Internet advertisement accurate putting system and method based on big data

Abstract

The invention relates to the field of advertisement putting, and particularly discloses an internet advertisement accurate putting system and method based on big data, and the method comprises the steps: firstly obtaining user basic data, user reading behavior data and user reading novel preference data, and then carrying out the feature extraction and correlation analysis of the three data through a deep learning technology, and finally, a personalized advertisement putting strategy is generated through a generator, so that accurate internet advertisement putting is realized, advertisement content is ensured to be related to interests and demands of users, and the click rate and conversion rate of advertisements are improved.

Description

Technical Field

[0001] The present application relates to the field of advertising delivery, and more specifically, to a system and method for precise delivery of Internet advertising based on big data. Background Art

[0002] With the rapid development of the Internet, Internet advertising has gradually become the fourth media after radio, television and print publications. It is favored by advertisers for its wide audience, super-time and space communication and interactivity. Recently, with the popularity of online novels and the emergence of a large number of novel platforms, advertisers have begun to place advertisements on these platforms.

[0003] However, due to the huge number of novels and the large number of users, advertising management faces the problem of blindness and randomness. In the current technology, advertising is delivered to each novel, resulting in a too large target range, without considering whether the novel is read by the current user, resulting in ineffective delivery. In addition, when delivering advertisements for novels read by users, the matching degree between delivery time and advertising content is not considered, which not only reduces the user's reading experience, but also affects the effectiveness of the advertisement.

[0004] Therefore, a system and method for precise delivery of Internet advertisements based on big data is desired. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides a system and method for precise Internet advertising delivery based on big data, which first obtains user basic data, user reading behavior data and user novel reading preference data, and then uses deep learning technology to perform feature extraction and correlation analysis on the three, and finally generates a personalized advertising delivery strategy through a generator, thereby achieving precise Internet advertising delivery, ensuring that the advertising content is relevant to the user's interests and needs, thereby improving the click-through rate and conversion rate of the advertisement.

[0006] According to one aspect of the present application, a system for accurate Internet advertising delivery based on big data is provided, which includes: Internet advertising data acquisition module, used to obtain user basic data, user reading behavior data and user novel reading preference data; An Internet advertising delivery data extraction module is used to extract an optimized user basic data semantic feature vector and a user reading multimodal association feature vector from the user basic data, the user reading behavior data and the user reading novel preference data; The advertisement delivery strategy generation module is used to generate a personalized advertisement delivery strategy based on the optimized user basic data semantic feature vector and the user reading multimodal association feature vector.

[0007] According to another aspect of the present application, a method for accurate delivery of Internet advertisements based on big data is provided, which includes: Obtain user basic data, user reading behavior data and user novel reading preference data; Extracting an optimized user basic data semantic feature vector and a user reading multimodal association feature vector from the user basic data, the user reading behavior data and the user novel reading preference data; A personalized advertising delivery strategy is generated based on the optimized user basic data semantic feature vector and the user reading multimodal association feature vector.

[0008] Compared with the prior art, the present application provides a system and method for precise Internet advertising delivery based on big data. The system and method first obtain basic user data, user reading behavior data, and user novel reading preference data, and then use deep learning technology to perform feature extraction and correlation analysis on the three. Finally, a generator is used to generate a personalized advertising delivery strategy, thereby achieving precise Internet advertising delivery and ensuring that the advertising content is relevant to the user's interests and needs, thereby improving the click-through rate and conversion rate of the advertisement. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 It is a block diagram of a system for precise Internet advertising delivery based on big data according to an embodiment of the present application.

[0011] Figure 2 It is a block diagram of an Internet advertising delivery data extraction module in a big data-based Internet advertising precision delivery system according to an embodiment of the present application.

[0012] Figure 3 A block diagram of a multimodal data feature extraction unit for reading in a big data-based Internet advertising precision delivery system according to an embodiment of the present application.

[0013] Figure 4 This is a block diagram of an advertising delivery strategy generation module in a big data-based Internet advertising precision delivery system according to an embodiment of the present application.

[0014] Figure 5 The present invention is a flowchart of a method for precise Internet advertising delivery based on big data according to an embodiment of the present application.

[0015] Figure 6 is a block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0016] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0017] Figure 1 Schematic diagram of the Internet advertising precision delivery system based on big data in the embodiment of the present application. Figure 1 As shown, according to the big data-based Internet advertising precision delivery system 100 according to the embodiment of the present application, it includes: an Internet advertising delivery data acquisition module 110, which is used to obtain user basic data, user reading behavior data and user novel reading preference data; an Internet advertising delivery data extraction module 120, which is used to extract optimized user basic data semantic feature vectors and user reading multimodal association feature vectors from the user basic data, the user reading behavior data and the user novel reading preference data; an advertising delivery strategy generation module 130, which is used to generate a personalized advertising delivery strategy based on the optimized user basic data semantic feature vector and the user reading multimodal association feature vector.

