Visual information pushing method and system for public health science popularization

Abstract

The invention relates to the technical field of visual information pushing, and particularly discloses a visual information pushing method and system for public health science popularization, which creates a detailed user portrait by analyzing online behavior data of a user. And then, semantic search query is carried out by utilizing the user portrait and each science popularization content in a public health science popularization database, so that public health science popularization semantic query response feature representation meeting user interests and requirements is matched, and a public health science popularization visual expression image is generated based on the public health science popularization semantic query response feature representation and is pushed to a target user object. The personalized public health science popularization visual information pushing method can ensure that the pushed content is not only matched with the interest of the user, but also adapts to the acceptance of the user in cognition degree, so that not only are the correlation and pertinence of information pushing improved, but also the user experience is enhanced, the public health science popularization effect is improved, and the user experience is improved. And improvement of health of the whole people is promoted.

Description

Technical Field

[0001] The present application relates to the technical field of visual information push technology, and more specifically, to a method and system for pushing visual information for public health popularization. Background Art

[0002] With the popularization of the Internet and mobile devices, the speed and scope of information dissemination have been unprecedentedly expanded. However, this has also brought about the problem of information overload, especially in the field of public health. The public often finds it difficult to distinguish the authenticity of massive amounts of information, and it becomes difficult to obtain knowledge that suits their needs. In addition, due to the different cognitive levels, cultural backgrounds, interests, and health challenges that different groups of people may face, how to effectively push scientific, accurate, and easy-to-understand public health science information to the target audience has become an urgent problem to be solved.

[0003] Therefore, a visual information push solution for public health popularization is desired. Summary of the invention

[0004] The present application provides a method and system for pushing visual information for public health popularization, which can ensure that the pushed content not only matches the user's interests, but also adapts to their acceptance ability in terms of cognitive level. It not only improves the relevance and pertinence of information push, but also enhances the user experience, helps to improve the effect of public health popularization, and promotes the improvement of the health literacy of the whole people.

[0005] In a first aspect, a method for pushing visual information for public health popularization is provided, comprising:

[0006] Obtaining a user profile of the target user object to be pushed, wherein the user profile of the target user object to be pushed includes basic attributes, interests, and health challenges faced;

[0007] Extracting multiple preliminary matching public health popular science contents from the public health popular science database;

[0008] Embedding and encoding the plurality of preliminary matched public health science popularization contents to obtain a plurality of public health science popularization contents embedding and encoding features;

[0009] Embedding the user portrait of the target user object to be pushed to obtain an embedded coding feature of the target user portrait;

[0010] The multiple public health science popularization content embedding coding features and the target user portrait embedding coding features are subjected to feature dynamic semantic search query processing to obtain coding features that meet the user's public health science popularization semantic query response, including: determining a selection range ratio based on the multiple public health science popularization content embedding coding features and the target user portrait embedding coding features; based on the selection range ratio, the multiple public health science popularization content embedding coding features and the target user portrait embedding coding features are subjected to dynamic semantic search coding to obtain coding features that meet the user's public health science popularization semantic query response;

[0011] Generate visual information based on the coding features of the response to the public health science semantic query that meets the user to obtain a public health science visual expression image;

[0012] The public health science popularization visual expression image is pushed to the target user object to be pushed so as to be displayed on the terminal device of the target user object to be pushed.

[0013] In the second aspect, a visual information push system for public health popularization is provided, comprising:

[0014] A user portrait acquisition module, used to acquire a user portrait of a target user object to be pushed, wherein the user portrait of the target user object to be pushed includes basic attributes, points of interest, and health challenges faced;

[0015] A preliminary matching public health science popularization content extraction module is used to extract a plurality of preliminary matching public health science popularization contents from a public health science popularization database;

[0016] A public health science popularization content embedding coding module, used for embedding and coding the plurality of preliminary matching public health science popularization contents to obtain a plurality of public health science popularization content embedding coding features;

[0017] A target user portrait embedding coding module is used to embed the user portrait of the target user object to be pushed to obtain a target user portrait embedding coding feature;

[0018] A feature dynamic semantic search query processing module is used to perform feature dynamic semantic search query processing on the multiple public health science popularization content embedding coding features and the target user portrait embedding coding features to obtain coding features that meet the user's public health science popularization semantic query response, including: a selection range ratio determination unit, used to determine the selection range ratio based on the multiple public health science popularization content embedding coding features and the target user portrait embedding coding features; a dynamic semantic search coding unit, used to perform dynamic semantic search coding on the multiple public health science popularization content embedding coding features and the target user portrait embedding coding features based on the selection range ratio to obtain the coding features that meet the user's public health science popularization semantic query response;

[0019] A visual information generation module, used to generate visual information based on the coding features of the response to the user's public health science semantic query to obtain a public health science visual expression image;

[0020] The visual expression image pushing module is used to push the public health science popularization visual expression image to the target user object to be pushed so as to display it on the terminal device of the target user object to be pushed.

[0021] The present application provides a method and system for pushing visual information for public health science popularization, which can create a detailed user portrait by analyzing the user's online behavior data, including social media interaction records and search history, and then understand the specific needs and preferences of each user. Then, this user portrait is used to perform semantic search queries with various popular science contents in the public health science popularization database to match the public health science popularization semantic query response feature representation that meets the user's interests and needs, and based on this, a public health science popularization visual expression image is generated to push it to the target user object. This personalized public health science popularization visual information push method can ensure that the pushed content not only matches the user's interests, but also adapts to their receptive ability in terms of cognitive level, which not only improves the relevance and pertinence of information push, but also enhances the user experience, which helps to improve the effect of public health science popularization and promote the improvement of health literacy for all people. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application are briefly introduced below. Obviously, the drawings described below only relate to some embodiments of the present application, and are not intended to limit the present application.

