A multi-source data fusion analysis push system and method based on cloud platform

Abstract

The present invention discloses a multi-source data fusion analysis push system and method based on a cloud platform, which belongs to the field of artificial intelligence technology. The present invention obtains artwork data through several data sources, provides virtual exhibitions and interactive displays of artworks through a cloud platform; uses a natural language processing algorithm to analyze artworks and integrates to form artwork features; collects access data of different users on the cloud platform, generates the popularity and trends of different users' visits in real time, and performs visual displays for administrators; analyzes user access data, performs data preprocessing and feature extraction, and establishes user portraits; uses a collaborative filtering algorithm to train a personalized recommendation model based on user portraits and artwork features to achieve personalized recommendations; establishes a multi-language support function on the cloud platform based on artwork data from different regions and cultural backgrounds, and designs exhibitions and interpretation services in combination with the characteristics of artworks from different regions and cultures.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a multi-source data fusion analysis and push system and method based on a cloud platform. Background Art

[0002] Traditional art display platforms often only provide simple picture displays and text introductions, and users cannot get a rich art experience. In addition, when faced with a large amount of art, users often find it difficult to find works that suit their interests. Existing technologies lack interactivity and personalized art recommendation functions, and cannot meet users' diverse needs for art appreciation.

[0003] In the context of globalization, users have shown an increasingly broad interest in artworks from different cultural backgrounds. However, existing platforms are often limited to specific regions or languages ​​and cannot effectively promote cross-cultural artistic exchanges and cooperation. On existing platforms, it is not possible to quickly and intuitively view the full picture of high-quality steles, calligraphy or paintings. In terms of operation, it is necessary to constantly switch pictures, and it is impossible to quickly locate the content in calligraphy works, and then it is impossible to provide a search function in the works. On existing platforms, if the resolution of the sliced ​​images of artworks is too small, the number of files will be too large; if the resolution of the sliced ​​images is too large, the clarity cannot be guaranteed. Summary of the invention

[0004] The purpose of the present invention is to provide a multi-source data fusion analysis push system and method based on a cloud platform to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] A multi-source data fusion analysis and push method based on a cloud platform, the method comprising the following steps:

[0007] S100, obtaining artwork data from a number of data sources, wherein the artwork data includes artwork display data, artist background information, artwork collection value, and artwork reviews, and integrating the data into a cloud platform, and providing virtual exhibitions and interactive displays of artworks through the cloud platform; using a natural language processing algorithm to analyze the historical background, cultural connotation, style characteristics, and artistic expression techniques of the artwork, and integrating them to form artwork features;

[0008] S200, collecting access data of different users on the cloud platform, wherein the access data includes the number of page views, dwell time, click hot spots, favorite preferences and user interaction behaviors of the exhibition page, generating the popularity and trends of different users' visits in real time, and performing visual display for the administrator;

[0009] S300, analyzing the user's access data, performing data preprocessing and feature extraction, and establishing a user profile; using a collaborative filtering algorithm, training a personalized recommendation model based on the user profile and artwork features. The goal of the model training is to predict the user's preference for artworks that have not been viewed based on the user's interests and needs, so as to achieve personalized recommendations;

[0010] S400. Based on the artwork data from different regions and cultural backgrounds, establish multi-language support functions on the cloud platform, design exhibition and interpretation services in combination with the characteristics of artworks from different regions and cultures, and provide cultural interpretation and exchange forum functions.

[0011] According to step S100, the exhibition space of the virtual exhibition is designed on the cloud platform through three-dimensional modeling, and the layout is designed according to different exhibition themes and characteristics of artworks. The exhibition space can simulate the exhibition hall of a traditional art gallery or be designed into a scene that matches the style of the artwork;

[0012] The design of the exhibition space includes: the size, shape, walls, floors, lighting, art display racks and cabinets, navigation and paths, interactive elements, sound effects and user behavior tracking of the exhibition hall; users can interact with the artwork by clicking, zooming, rotating or dragging, and set up an independent display area for each artwork in the exhibition space, providing an introduction to the artwork's historical background, cultural connotation, style characteristics and artistic expression techniques; users can freely browse the virtual exhibition space and choose exhibition areas of interest to explore and view.

[0013] According to step S100, the historical background, cultural connotation, style characteristics and artistic expression of the artwork are analyzed using a natural language processing algorithm, and text cleaning, word segmentation and part-of-speech tagging, stop word filtering, morphological restoration and stem extraction are performed through preprocessing;

[0014] In view of historical background and cultural connotation, we use the pre-trained NER model to identify and extract entities of time, place, and person in the text, use the LDA model to analyze the text, extract the main themes and keywords, analyze the sentiment tendency in the text, and use the VADER sentiment analysis model to obtain the cultural connotation of artworks and artists and the emotional color of historical background;

[0015] The NER model is a model used to identify named entities in text. The pre-trained NER model is based on deep learning technology and is trained through large-scale labeled text data. It can automatically identify and extract entities in the text, such as names of people, places, and organizations. In the present invention, a recurrent neural network is used, which can accurately identify entities in the text after training and can perform semantic understanding based on the context.