[0018] In the above-mentioned Internet advertising precision delivery system 100 based on big data, the Internet advertising delivery data acquisition module 110 is used to obtain user basic data, user reading behavior data and user reading novel preference data. It should be understood that with the rapid development of the Internet, Internet advertising has gradually emerged as the fourth largest communication channel after radio, television and print media. With its advantages such as wide coverage, dissemination anytime, anywhere and high interactivity, it has won the favor of advertisers. Especially with the popularity of online novels and the rapid growth of novel platforms, advertisers have placed advertisements on these emerging platforms. However, in the face of a large number of novels and users, advertising delivery management faces many challenges and often appears blind and arbitrary. Current advertising technology often delivers advertisements to all novels, which does not take into account whether users actually read these novels, resulting in a large number of invalid delivery. In addition, even when advertising for novels that users are reading, the degree of matching between the delivery time and content of the advertisement and the novel is not fully considered. This situation not only reduces the user's reading experience, but also significantly affects the effectiveness of the advertisement. Therefore, in the technical solution of this application, basic user data, user reading behavior data, and user novel reading preference data are obtained, and combined with deep learning technology, a personalized advertising delivery strategy is formulated to achieve accurate delivery of Internet advertising. This strategy ensures that the advertising content is closely related to the interests and needs of users, thereby effectively improving the click-through rate and conversion rate of advertisements.

[0019] Specifically, obtaining basic user data, user reading behavior data, and user reading novel preference data is the basis for achieving precise Internet advertising delivery, because these data can comprehensively depict the characteristics and interests of users, thereby helping the advertising system to better understand user needs. Basic user data includes basic information such as age, gender, and geographic location, which helps to determine the basic demographic characteristics of users and provide a basic background for advertising delivery strategies. User reading behavior data, such as browsing history, dwell time, and click records, can reflect the actual interests and behavior patterns of users, revealing the user's preferences for different content and their activity. User reading novel preference data provides information about the user's interest in specific types of literary content. Combining these three types of data, the advertising system can create a detailed user portrait and formulate a personalized advertising strategy by analyzing the user's basic information, behavior patterns, and preferences. This data-driven approach not only improves the accuracy of advertising delivery, but also improves the relevance of advertising and user acceptance, thereby increasing the effectiveness of advertising and the return on investment.

[0020] In the above-mentioned Internet advertising precision delivery system 100 based on big data, the Internet advertising delivery data extraction module 120 is used to extract optimized user basic data semantic feature vectors and user reading multimodal association feature vectors from the user basic data, the user reading behavior data and the user novel reading preference data. In this way, the advertising system can understand the needs and preferences of users more accurately, so as to formulate more targeted advertising strategies. The extraction of this multimodal feature vector enables the advertising system to not only pay attention to the basic information of users, but also deeply explore the personalized needs of users, improve the matching degree between advertising content and users' actual interests, and ultimately improve the effect of advertising delivery and user satisfaction.

[0021] Figure 2 FIG. 1 is a block diagram of an Internet advertising delivery data extraction module in an Internet advertising precision delivery system based on big data according to an embodiment of the present application. Figure 2 As shown, in a specific embodiment of the present application, the Internet advertising delivery data extraction module 120 includes: a user basic data feature extraction unit 121, used to perform feature extraction on the user basic data to obtain the optimized user basic data semantic feature vector; a reading multimodal data feature extraction unit 122, used to perform feature extraction on the user reading behavior data and the user novel reading preference data to obtain the user reading multimodal association feature vector.

[0022] It should be understood that the purpose of extracting features from user basic data to obtain optimized semantic feature vectors of user basic data is to convert the original, structured user information into a high-level abstract form that can be processed by the model. These basic data include basic demographic information such as the user's age, gender, and geographic location. Although these data are relatively simple in themselves, they play an important role in advertising. Through feature extraction, these data are converted into semantic feature vectors, which can better capture and represent the basic attributes of users and their potential needs. This process involves converting these raw data into a digital form that can be understood by machine learning algorithms, such as through embedding technology or neural network encoding. The resulting semantic feature vectors can effectively structure and standardize the user's basic information.

[0023] Furthermore, feature extraction of user reading behavior data and user novel reading preference data is to deeply analyze user interests and behavior patterns, so as to achieve more accurate advertising. User reading behavior data, including browsing history, click-through rate, dwell time, etc., reflects the user's interaction mode and preference in the actual use process. These data can reveal the user's real interest and behavioral tendency for different content, but the original data usually contains a lot of information noise and complex patterns. Directly using these data for advertising may be inefficient and have limited effect. Through feature extraction, these complex behavior data are converted into structured feature vectors, and the user's behavior pattern can be quantified and standardized to make it an input form that the model can process. User novel reading preference data provides detailed information about the user's interest and preference for a specific type of novel. By extracting features from these data, the user's preference for different novel categories, themes or authors can be captured, and semantic feature vectors related to the user's reading interest can be generated. Extracting user reading behavior data and novel reading preference data into multimodal correlation feature vectors can effectively merge the features of the two types of data, so as to more comprehensively capture the user's interests and needs. This multimodal feature vector not only contains the user's behavioral patterns, but also integrates the user's interest preferences, enabling the advertising system to analyze and make decisions in multiple dimensions.

