semantic cascade vector, and the resultant is added to the third bias parameter to obtain a multiplication coefficient; The cascade coefficient, the addition coefficient and the multiplication coefficient are respectively subjected to a sigmoid activation function to obtain the first equalization threshold dynamic fusion weight, the second equalization threshold dynamic fusion weight and the third equalization threshold dynamic fusion weight; Adaptively and dynamically integrating the semantic encoding vector of the feature of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern with the first balance threshold dynamic fusion weight, the second balance threshold dynamic fusion weight, and the third balance threshold dynamic fusion weight to obtain a dynamic interaction fusion item of the semantics of the feature of the advertisement to be placed and a dynamic interaction fusion item of the semantics of the user behavior, which includes: taking the average value of the first balance threshold dynamic fusion weight, the second balance threshold dynamic fusion weight, and the third balance threshold dynamic fusion weight as a user behavior semantic dynamic interaction fusion coefficient; The value of the user behavior semantic dynamic interaction fusion coefficient is subtracted from 1 as the semantic dynamic interaction fusion coefficient of the feature of the advertisement to be placed; Calculate the point multiplication of the semantic dynamic interaction fusion coefficient of the characteristic semantics of the advertisement to be placed and the semantic coding vector of the characteristic semantics of the advertisement to be placed to obtain the semantic dynamic interaction fusion item of the characteristic semantics of the advertisement to be placed; Calculate the point multiplication of the user behavior semantic dynamic interaction fusion coefficient and the user behavior pattern self-supervisory semantic representation vector to obtain the user behavior semantic dynamic interaction fusion term; The dynamic interaction fusion item of the semantics of the characteristic of the advertisement to be placed and the dynamic interaction fusion item of the semantics of the user behavior are fused element by element to obtain the semantics fusion representation vector of the advertisement to be placed-user behavior; Based on the to-be-delivered advertisement-user behavior semantic fusion representation vector, a screening result is determined.

2. The artificial intelligence-based advertising screening method according to claim 1, characterized in that: Calculating the user behavior semantic relevance between each context user behavior semantic encoding feature vector in the time series of the context user behavior semantic encoding feature vector and the cluster center of the self-supervised sequence distribution to obtain a sequence of user behavior semantic relevance, including: Encoding the context user behavior semantic encoding feature vector through a fully connected layer to obtain a context user behavior semantic fully connected encoding feature vector; The vector product between the transposed vector of the self-supervised sequence distribution cluster center and the contextual user behavior semantic fully connected encoding feature vector is calculated to obtain the user behavior semantic relevance.

3. The artificial intelligence-based advertising screening method according to claim 2, characterized in that: Determine the screening result based on the to-be-delivered advertisement-user behavior semantic fusion representation vector, including: The advertisement to be placed-user behavior semantic fusion representation vector is input into a classifier-based advertisement screening module to obtain the screening result, and the screening result is used to indicate whether to place the candidate advertisement to be placed.