[0023] Figure 1 This is a schematic flowchart of a method for pushing visual information for public health popularization according to an embodiment of the present application.

[0024] Figure 2This is a schematic flowchart of step S1 in the visual information push method for public health popularization according to an embodiment of the present application.

[0025] Figure 3 This is a schematic flowchart of step S5 in the visual information push method for public health popularization according to an embodiment of the present application.

[0026] Figure 4 This is a schematic flowchart of step S51 in the visual information push method for public health popularization according to an embodiment of the present application.

[0027] Figure 5 This is a schematic flowchart of step S52 in the visual information push method for public health popularization according to an embodiment of the present application.

[0028] Figure 6 This is a schematic block diagram of a visual information push system for public health popularization according to an embodiment of the present application. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

[0030] In response to the above technical problems, in the technical solution of this application, a method for pushing visual information for public health science is proposed, which can use advanced information technology such as big data analysis, machine learning, artificial intelligence to generate content, natural language processing, etc., to achieve customized visual information push. Specifically, by analyzing the user's online behavior data, including social media interaction records and search history, a detailed user portrait can be created to understand the specific needs and preferences of each user. Then, this user portrait is used to perform semantic search queries with various popular science contents in the public health science database to match the public health science semantic query response feature representation that meets the user's interests and needs, and based on this, a public health science visual expression image is generated to push it to the target user object. This personalized public health science visual information push method can ensure that the pushed content not only matches the user's interests, but also adapts to their receptive ability in terms of cognitive level, which not only improves the relevance and pertinence of information push, but also enhances the user experience, which helps to improve the effect of public health science popularization and promote the improvement of health literacy for all people.

[0031] Specifically, in the technical solution of this application, if Figure 1As shown, the visual information push method for public health popular science includes: S1, obtaining a user portrait of the target user object to be pushed, wherein the user portrait of the target user object to be pushed includes basic attributes, points of interest and health challenges faced; S2, extracting multiple preliminary matching public health popular science contents from a public health popular science database; S3, embedding and encoding the multiple preliminary matching public health popular science contents to obtain multiple public health popular science content embedding coding features; S4, embedding and encoding the user portrait of the target user object to be pushed to obtain the target user portrait embedding coding features; S5, performing feature dynamic semantic search query processing on the multiple public health popular science content embedding coding features and the target user portrait embedding coding features to obtain the public health popular science semantic query response coding features that meet the user; S6, generating visual information based on the public health popular science semantic query response coding features that meet the user to obtain the public health popular science visual expression image; S7, pushing the public health popular science visual expression image to the target user object to be pushed for display on the terminal device of the target user object to be pushed.

[0032] Exemplarily, in step S1, a user portrait of the target user object to be pushed is obtained, wherein the user portrait of the target user object to be pushed includes basic attributes, points of interest, and health challenges faced. It should be understood that by creating a detailed user portrait, the specific needs and preferences of each user can be understood more accurately, thereby ensuring that the pushed content not only matches the user's interests, but also adapts to their acceptance ability in terms of cognitive level. Doing so can improve the relevance and pertinence of information push, enhance user experience, and ultimately help improve the effect of public health science popularization and promote the improvement of health literacy for all people. Specifically, understanding the basic attributes of users (such as age, gender, geographic location, etc.) can help determine general health advice suitable for the population; identifying the user's points of interest enables the provision of content that is closer to personal preferences and increases user participation; and clarifying the health challenges faced by users allows guidance to be given for specific health problems or risk factors. These three points combined can provide customized services for users, making public health science popularization more effective.

[0033] In one embodiment, Figure 2 As shown, obtaining a user portrait of a target user object to be pushed includes: S11, collecting online behavior data of the target user object to be pushed from a social media platform, wherein the online behavior data includes search history record data and social media interaction record data; S12, embedding and encoding the online behavior data of the target user object to be pushed using an online behavior embedding matrix to obtain an online behavior embedding coding vector; S13, passing the online behavior embedding coding vector through a user portrait generator based on a large language model to obtain a user portrait of the target user object to be pushed.

[0034] Exemplarily, in step S11, online behavior data of the target user object to be pushed is collected from the social media platform, and the online behavior data includes search history record data and social media interaction record data. It should be understood that the online behavior data of the target user object can provide a basis for in-depth understanding of user interests, needs and dynamic changes. By analyzing the user's online activities, such as their query content in the search engine and their interaction on social media (such as likes, comments, shares, etc.), the topics they are concerned about and their personal preferences can be revealed. This not only helps to capture the user's immediate interests, but also tracks their trends over time, ensuring that the pushed content always remains relevant and targeted.

[0035] In one embodiment, an application programming interface (API) or partnership is used to legally obtain the user's public information on the social media platform. For search history, it can be collected through cooperation with search engine service providers or user authorization; for social media interaction records, it depends on the developer tools and service agreements provided by the social platform.

[0036] Exemplarily, in step S12, the online behavior data of the target user object to be pushed is embedded and encoded using an online behavior embedding matrix to obtain an online behavior embedding coding vector. It should be understood that since the user's online behavior data (such as clicks, likes, search keywords, browsing time, etc.) is usually high-dimensional and may contain a lot of noise. Through embedded coding, these complex behaviors can be converted into low-dimensional dense vector representations. This process not only reduces the dimension of the data and facilitates subsequent data processing, but also helps to capture the implicit patterns and embedded semantic relationships in the online behavior data of the target user object.