[0024] In a specific embodiment of the present application, the user basic data feature extraction unit 121 includes: passing the user basic data through a user basic data semantic encoder based on a bidirectional long short-term memory neural network to obtain a plurality of user basic data input vectors; passing the plurality of user basic data input vectors through a user basic data semantic feature extractor based on a convolutional neural network to obtain a user basic data semantic feature vector; and performing linear embedding modulation on the user basic data semantic feature vector based on a shift information position expression based on a user reading multimodal association feature vector to obtain an optimized user basic data semantic feature vector.

[0025] It should be understood that although basic user data, such as age, gender, and geographic location, are relatively simple in structure, the interrelationships of these data and their potential impact on user behavior can be better understood through sequence modeling. The Bi-LSTM network, through its bidirectional structure, can process information from both directions of the sequence simultaneously, capturing the contextual information and complex relationships in the basic user data. This method can not only extract the semantic features of the basic user data, but also transform these features into multiple high-dimensional input vectors, providing a richer expression of user information.

[0026] Furthermore, although user basic data (such as age, gender, geographic location, etc.) may seem simple, when converted into vectors, they may contain various complex internal connections. Through its unique convolutional layer, convolutional neural networks can effectively extract local features and important patterns from these input vectors. This operation helps to reduce computational complexity while retaining the core features of the data, making the semantic feature vector of user basic data more refined and effective. The semantic feature vector of user basic data processed by CNN can transform the original user data into a highly abstract and optimized representation. This representation not only covers the basic attributes of the user, but also integrates the complex relationships and patterns in the data, providing richer semantic information. This high-level feature expression enables the advertising system to understand user needs more accurately and formulate more targeted advertising strategies based on these features, thereby improving the effectiveness of advertising and user experience. Specifically, each layer of the user basic data semantic feature extractor based on convolutional neural network is used to perform convolution processing, mean pooling processing based on local feature matrix and nonlinear activation processing on the input data in the forward pass of the layer, so that the last layer of the user basic data semantic feature extractor based on convolutional neural network outputs the user basic data semantic feature vector, wherein the input of the user basic data semantic feature extractor based on convolutional neural network is the multiple user basic data input vectors.

[0027] In particular, in the technical solution of the present application, the basic data of users include static and structured information such as age, gender, and geographic location. These data are processed by a bidirectional long short-term memory neural network and a convolutional neural network, mainly focusing on extracting stable and contextual features from the basic attributes of users, with clear semantics and relatively stable. Relatively speaking, the user reading multimodal association feature vector integrates the user's reading behavior data and reading novel preference data. The processing of these data includes word embedding models, Clip models, and contextual convolutional neural networks, involving text features, semantic associations, and contextual understanding, emphasizing the diversity and dynamics of user behavior patterns and novel preferences. It not only includes user behavior characteristics, but also includes complex semantic information related to reading, with high granularity and contextual changes. This feature processing difference leads to inconsistent distribution of the user basic data semantic feature vector and the user reading multimodal association feature vector in the feature space. In the process of fusing the user basic data semantic feature vector and the user reading multimodal association feature vector to obtain the feature vector generated by the advertising delivery strategy, due to the distribution difference of the feature manifold in the high-dimensional space, abnormal stacking on the boundary may occur. Therefore, in the technical solution of the present application, based on the user reading multimodal association feature vector, the user basic data semantic feature vector is subjected to linear embedding modulation based on the position expression of the shift information to obtain an optimized user basic data semantic feature vector.

[0028] Among them, based on the user reading multimodal associated feature vector, the user basic data semantic feature vector is subjected to linear embedding modulation based on the position expression of the shift information to obtain an optimized user basic data semantic feature vector, including: calculating the position difference between the user basic data semantic feature vector and the user reading multimodal associated feature vector to obtain an Internet advertising shift position difference representation vector; after cascading the user basic data semantic feature vector and the user reading multimodal associated feature vector into an Internet advertising joint feature vector, the Internet advertising joint feature vector is input into a one-dimensional convolutional layer to obtain an Internet advertising shift convolutional coding representation vector; after using the first Internet advertising linear embedding weight matrix to perform Internet advertising linear embedding modulation coding on the user basic data semantic feature vector to obtain a user basic data semantic feature vector linear modulation vector, the user basic data semantic feature vector is subjected to Internet advertising linear embedding modulation coding, and ... The linear modulation vector of the user basic data semantic feature vector and the user reading multimodal associated feature vector are differentially positioned to obtain an Internet advertisement shift modulation differential representation vector; the Internet advertisement shifted position differential representation vector is used as a value vector, the Internet advertisement shifted convolutional coding representation vector and the Internet advertisement shift modulation differential representation vector are used as a query vector and a key vector, and the Internet advertisement shifted position differential representation vector, the Internet advertisement shifted convolutional coding representation vector and the Internet advertisement shift modulation differential representation vector are input into a shift corrector based on an imitation transformer structure to obtain an Internet advertisement multi-scale hierarchical fusion shift modulation weight vector; the position point multiplication between the Internet advertisement multi-scale hierarchical fusion shift modulation weight vector and the user basic data semantic feature vector is calculated to obtain the optimized user basic data semantic feature vector.