4. An advertisement screening system based on artificial intelligence, characterized in that: include: User behavior data acquisition module, used to obtain the time series of user behavior data; A module for obtaining candidate advertisements to be placed, used for obtaining text descriptions of candidate advertisements to be placed; A context semantic analysis module is used to perform context semantic analysis on the time series of the user behavior data to obtain a time series of context user behavior semantic encoding feature vectors, which includes: Embedding and encoding each user behavior data in the time series of the user behavior data to obtain a time series of user behavior embedded coding vectors; Inputting the time series of the user behavior embedding coding vector into a user behavior context encoder based on a Transformer layer to obtain the time series of the context user behavior semantic coding feature vector; A vector fusion module is used to input the time series of the context user behavior semantic encoding feature vector into the self-supervised feature sequence significant fusion network to obtain a user behavior pattern self-supervised semantic representation vector, which includes: Calculating the mean vector of the time series of the contextual user behavior semantic encoding feature vector as the clustering center of the self-supervised sequence distribution; Calculating the user behavior semantic relevance between each context user behavior semantic encoding feature vector in the time series of the context user behavior semantic encoding feature vector and the cluster center of the self-supervised sequence distribution to obtain a sequence of user behavior semantic relevance; Inputting the sequence of user behavior semantic relevance into a softmax function for soft maximum normalization processing to obtain a sequence of self-supervised context user behavior semantic relevance coefficients; Taking the sequence of the self-supervised context user behavior semantic relevance coefficients as a weight sequence, calculating the vector-by-vector weighted sum of the time series of the context user behavior semantic encoding feature vectors to obtain the user behavior pattern self-supervised semantic representation vector; A semantic coding module is used to semantically code the text description of the candidate advertisement to be placed to obtain a semantic coding vector of the characteristics of the advertisement to be placed, which includes: Performing word segmentation processing on the text description of the candidate advertisement to be placed so as to convert the text description of the candidate advertisement to be placed into a word sequence consisting of a plurality of words; Mapping each word in the word sequence to a word vector using a word embedding layer of a semantic encoder including a word embedding layer to obtain a sequence of word vectors; Using the semantic encoder including the word embedding layer, the sequence of word vectors is encoded based on global context semantics to obtain a sequence of encoding vectors of advertisements to be delivered; Cascading the sequence of the to-be-delivered advertisement coding vectors to obtain the to-be-delivered advertisement characteristic semantic coding vector; The semantic fusion module is used to input the semantic encoding vector of the feature of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern into the balanced threshold feature vector adaptive fusion network to obtain the semantic fusion representation vector of the advertisement to be placed-user behavior, and determine the first balanced threshold dynamic fusion weight, the second balanced threshold dynamic fusion weight and the third balanced threshold dynamic fusion weight based on the semantic encoding vector of the feature of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern, which includes: Cascading the semantic encoding vector of the feature of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern to obtain a semantic cascade vector of the advertisement to be placed-user behavior; Calculate the position-by-position addition of the semantic encoding vector of the feature of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern to obtain an element-by-element addition vector of the semantics of the advertisement to be placed-user behavior; Calculate the position-by-position element-by-element multiplication of the semantic encoding vector of the feature of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern to obtain an element-by-element multiplication vector of the semantics of the advertisement to be placed-user behavior; The first transformation vector is multiplied by the advertisement to be placed-user behavior semantic cascade vector, and the resultant is added to the first bias parameter to obtain a cascade coefficient; the second transformation vector is multiplied by the advertisement to be placed-user behavior semantic cascade vector, and the resultant is added to the second bias parameter to obtain an addition coefficient; the third transformation vector is multiplied by the advertisement to be placed-user behavior semantic cascade vector, and the resultant is added to the third bias parameter to obtain a multiplication coefficient; The cascade coefficient, the addition coefficient and the multiplication coefficient are respectively subjected to a sigmoid activation function to obtain the first equalization threshold dynamic fusion weight, the second equalization threshold dynamic fusion weight and the third equalization threshold dynamic fusion weight; Adaptively and dynamically integrate the semantic encoding vector of the feature of the advertisement to be placed and the self-supervised semantic representation vector of the user behavior pattern with the first balance threshold dynamic fusion weight, the second balance threshold dynamic fusion weight and the third balance threshold dynamic fusion weight to obtain a dynamic interaction fusion item of the semantics of the feature of the advertisement to be placed and a dynamic interaction fusion item of the semantics of the user behavior, which includes: taking the average value of the first balance threshold dynamic fusion weight, the second balance threshold dynamic fusion weight and the third balance threshold dynamic fusion weight as a user behavior semantic dynamic interaction fusion coefficient; The value of the user behavior semantic dynamic interaction fusion coefficient is subtracted from 1 as the semantic dynamic interaction fusion coefficient of the feature of the advertisement to be placed; Calculate the point multiplication of the semantic dynamic interaction fusion coefficient of the characteristic semantics of the advertisement to be placed and the semantic coding vector of the characteristic semantics of the advertisement to be placed to obtain the semantic dynamic interaction fusion item of the characteristic semantics of the advertisement to be placed; Calculate the point multiplication of the user behavior semantic dynamic interaction fusion coefficient and the user behavior pattern self-supervisory semantic representation vector to obtain the user behavior semantic dynamic interaction fusion term; The dynamic interaction fusion item of the semantics of the characteristic of the advertisement to be placed and the dynamic interaction fusion item of the semantics of the user behavior are fused element by element to obtain the semantics fusion representation vector of the advertisement to be placed-user behavior; The screening module is used to determine the screening result based on the to-be-delivered advertisement-user behavior semantic fusion representation vector.

Images