[0016] LDA is a generative probabilistic model for document topic modeling. It treats each document as a mixture of multiple topics, each of which consists of multiple words. The LDA model automatically discovers topics from a document collection by analyzing the distribution of words in the document and assigns a set of keywords to each topic.

[0017] VADER is a pre-trained model for sentiment analysis. It is based on a rule-based and lexicon approach and can classify sentiment in text, including positive, negative, and neutral polarity. The VADER model uses a lexicon containing sentiment words and their sentiment intensity, and takes into account factors such as negation, degree adverbs, and sentiment expressions, so that it can accurately determine the sentiment tendency of the text.

[0018] Based on the style characteristics and artistic expressions, TF-IDF is used to extract keywords describing the artistic style and expressions, and the syntax analysis tool SpaCy is used to parse the syntactic structure of the text to obtain the relationships in the description of the artwork;

[0019] The extracted entities, topics, emotions, keywords and syntactic structures are converted into feature vectors, and the word embedding model Word2Vec is used to convert text features into feature vectors; the feature vectors are concatenated and weighted averaged to form artwork features.

[0020] TF-IDF is a common technique for text feature extraction, which measures the importance of a word in a document. SpaCy is an open source natural language processing library that provides a wealth of text processing functions, including word segmentation, part-of-speech tagging, named entity recognition, syntactic analysis, etc. Word2Vec is a word embedding technology used to convert words in a text into dense vector representations. It is based on a neural network model and generates word vectors by learning the contextual information of words. The Word2Vec model can map words with similar contexts into similar vector spaces, thereby achieving semantic similarity.

[0021] According to step S200, the page views record the number of page views of each exhibition page, the dwell time records the time the user stays on each page, the click hotspot records the user's click behavior on the page to determine the hotspot area, the collection preference records the artworks and exhibitions collected by the user, and the interactive behavior records the user's interactive behavior in the exhibition;

[0022] When collecting data, JavaScript code is embedded in the front-end page to capture users' browsing and interaction behaviors, record user request data through server logs, and use Google Analytics to collect user behavior data.

[0023] Step S300 includes:

[0024] S301, collect user access data, analyze the user access data, perform data preprocessing and feature extraction, generate user access data feature vectors, and establish user portraits;

[0025] S302, constructing a user-artwork rating matrix, wherein each element in the matrix corresponds to a user's rating of the artwork;

[0026] S303, using cosine similarity to calculate the similarity between two users to form a user similarity matrix; based on the user similarity matrix, predicting the user's rating for the artwork that has not been visited;

[0027] S304, using cosine similarity to calculate the similarity between two artworks to form an artwork similarity matrix; based on the artwork similarity matrix, predicting the user's rating for the artwork that has not been visited;

[0028] S305, decomposing the user-artwork rating matrix into two low-rank matrices by matrix decomposition, wherein the two low-rank matrices include a user feature matrix and an artwork feature matrix;

[0029] S306, solving the user feature matrix and the artwork feature matrix by minimizing the loss function;

[0030] S307, using the gradient descent method to update the user feature matrix and the artwork feature matrix;

[0031] S308, using the trained user feature matrix and artwork feature matrix to predict the user's rating of the artwork;

[0032] S309. Use the cross-validation method to evaluate the performance of the model. The evaluation indicator used is the root mean square error. According to the predicted score, recommend the highest-rated unviewed artwork to each user.

[0033] According to step S302, a user-artwork rating matrix R is constructed, in which each element represents the rating of user u on artwork i, which is weighted based on the feature vectors of browsing volume, dwell time, click hotspots, collection preferences, and user interaction behavior in the user profile. The user-artwork rating matrix R is expressed as follows:

[0034]

[0035] Where n represents the number of users and m represents the number of artworks.

[0036] According to step S303, the similarity between users u and v is calculated using cosine similarity, and the formula is as follows:

[0037]

[0038] in, Represents the similarity between users u and v;

[0039] Based on the user similarity matrix, predict the score of user u for artwork i that has not been visited. The formula is as follows:

[0040]

[0041] in, is the predicted rating of user u on artwork i, is the average rating of user u, is a set of neighbor users similar to user u;

[0042] According to step S304, the similarity between artworks i and j is calculated using cosine similarity, and the formula is as follows:

[0043]

[0044] in, is the similarity between artworks i and j;

[0045] Based on the artwork similarity matrix, predict the score of user u for artwork i that has not been visited. The formula is as follows:

[0046]

[0047] in, is the average rating of artwork i, is the set of neighbor artworks similar to artwork i.