[0029] The following optimization formula is used to express that the user basic data semantic feature vector is subjected to linear embedding modulation based on the position expression of the shift information based on the user reading multimodal association feature vector to obtain an optimized user basic data semantic feature vector; Wherein, the optimization formula is: ; ; ; ; ;in, Represents the Internet advertising shift by position difference representation vector, Represents the semantic feature vector of user basic data, Represents the multimodal correlation feature vector of user reading, represents the Internet advertising shift convolutional coding representation vector, represents the Internet advertising shift modulation differential representation vector, represents the first Internet advertisement linear embedding weight matrix, represents the Internet advertising modulation bias vector, represents the multi-scale hierarchical fusion shift modulation weight vector of Internet advertising, represents the transposed vector of the Internet advertising shift modulation differential representation vector, Indicates the length of the Internet advertising shift convolutional coding representation vector represents the difference by position, represents a one-dimensional convolutional layer, represents cascading the user basic data semantic feature vector and the user reading multimodal association feature vector, represents the linear rectification function, represents matrix multiplication, represents the normalized exponential function, It means point multiplication by position. Represents the optimized semantic feature vector of user basic data.

[0030] In the technical solution of the present application, in the process of fusing the semantic feature vector of the user basic data and the multimodal association feature vector of the user reading, the fine-grained inconsistency of the feature distribution between the semantic feature vector of the user basic data and the multimodal association feature vector of the user reading is taken into consideration, which will cause the fused advertising delivery strategy generation feature vector to have abnormal stacking on the boundary of the feature manifold. Based on this, in the technical solution of the present application, before fusing the semantic feature vector of the user basic data and the multimodal association feature vector of the user reading to obtain the advertising delivery strategy generation feature vector, the linear embedding modulation based on the shift information position expression is performed on the semantic feature vector of the user basic data based on the multimodal association feature vector of the user reading.

[0031] In the technical solution of the present application, a linear embedding modulation method based on the position expression of shifted information is proposed to address the problem of fine-grained inconsistency in feature distribution that occurs in the process of fusing the semantic feature vector of user basic data and the multimodal association feature vector of user reading. The method aims to reduce the abnormal stacking phenomenon of feature vectors generated by advertising delivery strategies on the boundary of feature manifolds, so as to improve the robustness and classification performance of the model.

[0032] Specifically, the method first calculates the element-by-element difference between the user basic data semantic feature vector and the user reading multimodal association feature vector to obtain the Internet advertising shift element-by-element differential representation vector, thereby capturing the difference between the two feature vectors. Then, the user basic data semantic feature vector and the user reading multimodal association feature vector are cascaded into an Internet advertising joint feature vector and input into a one-dimensional convolution layer to generate an Internet advertising shift convolution coding representation vector. That is, the deep-level shift implicit information between the user basic data semantic feature vector and the user reading multimodal association feature vector is captured by one-dimensional convolution coding. At the same time, the user basic data semantic feature vector is subjected to Internet advertising linear embedding modulation coding using the first Internet advertising linear embedding weight matrix to be a user basic data semantic feature vector linear modulation vector, and the element-by-element difference between the user basic data semantic feature vector linear modulation vector and the user reading multimodal association feature vector feature vector is calculated to obtain an Internet advertising shift modulation differential representation vector. The Internet advertisement shift modulation differential representation vector is used to represent the positional expression of the shift information of the user basic data semantic feature vector and the user reading multimodal association feature vector in the linear space. That is, in the technical solution of the present application, the Internet advertisement shift position difference, the Internet advertisement shift convolutional coding representation vector and the Internet advertisement shift modulation differential representation vector respectively represent the explicit expression of the fine-grained distribution difference shift information between the user basic data semantic feature vector and the user reading multimodal association feature vector in different feature spaces.

[0033] Next, the Internet advertisement shifted element-by-element differential representation vector is used as the value vector, the Internet advertisement shifted convolutional coding representation vector and the Internet advertisement shifted modulation differential representation vector are used as the query vector and the key vector respectively, and these vectors are input into the shift corrector based on the imitation transformer structure to generate the Internet advertisement multi-scale hierarchical fusion shift modulation weight vector. In this way, the importance of features of different scales is evaluated by the attention mechanism, that is, the attention mechanism is used to simulate the hierarchical precipitation expression of the fine-grained difference feature distribution between the user basic data semantic feature vector and the user reading multimodal association feature vector in different feature spaces. Finally, the element-by-element product of the Internet advertisement multi-scale hierarchical fusion shift modulation weight vector and the user basic data semantic feature vector is calculated to obtain the optimized user basic data semantic feature vector, that is, the weighted operation is used to assign weights to each element in the user basic data semantic feature vector, which reduces the influence of local disturbances and improves the quality of the feature vector.