[0037] Exemplarily, in step S13, the online behavior embedding coding vector is passed through a user portrait generator based on a large language model to obtain a user portrait of the target user object to be pushed, wherein the user portrait of the target user object to be pushed includes basic attributes, points of interest, and health challenges faced. In particular, since large language models are built based on deep learning technology, they have powerful natural language processing capabilities and can understand complex patterns and semantic information in online behavior embedded coding content. Through these models, users' online behavior data can be better parsed to capture deep information such as user interests, needs, and potential health challenges. Based on this, in the technical solution of the present application, by introducing a user portrait generator based on a large language model, a more comprehensive and detailed user portrait can be generated based on the user's online behavior embedding coding features, and this user portrait helps to formulate a more personalized information push strategy.

[0038] Exemplarily, in step S2, a plurality of preliminary matching public health science popularization contents are extracted from the public health science popularization database. That is to say, through preliminary screening, entries that may be related to the target user profile can be quickly located from a large amount of public health science popularization contents. This method can greatly reduce the amount of data to be processed later and improve the response speed and efficiency of the system.

[0039] In one embodiment, extracting multiple preliminary matching public health science popularization contents from the public health science popularization database includes: first, it is necessary to build a rich public health science popularization database, which contains a large amount of diverse science popularization contents, covering different health topics, cognitive levels and forms of expression (such as text, images, videos, etc.). When receiving a user's request or automatically triggered based on a user profile, algorithms and technical means will be used to efficiently query this huge database, and several items of content that meet the user's characteristics will be screened out as preliminary matching results.

[0040] In a specific embodiment, considering that in a large-scale and complex data set such as a public health science popularization database, it is very time-consuming to directly traverse all records for matching. By creating an index in advance, the content that is most likely to meet user needs can be located in a short time, thereby greatly improving the performance of the system and user experience. Specifically, establishing an index structure involves selecting a suitable index type, determining index fields, using embedded coding and feature vectors, and maintaining an index update mechanism. For example, in a public health science popularization database, you can choose index types such as inverted indexes, tree structures (such as B-trees or Trie trees), or hash tables. Inverted indexes are particularly effective for situations with a lot of text content; for small data sets that need to be updated frequently, hash tables can be considered. When determining index fields, keywords (important words in article titles and abstracts), classification tags (classify content according to health topics), release time (screen the latest information in chronological order), cognitive level (customized content for audiences of different age groups or educational backgrounds), and media types (distinguishing between content forms such as text, images, and videos). In addition, deep learning models can be used to embed and encode each popular science content, generate high-dimensional feature vectors, and store these vectors as part of the index so that these vectors can be directly operated when performing similarity calculations, further speeding up the matching speed.

[0041] In a specific embodiment, suppose that you want to create an index for a popular science article on "Prevention of Hypertension". First, identify the key fields of the article, such as "hypertension" and "prevention" in the title, and key terms in the text, such as "diet control" and "exercise habits". Then, record these keywords and their locations to form an inverted index in the following format: The index table contains each keyword and its associated article ID list. When the user enters the query conditions, the system can quickly find a collection of related articles based on these keywords without having to check the entire database one by one. For example, for the case of multiple keyword combinations or fuzzy matches, it can be supported by expanding the function of the inverted index, introducing a weight scoring mechanism to assign different scores according to the importance of keywords, or using the TF-IDF algorithm to evaluate the universality and uniqueness of keywords in the entire database, so as to optimize the quality of search results.

[0042] Exemplarily, in step S3, the multiple preliminary matching public health science popularization contents are embedded and encoded to obtain multiple public health science popularization content embedding coding features. It should be understood that the user portrait embedding matrix is ​​used to embed the user portrait of the target user object to be pushed to obtain the target user portrait embedding coding vector. It should be understood that since user portraits usually contain various types of unstructured or semi-structured data (such as text descriptions, category labels, etc.). Through embedded coding, these data information can be converted into a numerical vector form, so that the machine learning model can process and analyze these data. In addition, embedded coding can also map these data into a common embedding space to help capture complex relationships and patterns in user portraits. For example, the association between different points of interest, the connection between basic attributes and health challenges, etc. This encoding method can effectively retain these complex semantic information, rather than just simple keyword matching.

[0043] In one embodiment, the multiple preliminary matching public health science popularization contents are embedded and encoded to obtain multiple public health science popularization content embedding coding features, including: using the popular science content embedding matrix to embed and encode the multiple preliminary matching public health science popularization contents respectively to obtain multiple public health science popularization content embedding coding vectors as the multiple public health science popularization content embedding coding features.

[0044] Exemplarily, in step S4, the user portrait of the target user object to be pushed is embedded coded to obtain the target user portrait embedded coding feature. In one embodiment, the user portrait of the target user object to be pushed is embedded coded to obtain the target user portrait embedded coding feature, including: using the user portrait embedding matrix to embed the user portrait of the target user object to be pushed to obtain the target user portrait embedded coding vector as the target user portrait embedded coding feature. It should be understood that by means of embedded coding, each preliminary matching public health popular science content can be mapped to a common space, and the embedded coding feature information in each preliminary matching public health popular science content can be captured. Compared with simple keyword matching, the high-dimensional embedded semantic matching method based on embedded coding can more accurately identify content that truly meets the user's interests and cognitive level, thereby improving the quality of recommendation.

[0045] Exemplarily, in step S5, the multiple public health science popularization content embedded coding features and the target user portrait embedded coding features are subjected to feature dynamic semantic search query processing to obtain coding features that meet the user's public health science popularization semantic query response. It should be understood that since the multiple public health science popularization content embedded coding vectors and the target user portrait embedded coding vectors respectively contain embedded coding features related to each public health science popularization content and embedded coding features of the target user portrait. And each user's health needs, points of interest and cognitive level are unique. Therefore, in order to provide public health science popularization content information related to different individual users according to their needs and cognition, it is necessary to be able to understand and match these personal characteristics to the most suitable popular science content. Based on this, in the technical solution of the present application, the multiple public health science popularization content embedded coding features and the target user portrait embedded coding features are further subjected to feature dynamic semantic search query processing to obtain coding features that meet the user's public health science popularization semantic query response. In particular, the process of feature dynamic semantic search query processing can handle efficient query matching and semantic understanding tasks in a large-scale public health science popularization content embedding coding vector library. It can automatically identify the public health science popularization content embedding coding features (i.e., dynamic search anchor points) that best represent the user's profile query intent in a complex high-dimensional data space by introducing a dynamic selection mechanism and semantic coding technology, and define an optimized selection range based on the dynamic search anchor points of the public health science popularization content embedding coding features. Within this range, each public health science popularization content embedding coding vector is further screened and evaluated, and finally semantic coding is performed through a deep learning model to generate a semantic embedding representation of the public health science popularization semantic query response that can accurately reflect the user's profile and meet the query intent.

[0046] Specifically, compared with the traditional query response encoding network, the feature dynamic semantic search query processing can focus on the public health science content embedding encoding features that are most relevant to the user portrait by defining the dynamic selection range ratio, thereby effectively reducing the interference of irrelevant information and achieving preliminary noise reduction and dimensionality compression of the data. At the same time, with the powerful expression ability of deep neural networks, the complex nonlinear relationship between the input multiple public health science content embedding encoding vectors and the target user portrait embedding encoding vector can be learned. In addition, the "anchoring based on the selection range ratio" method allows the system to dynamically adjust the search strategy according to the current query, ensuring that the most relevant public health science content that matches the user can always be found, while maintaining good adaptability to new data, providing a basis for the subsequent push of public health science visual information.

[0047] In one embodiment, Figure 3 As shown, the multiple public health science popularization content embedded coding features and the target user portrait embedded coding features are subjected to feature dynamic semantic search query processing to obtain coding features that meet the user's public health science popularization semantic query response, including: S51, determining a selection range ratio based on the multiple public health science popularization content embedded coding features and the target user portrait embedded coding features; S52, based on the selection range ratio, performing dynamic semantic search encoding on the multiple public health science popularization content embedded coding features and the target user portrait embedded coding features to obtain coding features that meet the user's public health science popularization semantic query response.

[0048] In one embodiment, Figure 4 As shown, based on the multiple public health science popularization content embedding coding features and the target user portrait embedding coding features, determining the selection range ratio includes: S511, calculating the internal relationship score value of the target user portrait embedding coding vector relative to each public health science popularization content embedding coding vector in the multiple public health science popularization content embedding coding vectors to obtain a sequence of target user portrait-public health science popularization content semantic internal relationship score values. Specifically, the process can be expressed by the formula:

[0049]

[0050]

[0051] in, are the multiple public health science popularization content embedding coding vectors, are respectively the first, second, and third public health science popularization content embedded coding vectors. and Public health popular science content embedding coding vector, is the target user portrait embedding encoding vector, and They are the public health science content weight matrix and the target user portrait weight matrix, for function, is the transposed vector of the modulation vector, The first in the sequence of the semantic internal relationship score values ​​of the target user portrait-public health popular science content Target user portrait-semantic internal relationship score of public health popular science content.

[0052] S512, using the public health science popularization content embedding coding vector corresponding to the maximum value in the sequence of the target user portrait-public health science popularization content semantic internal relationship score value as a dynamic search anchor vector to obtain the target user portrait-public health science popularization content semantic dynamic search anchor vector. Specifically, the process can be expressed by the formula:

[0053]

[0054] in, To return the public health science popularization content embedding encoding vector corresponding to the maximum value in the sequence of the target user portrait-public health science popularization content semantic internal relationship score values, For target user portraits - semantic dynamic search anchor vectors for public health popular science content.

[0055] S513, based on the characteristic distribution characteristics of the target user portrait-public health science popularization content semantic dynamic search anchor vector, determine the selection range ratio, wherein the vector of the starting position of the selection range ratio is the target user portrait-public health science popularization content semantic dynamic search anchor vector. Specifically, the process can be expressed by the formula:

[0056]

[0057] in, is the average value of the target user portrait-public health science content semantic dynamic search anchor quantity, The variance of the anchoring quantity for the target user portrait-public health popular science content semantic dynamic search, is a trainable hyperparameter, Indicates rounding up operation. To select the range ratio.

[0058] Specifically, the correlation between the user portrait and each popular science content is quantified by calculating the internal relationship score of the target user portrait embedding coding vector relative to the embedding coding vectors of multiple public health popular science contents to form a sequence. Subsequently, the popular science content embedding coding vector corresponding to the maximum score is selected as the dynamic search anchor vector. This step ensures that the initial reference point that best represents the user's interests and needs can be found. Next, the selection range ratio is determined based on the characteristic distribution characteristics of this dynamic search anchor vector. The selection range ratio here defines an optimized search space, so that subsequent screening and evaluation work can be carried out in a more focused and relevant area. The starting position of the selection range ratio is occupied by the above-mentioned dynamic search anchor vector, which means that the entire search process will revolve around the content that is most likely to arouse the user's interest, thereby improving the search efficiency and accuracy. The effect of this mechanism is that it can significantly improve the quality of personalized push. First, by calculating the internal relationship score value and selecting the best match as the anchor point, the user's current focus of interest can be accurately captured, avoiding the interference of irrelevant or low-relevance content. Secondly, the search strategy is adjusted based on dynamic search anchoring, allowing for flexible positioning of the content that best suits the user based on real-time conditions while maintaining good adaptability to new data. Finally, by determining the selection range ratio, the search scope is further narrowed, focusing on the information that is most likely to be accepted and understood by users, thereby achieving preliminary noise reduction and dimensionality compression of the data, and enhancing the relevance and pertinence of the final push content.