[0034] In summary, in the method, the Internet advertisement shift position difference, the Internet advertisement shift convolutional coding representation vector and the Internet advertisement shift modulation difference representation vector respectively represent the explicit expression of the fine-grained distribution difference shift information between the user basic data semantic feature vector and the user reading multimodal association feature vector in different feature spaces. The generation of the Internet advertisement multi-scale hierarchical fusion shift modulation weight vector uses the attention mechanism to evaluate the importance of features at different scales and provide weight information for feature shift correction. Ultimately, the generation of the optimized user basic data semantic feature vector reduces the impact of disturbances, improves the model's robustness to local disturbances in the input data, enhances the model's adaptability to noisy or variant data sets, improves the fusion quality of feature vectors, and improves the overall classification performance and training efficiency of the model.

[0035] Figure 3 FIG. 1 is a block diagram of a multimodal data feature extraction unit in a big data-based Internet advertising precision delivery system according to an embodiment of the present application. Figure 3 As shown, in a specific embodiment of the present application, the reading multimodal data feature extraction unit 122 includes: a reading behavior data feature encoding subunit 1221, used to feature encode the user reading behavior data to obtain a user reading behavior text feature vector; a reading novel preference data feature encoding subunit 1222, used to feature encode the user reading novel preference data to obtain a user reading novel preference semantic association feature vector; a reading multimodal feature association subunit 1223, used to associate the user reading behavior text feature vector with the user reading novel preference semantic association feature vector to obtain the user reading multimodal association feature vector.

[0036] It should be understood that the process of feature encoding can condense these complex and diverse behavior information into a form that can be processed by machine learning models by converting user reading behavior data into feature vectors. Feature encoding usually involves converting reading behavior data into numerical features and representing them as points in a high-dimensional feature space through vectorization technology. In this way, information such as user reading preferences, interest intensity, and behavior patterns can be integrated into a unified representation to provide a richer user portrait.

[0037] Furthermore, user novel reading preference data usually includes user reading records and evaluations of different novel types, authors, and themes. These data may be scattered in their original form and difficult to be directly used for analysis, so they need to be converted into highly semantic feature vectors through feature encoding. The feature encoding process includes extracting and converting user reading behavior and preference information, representing them as feature vectors, so that these vectors can accurately reflect the user's interests and preferences. By encoding information such as the types of novels read by users, favorite authors, and reading frequency, these data can be converted into multidimensional features. These features can include the user's reading time for a specific type of novel, the evaluation score for a specific author, the strength of preference for certain themes, etc.

[0038] Furthermore, associating the text feature vector of user reading behavior with the semantic association feature vector of user reading novel preference to obtain the multimodal association feature vector of user reading is to fully and deeply understand the user's reading habits and preferences, so as to provide more accurate and personalized services. This process involves fusing user feature information from different sources to form a comprehensive and multi-dimensional user feature representation. Among them, the text feature vector of user reading behavior usually includes various behavioral data of users during the reading process, such as reading time, click records, and dwell time. These feature vectors can reflect the actual reading habits and behavior patterns of users. However, although these vectors provide information about user behavior, they may not be able to fully capture the deep semantics of user preferences and interests. On the other hand, the semantic association feature vector of user reading novel preference is obtained by encoding the user's preferences for different novel types, authors, and themes. These feature vectors can accurately reflect the user's reading preferences and interests, but may lack the description of the specific details and dynamic changes of the actual reading behavior. By associating these two feature vectors, the user's reading behavior and preference data can be combined to form a more comprehensive and accurate user portrait. This comprehensive feature vector can not only show the user's preference for a certain type of novel, but also be combined with the user's actual reading behavior, such as reading frequency, reading time, etc., to better reveal the user's real needs and interests.

[0039] In a specific embodiment of the present application, the reading behavior data feature encoding subunit 1221 includes: embedding the user reading behavior data into a user reading behavior word model to obtain a plurality of user reading behavior text data items; and passing the plurality of user reading behavior text data items through a user reading behavior text editor based on a Clip model to obtain the user reading behavior text feature vector.

[0040] It should be understood that user reading behavior data usually contains information such as user click records, reading time, dwell time, article type, etc. These data may be unstructured in their original state and difficult to use directly in machine learning models. Through the word embedding model, these behavior data are converted into word vectors (or called embedding vectors), which can be mapped to a high-dimensional vector space, so that these data have better semantic expression ability and computational efficiency. Specifically, word embedding models (such as Word2Vec, GloVe or BERT, etc.) can map each item in the user behavior data (such as different articles, click actions, reading time, etc.) into a vector of fixed dimension through training. This process not only converts the original behavior data into numerical representation, but also captures the semantic relationship and contextual information between data items.