[0059] In one embodiment, Figure 5 As shown, based on the selection range ratio, the multiple public health science popularization content embedding coding features and the target user portrait embedding coding features are dynamically semantically searched and encoded to obtain the coding features that meet the user's public health science popularization semantic query response, including: S521, based on the selection range ratio, determining multiple dynamic search optimization public health science popularization content embedding coding vectors, wherein each feature vector in the selection range ratio is defined as a dynamic search optimization public health science popularization content embedding coding vector. Specifically, the process can be expressed by the formula:

[0060]

[0061] in, is the vector of the starting position of the selected range ratio, that is, , is the vector of the end position of the selected range ratio, embedding coding vectors for each of the multiple dynamic search optimization public health science popularization content embedding coding vectors, Optimize the public health science popularization content embedding coding vectors for the multiple dynamic searches.

[0062] S522, input the target user portrait-public health science popularization content semantic dynamic search anchor vector and the multiple dynamic search optimized public health science popularization content embedding coding vectors into a dynamic semantic search encoder to obtain a coding vector that meets the user's public health science popularization semantic query response as the coding feature that meets the user's public health science popularization semantic query response. Specifically, the process can be expressed by the formula:

[0063]

[0064]

[0065] in, is the one-norm of the vector, Cosine similarity for dynamic semantic search, The encoding vector is a response to the user's public health science semantic query.

[0066] Specifically, after a preliminary matching content has been found as a dynamic search anchor point based on the user profile, the next task is to conduct more detailed screening within a defined selection range around this anchor point. The selection range here is actually a calculated set of parameters that helps identify those popular science contents that are most likely to resonate with users and are easy to understand. Within this range, each feature vector is considered as a potential optimization candidate, that is, the dynamic search optimization public health popular science content embedding encoding vector. This means that the relationship between these candidates and the user profile will be carefully evaluated to find the best combination. Then, all selected dynamic search optimization public health popular science content embedding encoding vectors are fed into the dynamic semantic search encoder together with the initially selected dynamic search anchor vector. The role of the encoder is to deeply process these vectors, capture the complex semantic associations between them, and generate a final encoding vector that meets the user's public health popular science semantic query response. This process is not just a simple data aggregation, but an in-depth analysis of the interactions between the various vectors through a machine learning model to extract the semantic feature representation that best reflects user needs. In other words, the encoder tries to understand which combination of popular science content can best meet the user's specific query intent while ensuring the quality and relevance of the content.

[0067] Exemplarily, in step S6, visual information is generated based on the coding features of the public health science popularization semantic query response that meets the user to obtain a public health science popularization visual expression image. In one embodiment, visual information is generated based on the coding features of the public health science popularization semantic query response that meets the user to obtain a public health science popularization visual expression image, including: passing the coding vector of the public health science popularization semantic query response that meets the user through a visual information generator based on AIGC to obtain the public health science popularization visual expression image. It should be understood that in the field of public health science popularization, although text information is important, it is often difficult to attract the attention of all users, especially for those who are more inclined to visual learning. Visual expression images can intuitively display complex scientific concepts or health suggestions in the form of charts, illustrations, animations, etc., making the information more vivid, easy to understand and remember. In addition, personalized visual content can be customized according to the user's interest points and cognitive level, so as to better meet the needs of different groups of people and improve the effectiveness of information transmission. Therefore, using AIGC technology to generate a visual expression image that is highly matched with the user portrait can significantly enhance the dissemination effect of popular science information. In a specific embodiment, the coding vector of the public health science popularization semantic query response that meets the user is input into a visual information generator based on AIGC. This generator is usually a deep learning model that has been trained to understand and convert text or numerical data into visual elements. The generator interprets the information in the encoded vector and generates corresponding images, charts, or other forms of visualization based on its instructions.

[0068] Preferably, considering that the target user portrait embedding coding vector and the multiple public health science popularization content embedding coding vectors respectively represent the embedded coding features of the target user portrait and the semantic coding features of each of the multiple preliminary matching public health science popularization contents, when performing feature dynamic semantic search encoding based on selection range ratio anchoring, the complexity of the selection range ratio anchoring mechanism caused by different encoding modes under modal differences will lead to insufficient long-distance dynamic semantic search coding representation of the public health science popularization semantic query response coding vector that meets the user, thereby reducing the expression effect of the public health science popularization semantic query response coding vector that meets the user, and affecting the image quality of the public health science popularization visual expression image obtained by inputting the AIGC-based visual information generator.

[0069] Therefore, in one example, when the encoding vector of the response to the public health science semantic query that meets the user is input into the visual information generator based on AIGC, the encoding vector of the response to the public health science semantic query that meets the user is optimized, and the optimization includes the following steps:

[0070] The encoded vector that meets the user's public health science semantic query response is reconstructed in feature topological continuity according to the size of the feature value at each position to obtain a topological sequential generation vector that meets the user's public health science semantic query response;

[0071] If the user's public health science semantic query response topological sequential generation vector The eigenvalue and The pedigree distance between the eigenvalues ​​is less than or equal to the cross-order neighborhood judgment threshold, and the first The eigenvalue and The bilinear coupling between the eigenvalues ​​is the first The eigenvalue is expressed as:

[0072]

[0073]

[0074] in, Indicates if, and represents the topological sequential generation vector of the user's public health science semantic query response. Eigenvalues ​​and Eigenvalues, represents the topological sequential generation vector of the user's public health science semantic query response. The eigenvalue and The spectral distance between eigenvalues, and denote the first predetermined weight and the second predetermined weight respectively, represents the cross-order neighborhood judgment threshold, Represents the optimized Eigenvalue.

[0075] If the user's public health science semantic query response topological sequential generation vector The eigenvalue and The spectral distance between the eigenvalues ​​is greater than the cross-order neighborhood judgment threshold, and the global eigenvalue spectral norm hyperbolic projection is performed on the topological sequential generation vector that meets the user's public health science popularization semantic query response to obtain a global spectral norm projection factor, and a multimodal generalization space implicit kernel value is constructed based on the global spectral norm projection factor and the vector length of the topological sequential generation vector that meets the user's public health science popularization semantic query response, and then the first The eigenvalues ​​are fine-tuned to obtain the optimized Eigenvalue;

[0076]

[0077]

[0078]

[0079]

[0080] in, It represents the modulus value of the topological sequentially generated vector that meets the user's public health science semantic query response. represents the global spectral norm projection factor, represents the length of the vector that meets the topological sequential generation of the user's public health science semantic query response, represents the third predetermined weight, represents the fourth predetermined weight, represents the fifth predetermined weight, Here, the selection of the predetermined weight and the cross-order neighborhood determination threshold can be based on empirical estimation and can be tuned by model training. For example, the first predetermined weight = 0.7, the second predetermined weight = 0.3, the third predetermined weight = 0.6, the fourth predetermined weight = 0.4, the fifth predetermined weight = 0.5, the sixth predetermined weight = 0.5, and the cross-order neighborhood determination threshold = 0.85 can be preset. Of course, this is only an exemplary and non-restrictive example, not a specific limitation.

[0081] based on , the first Eigenvalue To obtain an optimized encoding vector that meets the user's public health science semantic query response.

[0082] Thus, in this preferred embodiment, the multimodal generalized space implicit architecture based on heterogeneous tensor aggregation of the user's public health science semantic query response coding vector is used to capture its spectral domain holographic network coupling complex structure, so as to reconstruct the functional association topological mapping of the user's public health science semantic query response coding vector through the reverse projection multimodal generalized space implicit architecture, so as to realize the coding reconstruction of the cross-scale topological sequence generation mechanism of the user's public health science semantic query response coding vector, improve the coding expression effect of the user's public health science semantic query response coding vector, and improve the image quality of the public health science visual expression image obtained by inputting the AIGC-based visual information generator. In this way, it is possible to customize the push of visual public health science information suitable for different groups of people, thereby ensuring that the pushed content not only matches the user's interests, but also adapts to their acceptance ability in terms of cognitive level, not only improving the relevance and pertinence of information push, but also enhancing the user experience, which is helpful to improve the effect of public health science popularization and promote the improvement of health literacy of all people.

[0083] Exemplarily, in step S7, the public health science popularization visual expression image is pushed to the target user object to be pushed for display on the terminal device of the target user object to be pushed. It should be understood that through the push mechanism, modern communication technologies and network platforms, such as mobile applications, social media, emails or websites, can be used to accurately send the generated visual expression images to the user's smartphone, tablet computer, personal computer and other terminal devices. This method ensures that no matter where the user is, as long as they are connected to the Internet, they can instantly receive the latest and highly relevant health knowledge. More importantly, these visual contents are designed to be easy to understand and share, encouraging users to actively participate in health education, and can be spread within social circles to further expand their influence.

[0084] In one embodiment, in the specific implementation of push, it is first necessary to confirm the user's device type and its preferences, such as notification permissions, receiving time windows, etc. Then, the most appropriate way to distribute content is selected based on this information. For example, for young people who actively use social media platforms, links or embedded pictures can be sent through social messages or the platform's built-in notification function; for users who prefer traditional channels, email or SMS reminders may be used. In addition, in order to ensure the success rate of push and user experience, network conditions will be monitored, data transmission protocols will be optimized, and delays and freezes will be reduced.

[0085] In summary, the visual information push method for public health science popularization according to the embodiment of the present application is explained, which can create a detailed user portrait by analyzing the user's online behavior data, including social media interaction records and search history, and then understand the specific needs and preferences of each user. Then, this user portrait is used to perform semantic search queries with the various popular science contents in the public health science popularization database to match the public health science popularization semantic query response feature representation that meets the user's interests and needs, and based on this, a public health science popularization visual expression image is generated to push it to the target user object. This personalized public health science popularization visual information push method can ensure that the pushed content not only matches the user's interests, but also adapts to their receptive ability in terms of cognitive level, which not only improves the relevance and pertinence of information push, but also enhances the user experience, which helps to improve the effect of public health science popularization and promote the improvement of health literacy for all people.