[0041] Furthermore, multiple user reading behavior text data items are processed through a user reading behavior text editor based on the Clip (Contrastive Language-Image Pretraining) model in order to map user behavior data into feature representations with rich semantic information through an advanced multimodal learning method. This process helps to improve the accuracy and effectiveness of user behavior analysis, thereby enhancing the performance of personalized recommendation systems. Among them, the Clip model was originally designed to handle multimodal learning tasks of image and text data. Its core advantage is that it can map image and text information into a shared vector space. With this technology, users' reading behavior data (such as reading time, click records, article types, etc.) can be aligned with related text descriptions to obtain a comprehensive, multi-dimensional feature representation. Specifically, the editor first pairs the user's behavior data items with the corresponding text descriptions and maps them into a shared semantic space. Specifically, the embedding layer of the Clip model-based user reading behavior text editor is used to map each of the multiple user reading behavior text data items into a word embedding vector to obtain a sequence of reading behavior word embedding vectors; the converter-based Bert model of the Clip model-based user reading behavior text editor is used to perform global contextual semantic encoding on the sequence of reading behavior word embedding vectors to obtain multiple reading behavior feature vectors; and, the multiple reading behavior feature vectors are cascaded to obtain the user reading behavior text feature vector.

[0042] In a specific embodiment of the present application, the novel reading preference data feature encoding subunit 1222 includes: passing the user novel reading preference data through a user novel reading preference semantic encoder based on a context convolutional neural network to obtain a plurality of user novel reading preference semantic feature vectors; and cascading the plurality of user novel reading preference semantic feature vectors into the user novel reading preference semantic association feature vector.

[0043] It should be understood that user novel reading preference data usually includes user evaluations and interaction records of different novel types, authors, themes, etc. These data are often in text form, such as user book reviews, ratings and comments. These text data contain rich semantic information, but may be difficult to use directly for model training and analysis in the original state. Through the encoder based on the context convolutional neural network, these unstructured text data can be effectively converted into structured feature representations. Specifically, the embedding layer of the user novel reading preference semantic encoder based on the context convolutional neural network is used to convert the user novel reading preference data into a user preference embedding vector to obtain a sequence of user preference embedding vectors; the converter-based Bert model of the user novel reading preference semantic encoder based on the context convolutional neural network is used to perform global context semantic encoding on the sequence of user preference embedding vectors to obtain multiple user novel reading preference semantic feature vectors.

[0044] Furthermore, multiple user novel reading preference semantic feature vectors are usually extracted by encoding the user's reading behavior, evaluation and comment data to extract the user's preference information for different novel types or themes. However, these feature vectors may not fully reflect the user's complex preferences when analyzed separately. For example, a feature vector may reflect the user's interest in a certain type of novel, but does not take into account the user's special preference for a specific author or plot. Therefore, cascading multiple feature vectors can integrate the user's multiple preference dimensions into a unified representation. The cascading operation involves connecting different feature vectors (such as interest in different novel types, preference for specific themes, etc.) in sequence to form a new, longer feature vector. This method not only retains the information contained in each original feature vector, but also integrates the relationship between different features, thereby providing a richer description of user preferences.

[0045] In the above-mentioned Internet advertising precision delivery system 100 based on big data, the advertising delivery strategy generation module 130 is used to generate a personalized advertising delivery strategy based on the optimized user basic data semantic feature vector and the user reading multimodal association feature vector. It should be understood that the optimized user basic data semantic feature vector usually includes basic information such as the user's age, gender, geographic location, occupation, etc., which provides the user's static background and basic characteristics. The user reading multimodal association feature vector captures the user's dynamic interests and behavior patterns by encoding the user's reading behavior on different media (such as text, pictures, videos, etc.). Combining these two types of feature vectors can form a comprehensive and multidimensional user portrait, and this combination enables the advertising delivery strategy to more accurately locate the user's points of interest.

[0046] Figure 4 FIG. 1 is a block diagram of an advertisement delivery strategy generation module in an Internet advertisement precision delivery system based on big data according to an embodiment of the present application. Figure 4 As shown, in a specific embodiment of the present application, the advertising delivery strategy generation module 130 includes: an advertising delivery feature fusion unit 131, which is used to fuse the optimized user basic data semantic feature vector and the user reading multimodal association feature vector to obtain an advertising delivery strategy generation feature vector; a personalized advertising delivery strategy generation unit 132, which is used to pass the advertising delivery strategy generation feature vector through a generator to generate a personalized advertising delivery strategy.

[0047] It should be understandable that the fusion of the optimized semantic feature vector of user basic data and the multimodal association feature vector of user reading can realize the construction of a more comprehensive user portrait to better reflect the overall needs and preferences of users and provide a more accurate basis for the formulation of advertising delivery strategies.

[0048] Furthermore, the ad delivery strategy generation feature vector is obtained by fusing the user's basic data semantic feature vector and the reading multimodal association feature vector. These feature vectors integrate the user's static background information and dynamic interest data, providing a comprehensive basis for generating personalized ad strategies. As a powerful model, the generator can accept these feature vectors as input and generate corresponding ad strategies based on the user's unique needs and preferences. The generator usually uses deep learning techniques such as generative adversarial networks (GANs) or variational autoencoders (VAEs), which can learn from complex input data and generate high-quality outputs. In the process of generating ad delivery strategies, the generator can generate personalized and targeted ad content by learning the feature vectors. This method uses the powerful capabilities of the generative model to achieve automatic generation and optimization of ad content, thereby improving the relevance and delivery effect of ads. Ultimately, this intelligently generated ad strategy not only improves the user experience, but also enhances the advertiser's return on investment, making ad delivery more accurate and efficient.

[0049] In summary, the embodiment of the present application first obtains user basic data, user reading behavior data and user novel reading preference data, and then uses deep learning technology to perform feature extraction and correlation analysis on the three, and finally generates a personalized advertising delivery strategy through a generator, thereby achieving precise delivery of Internet advertisements, ensuring that the advertising content is relevant to the user's interests and needs, thereby improving the click-through rate and conversion rate of advertisements.

[0050] As described above, the Internet advertising precision delivery system 100 based on big data according to the embodiment of the present application can be implemented in various terminal devices. In one example, the Internet advertising precision delivery system 100 based on big data can be integrated into the terminal device as a software module and / or a hardware module. For example, the Internet advertising precision delivery system 100 based on big data can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the Internet advertising precision delivery system 100 based on big data can also be one of the many hardware modules of the terminal device.

[0051] Alternatively, in another example, the Internet advertising precision delivery system 100 based on big data and the terminal device may also be separate devices, and the Internet advertising precision delivery system 100 based on big data may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0052] Figure 5 Flow chart of a method for accurate Internet advertising delivery based on big data according to an embodiment of the present application. Figure 5As shown, according to the method for precise Internet advertising delivery based on big data according to the embodiment of the present application, it includes: S110, acquiring user basic data, user reading behavior data and user novel reading preference data; S120, extracting optimized user basic data semantic feature vectors and user reading multimodal association feature vectors from the user basic data, the user reading behavior data and the user novel reading preference data; S130, generating a personalized advertising delivery strategy based on the optimized user basic data semantic feature vectors and the user reading multimodal association feature vectors.

[0053] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned method for accurate Internet advertising delivery based on big data have been referred to above. Figures 1 to 4 The invention has been introduced in detail in the description of the Internet advertising precision delivery system based on big data, and therefore, its repeated description will be omitted.

[0054] Below, reference Figure 6 To describe an electronic device according to an embodiment of the present application.

[0055] See also Figure 6 , Figure 6 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes: The processor 11 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solution provided in the embodiment of the present application; the memory 12 can be implemented in the form of a read-only memory (Read Only Memory, ROM), a static storage device, a dynamic storage device or a random access memory (Random Access Memory, RAM), etc.

[0056] The memory 12 can store an operating system and other application programs. When the technical solution provided in the embodiments of this specification is implemented by software or firmware, the relevant program code is stored in the memory 12, and the processor 11 calls and executes the Internet advertising precision delivery method based on big data in the embodiments of this application; Input / output interface 13, used to implement information input and output; The communication interface 14 is used to realize the communication interaction between the device and other devices. The communication can be realized through a wired manner (such as USB, network cable, etc.) or a wireless manner (such as mobile network, WIFI, Bluetooth, etc.); A bus 15 that transmits information between the various components of the device (e.g., the processor 11, the memory 12, the input / output interface 13, and the communication interface 14); The processor 11 , the memory 12 , the input / output interface 13 and the communication interface 14 are connected to each other in communication within the device via a bus 15 .

[0057] An embodiment of the present application also provides a storage medium, which is a computer-readable storage medium. The storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned method for precise Internet advertising delivery based on big data is implemented.

[0058] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0059] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0060] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0061] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0062] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0063] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0064] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0065] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0066] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0067] In addition, each functional unit in each embodiment of the present application can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit. If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method of each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs. The preferred embodiment of the embodiment of the present application is described above with reference to the accompanying drawings, and the scope of rights of the embodiment of the present application is not limited thereto. Any modifications, equivalent substitutions and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall be within the scope of rights of the embodiments of the present application.

Claims

1. A system for accurate Internet advertising delivery based on big data, characterized in that: include: Internet advertising data acquisition module, used to obtain user basic data, user reading behavior data and user novel reading preference data; An Internet advertising delivery data extraction module is used to extract an optimized user basic data semantic feature vector and a user reading multimodal association feature vector from the user basic data, the user reading behavior data and the user reading novel preference data; The advertisement delivery strategy generation module is used to generate a personalized advertisement delivery strategy based on the optimized user basic data semantic feature vector and the user reading multimodal association feature vector.

2. The Internet advertising precision delivery system based on big data according to claim 1 is characterized in that: The Internet advertisement delivery data extraction module includes: A user basic data feature extraction unit, used for performing feature extraction on the user basic data to obtain the optimized user basic data semantic feature vector; The reading multimodal data feature extraction unit is used to extract features from the user reading behavior data and the user novel reading preference data to obtain the user reading multimodal correlation feature vector.