[0086] Figure 6 Schematic block diagram of a visual information push system for public health science popularization according to an embodiment of the present application. Figure 6As shown, the visual information push system 100 for public health science popularization includes: a user portrait acquisition module 110, which is used to obtain the user portrait of the target user object to be pushed, wherein the user portrait of the target user object to be pushed includes basic attributes, points of interest and health challenges faced; a preliminary matching public health science popularization content extraction module 120, which is used to extract multiple preliminary matching public health science popularization contents from the public health science popularization database; a public health science popularization content embedding coding module 130, which is used to embed the multiple preliminary matching public health science popularization contents to obtain multiple public health science popularization content embedding coding features; a target user portrait embedding coding module 140, which is used to embed the user portrait of the target user object to be pushed The image is embedded and encoded to obtain the embedded coding features of the target user portrait; the feature dynamic semantic search query processing module 150 is used to perform feature dynamic semantic search query processing on the multiple public health science popularization content embedded coding features and the target user portrait embedded coding features to obtain the coding features that meet the user's public health science popularization semantic query response; the visual information generation module 160 is used to generate visual information based on the coding features that meet the user's public health science popularization semantic query response to obtain a public health science popularization visual expression image; the visual expression image pushing module 170 is used to push the public health science popularization visual expression image to the target user object to be pushed so as to display it on the terminal device of the target user object to be pushed.

[0087] In one embodiment, a feature dynamic semantic search query processing module is used to perform feature dynamic semantic search query processing on the multiple public health science popularization content embedded coding features and the target user portrait embedded coding features to obtain coding features that meet the user's public health science popularization semantic query response, including: a selection range ratio determination unit, used to determine the selection range ratio based on the multiple public health science popularization content embedded coding features and the target user portrait embedded coding features; a dynamic semantic search coding unit, used to perform dynamic semantic search coding on the multiple public health science popularization content embedded coding features and the target user portrait embedded coding features based on the selection range ratio to obtain the coding features that meet the user's public health science popularization semantic query response.

[0088] In one embodiment, the user portrait acquisition module is used to: collect online behavior data of the target user object to be pushed from the social media platform, the online behavior data including search history record data and social media interaction record data; use the online behavior embedding matrix to embed the online behavior data of the target user object to be pushed to obtain an online behavior embedding coding vector; and pass the online behavior embedding coding vector through a user portrait generator based on a large language model to obtain a user portrait of the target user object to be pushed.

[0089] In one embodiment, the public health science popularization content embedding coding module is used to: use the popular science content embedding matrix to respectively embed code the multiple preliminary matching public health science popularization contents to obtain multiple public health science popularization content embedding coding vectors as the multiple public health science popularization content embedding coding features.

[0090] Here, those skilled in the art can understand that the specific operations of the various modules and units in the above-mentioned visual information push system for public health science popularization have been referred to above. Figures 1 to 5 The description of the visual information push method for public health popularization has been introduced in detail, and therefore, its repeated description will be omitted.

[0091] An embodiment of the present application further provides a computer program product, which includes a computer program code. When the computer program code runs on a computer, the computer implements the methods in the above embodiments of the present application.

[0092] An embodiment of the present application further provides a computer-readable storage medium, which stores computer instructions. When the computer instructions are executed on a computer, the computer implements the methods in the above embodiments of the present application.

[0093] An embodiment of the present application also provides a chip, including a circuit, for executing the methods in the above embodiments of the present application.

[0094] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0095] In the description of the embodiments of the present application, unless otherwise specified, “ / ” means or, for example, A / B can mean A or B; “and / or” in this article is a description of the association relationship of associated objects, indicating that three relationships can exist, for example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In this application, “at least one” means one or more, and “more than one” means two or more. “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, ,or , where a, b, c can be single or multiple.

[0096] The prefixes such as "first" and "second" used in the embodiments of the present application are only used to distinguish different description objects, and have no limiting effect on the position, order, priority, quantity or content of the described objects. The use of prefixes such as ordinal numbers used to distinguish description objects in the embodiments of the present application does not constitute a limitation on the described objects. For the statement of the described objects, please refer to the description in the context of the claims or embodiments, and the use of such prefixes should not constitute an unnecessary limitation.

[0097] In the several embodiments provided in the present application, it should be understood that the disclosed systems, 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 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.

[0098] In the various embodiments of the present application, unless otherwise specified or logically conflicting, the terms and / or descriptions between the various embodiments are consistent and can be referenced to each other, and the technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationships.

[0099] The units described 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.

[0100] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0101] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for pushing visual information for public health popularization, characterized in that: include: Obtaining a user profile of the target user object to be pushed, wherein the user profile of the target user object to be pushed includes basic attributes, points of interest, and health challenges faced, and the basic attributes include the age, gender, and geographic location of the target user object to be pushed; Extracting multiple preliminary matching public health popular science contents from the public health popular science database; Embedding and encoding the plurality of preliminary matched public health science popularization contents to obtain a plurality of public health science popularization contents embedding and encoding features; Embedding the user portrait of the target user object to be pushed to obtain an embedded coding feature of the target user portrait; The multiple public health science popularization content embedding coding features and the target user portrait embedding coding features are subjected to feature dynamic semantic search query processing to obtain coding features that meet the user's public health science popularization semantic query response, including: determining a selection range ratio based on the multiple public health science popularization content embedding coding features and the target user portrait embedding coding features; based on the selection range ratio, the multiple public health science popularization content embedding coding features and the target user portrait embedding coding features are subjected to dynamic semantic search coding to obtain coding features that meet the user's public health science popularization semantic query response; Generate visual information based on the coding features of the response to the public health science semantic query that meets the user to obtain a public health science visual expression image; The public health science popularization visual expression image is pushed to the target user object to be pushed so as to be displayed on the terminal device of the target user object to be pushed.