3. The Internet advertising precision delivery system based on big data according to claim 2 is characterized in that: The user basic data feature extraction unit comprises: The user basic data is passed through a user basic data semantic encoder based on a bidirectional long short-term memory neural network to obtain a plurality of user basic data input vectors; Passing the plurality of user basic data input vectors through a user basic data semantic feature extractor based on a convolutional neural network to obtain a user basic data semantic feature vector; Based on the user reading multimodal association feature vector, the user basic data semantic feature vector is subjected to linear embedding modulation based on the position expression of the shift information to obtain the optimized user basic data semantic feature vector.

4. The Internet advertising precision delivery system based on big data according to claim 3 is characterized in that: Based on the user reading multimodal association feature vector, the user basic data semantic feature vector is subjected to linear embedding modulation based on the position expression of the shift information to obtain an optimized user basic data semantic feature vector, including: Calculating the position difference between the user basic data semantic feature vector and the user reading multimodal association feature vector to obtain an Internet advertisement shift position difference representation vector; After concatenating the user basic data semantic feature vector and the user reading multimodal association feature vector into an Internet advertisement joint feature vector, the Internet advertisement joint feature vector is input into a one-dimensional convolutional layer to obtain an Internet advertisement shift convolutional coding representation vector; After performing Internet advertisement linear embedding modulation encoding on the user basic data semantic feature vector using the first Internet advertisement linear embedding weight matrix to obtain a user basic data semantic feature vector linear modulation vector, calculating the position difference between the user basic data semantic feature vector linear modulation vector and the user reading multimodal association feature vector to obtain an Internet advertisement shift modulation difference representation vector; Taking the Internet advertisement shifted position differential representation vector as a value vector, the Internet advertisement shifted convolutional coding representation vector and the Internet advertisement shifted modulation differential representation vector as a query vector and a key vector, the Internet advertisement shifted position differential representation vector, the Internet advertisement shifted convolutional coding representation vector and the Internet advertisement shifted modulation differential representation vector are input into a shift corrector based on an imitation transformer structure to obtain an Internet advertisement multi-scale hierarchical fusion shift modulation weight vector; The optimized user basic data semantic feature vector is obtained by calculating the point-by-point multiplication between the Internet advertisement multi-scale layered fusion shift modulation weight vector and the user basic data semantic feature vector.

5. The Internet advertising precision delivery system based on big data according to claim 4 is characterized in that: The reading multimodal data feature extraction unit comprises: A reading behavior data feature encoding subunit, used for performing feature encoding on the user reading behavior data to obtain a user reading behavior text feature vector; A novel reading preference data feature encoding subunit, used for performing feature encoding on the user's novel reading preference data to obtain a user's novel reading preference semantic association feature vector; The reading multimodal feature association subunit is used to associate the user reading behavior text feature vector with the user reading novel preference semantic association feature vector to obtain the user reading multimodal association feature vector.

6. The Internet advertising precision delivery system based on big data according to claim 5 is characterized in that: The reading behavior data feature encoding subunit includes: The user reading behavior data is embedded into a user reading behavior word model to obtain a plurality of user reading behavior text data items; The plurality of user reading behavior text data items are passed through a user reading behavior text editor based on a Clip model to obtain the user reading behavior text feature vector.

7. The Internet advertising precision delivery system based on big data according to claim 6 is characterized in that: The novel reading preference data feature encoding subunit includes: Passing the user novel reading preference data through a user novel reading preference semantic encoder based on a contextual convolutional neural network to obtain a plurality of user novel reading preference semantic feature vectors; The multiple user novel reading preference semantic feature vectors are concatenated into the user novel reading preference semantic association feature vector.

8. The Internet advertising precision delivery system based on big data according to claim 7 is characterized in that: The advertisement delivery strategy generation module includes: An advertisement placement feature fusion unit, used for fusing the optimized user basic data semantic feature vector and the user reading multimodal association feature vector to obtain an advertisement placement strategy generation feature vector; The personalized advertisement delivery strategy generating unit is used to generate a personalized advertisement delivery strategy by passing the advertisement delivery strategy generating feature vector through a generator.

9. A method for accurate Internet advertising delivery based on big data, characterized in that: include: Obtain user basic data, user reading behavior data and user novel reading preference data; Extracting an optimized user basic data semantic feature vector and a user reading multimodal association feature vector from the user basic data, the user reading behavior data and the user novel reading preference data; A personalized advertising delivery strategy is generated based on the optimized user basic data semantic feature vector and the user reading multimodal association feature vector.

10. The method for accurate Internet advertising delivery based on big data according to claim 9, characterized in that: Extracting an optimized user basic data semantic feature vector and a user reading multimodal association feature vector from the user basic data, the user reading behavior data and the user novel reading preference data, including: Performing feature extraction on the user basic data to obtain the optimized user basic data semantic feature vector; Feature extraction is performed on the user reading behavior data and the user novel reading preference data to obtain the user reading multimodal association feature vector.

Images