2. The method for pushing visual information for public health science popularization according to claim 1, characterized in that: Get the user profile of the target user object to be pushed, including: Collect online behavior data of the target user object to be pushed from the social media platform, wherein the online behavior data includes search history record data and social media interaction record data; Using an online behavior embedding matrix, embedding and coding the online behavior data of the target user object to be pushed to obtain an online behavior embedding coding vector; The online behavior embedding coding vector is passed through a user portrait generator based on a large language model to obtain a user portrait of the target user object to be pushed.

3. The method for pushing visual information for public health science popularization according to claim 2, characterized in that: The multiple preliminary matching public health science popularization contents are embedded and encoded to obtain multiple public health science popularization content embedding coding features, including: using the popular science content embedding matrix to respectively embed and encode the multiple preliminary matching public health science popularization contents to obtain multiple public health science popularization content embedding coding vectors as the multiple public health science popularization content embedding coding features.

4. The method for pushing visual information for public health popularization according to claim 3, characterized in that: The user portrait of the target user object to be pushed is embedded and encoded to obtain a target user portrait embedded coding feature, including: using a user portrait embedding matrix to embed the user portrait of the target user object to be pushed to obtain a target user portrait embedded coding vector as the target user portrait embedded coding feature.

5. The method for pushing visual information for public health science popularization according to claim 4, characterized in that: Determining a selection range ratio based on the multiple public health science popularization content embedding coding features and the target user portrait embedding coding features includes: Calculate the internal relationship score value of the target user portrait embedding coding vector relative to each public health science popularization content embedding coding vector in the multiple public health science popularization content embedding coding vectors to obtain a sequence of target user portrait-public health science popularization content semantic internal relationship score values; Using the public health science popularization content embedding coding vector corresponding to the maximum value in the sequence of the target user portrait-public health science popularization content semantic internal relationship score value as the dynamic search anchor vector to obtain the target user portrait-public health science popularization content semantic dynamic search anchor vector; Based on the characteristic distribution characteristics of the target user portrait-public health popular science content semantic dynamic search anchor vector, the selection range ratio is determined, wherein the vector of the starting position of the selection range ratio is the target user portrait-public health popular science content semantic dynamic search anchor vector.

6. The method for pushing visual information for public health science popularization according to claim 5, characterized in that: Based on the selection range ratio, the plurality of public health science popularization content embedding coding features and the target user portrait embedding coding features are dynamically semantically searched and coded to obtain the coding features that meet the user's public health science popularization semantic query response, including: Based on the selection range ratio, determining a plurality of dynamic search optimization public health science popularization content embedding coding vectors, wherein each feature vector in the selection range ratio is defined as a dynamic search optimization public health science popularization content embedding coding vector; The target user portrait-public health science popularization content semantic dynamic search anchor vector and the multiple dynamic search optimized public health science popularization content embedding coding vectors are input into the dynamic semantic search encoder to obtain a coding vector that meets the user's public health science popularization semantic query response as the coding feature that meets the user's public health science popularization semantic query response.

7. The method for pushing visual information for public health science popularization according to claim 6, characterized in that: Based on the coding features of the response to the public health science popularization semantic query that meets the user, visual information is generated to obtain a public health science popularization visual expression image, including: passing the coding vector that meets the public health science popularization semantic query response of the user through a visual information generator based on AIGC to obtain the public health science popularization visual expression image.

8. A visual information push system for public health popularization, characterized in that: include: A user portrait acquisition module, used to acquire a user portrait of a target user object to be pushed, wherein the user portrait of the target user object to be pushed includes basic attributes, points of interest, and health challenges faced, and the basic attributes include the age, gender, and geographic location of the target user object to be pushed; A preliminary matching public health science popularization content extraction module is used to extract a plurality of preliminary matching public health science popularization contents from a public health science popularization database; A public health science popularization content embedding coding module, used for embedding and coding the plurality of preliminary matching public health science popularization contents to obtain a plurality of public health science popularization content embedding coding features; A target user portrait embedding coding module is used to embed the user portrait of the target user object to be pushed to obtain a target user portrait embedding coding feature; A feature dynamic semantic search query processing module is used to perform feature dynamic semantic search query processing on the multiple public health science popularization content embedding coding features and the target user portrait embedding coding features to obtain coding features that meet the user's public health science popularization semantic query response, including: a selection range ratio determination unit, used to determine the selection range ratio based on the multiple public health science popularization content embedding coding features and the target user portrait embedding coding features; a dynamic semantic search coding unit, used to perform dynamic semantic search coding on the multiple public health science popularization content embedding coding features and the target user portrait embedding coding features based on the selection range ratio to obtain the coding features that meet the user's public health science popularization semantic query response; A visual information generation module, used to generate visual information based on the coding features of the response to the user's public health science semantic query to obtain a public health science visual expression image; The visual expression image pushing module is used to push the public health science popularization visual expression image to the target user object to be pushed so as to display it on the terminal device of the target user object to be pushed.

9. The visual information push system for public health science popularization according to claim 8, characterized in that: The user portrait acquisition module is used to collect online behavior data of the target user object to be pushed from the social media platform, wherein the online behavior data includes search history record data and social media interaction record data; The online behavior data of the target user object to be pushed is embedded and encoded using an online behavior embedding matrix to obtain an online behavior embedding coding vector; and the online behavior embedding coding vector is passed through a user portrait generator based on a large language model to obtain a user portrait of the target user object to be pushed.

10. The visual information push system for public health science popularization according to claim 9, characterized in that: The public health science popularization content embedding coding module is used to: use the science popularization content embedding matrix to embed the multiple preliminary matching public health science popularization contents respectively to obtain multiple public health science popularization content embedding coding vectors as the multiple public health science popularization content embedding coding features.

Images