[0048] According to step S305, the user-artwork rating matrix R is decomposed into two low-rank matrices P and Q by matrix decomposition, where P is the user feature matrix and Q is the artwork feature matrix, and the formula is as follows:

[0049]

[0050] in, , , k is the dimension of the eigenvector, and T represents the transpose of the matrix;

[0051] According to step S306, P and Q are solved by minimizing the following loss function, and the formula is as follows:

[0052]

[0053] in, It means to find the values ​​of feature matrices P and Q so that the value of the loss function is minimized. is the feature vector of user u, is the feature vector of artwork i, is the set of rated values ​​for user u and artwork i, is the regularization parameter, which controls the influence of the regularization term and balances the prediction error and model complexity in the loss function;

[0054] According to step S307, the matrices of P and Q are updated using the gradient descent method:

[0055]

[0056] in, is the learning rate;

[0057] According to step S308, the trained P and Q matrices are used to predict the rating of user u on artwork i, and the formula is as follows:

[0058] .

[0059] According to step S309, the performance of the model is evaluated using the root mean square error, and the formula is as follows:

[0060]

[0061] Among them, RMSE is the root mean square error result;

[0062] Based on the predicted ratings, recommend the highest-rated unviewed artworks to each user:

[0063]

[0064] in, is the function that generates the recommendation list. Indicates that we want to find the artwork collection that user u has not visited yet, such that The largest artwork i.

[0065] According to step S400, a multi-language support function is established on the cloud platform, including multi-language display of the interface, exhibition introduction, artwork explanation, etc. Users can select the interface language according to their language preference, and when browsing artworks, they can select the corresponding language for explanation and understanding.

[0066] Design the exhibition content and layout according to the characteristics of artworks from different regions and cultural backgrounds. For example, for Western artworks, you can design a special exhibition to display paintings from different historical periods; for Eastern artworks, you can design an exhibition to display traditional artworks such as Chinese calligraphy and painting, and Japanese Ukiyo-e. The exhibition content should be representative and reflect the artistic style and characteristics of each region and culture.

[0067] Provide cultural interpretation services for different exhibits, including the historical background, cultural connotation, artist introduction, etc. Provide interpretation services in corresponding languages ​​for exhibits from different regions and cultures to help users better understand and appreciate artworks.

[0068] Provides a communication forum function, allowing users to discuss and exchange ideas on exhibition content, artwork themes, etc. The forum supports multilingual communication, providing a platform for users from different regions and cultural backgrounds to promote cross-cultural art communication and understanding.

[0069] A multi-source data fusion analysis and push system based on a cloud platform, comprising:

[0070] Artwork data acquisition and display module: including: data source acquisition unit, data integration unit and virtual exhibition and interactive display unit; wherein, the data source acquisition unit is responsible for acquiring artwork data through several data sources, and the artwork data includes artwork display data, artist background information, artwork collection value and artwork reviews; the data integration unit is responsible for integrating the artwork data acquired from various data sources and integrating it into the database of the cloud platform; the virtual exhibition and interactive display unit is responsible for providing virtual exhibition and interactive display functions of artworks based on the cloud platform, and users can browse artworks through the cloud platform and perform interactive operations;

[0071] User access data collection and visualization module: including: access data collection unit, real-time data analysis unit and visualization display unit; the access data collection unit is responsible for collecting access data of different users on the cloud platform, the access data includes the number of page views, dwell time, click hot spots, collection preferences and user interactive behaviors of the exhibition page, the real-time data analysis unit is responsible for real-time analysis of user access data, generating the popularity and trends of different users' visits, and the visualization display unit is responsible for displaying the analyzed user access data to the administrator in a visualized form;

[0072] Personalized recommendation module: including: data preprocessing and feature extraction unit, collaborative filtering algorithm training unit and personalized recommendation generation unit; among them, the data preprocessing and feature extraction unit is responsible for preprocessing and feature extraction of user access data and establishing user portraits; the collaborative filtering algorithm training unit is responsible for training personalized recommendation models based on user portraits and artwork features; the personalized recommendation generation unit is responsible for predicting the user's preference for artworks that have not been browsed based on the user's interests and needs, using the trained personalized recommendation model, and thus generating personalized artwork recommendations;

[0073] Cross-cultural art exchange module: includes: cross-cultural artwork data collection unit, multilingual support and cultural interpretation unit and exchange forum and community function unit. Among them, the cross-cultural artwork data collection unit is responsible for acquiring artwork data from different regions and cultural backgrounds, the multilingual support and cultural interpretation unit is responsible for establishing multilingual support functions on the cloud platform to provide users with multilingual exhibition and interpretation services, and the exchange forum and community function unit is responsible for providing cultural interpretation and exchange forum functions.

[0074] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0075] The present invention utilizes a cloud platform to integrate multiple data sources, including artwork display data, artist background information, artwork collection value, etc., to achieve comprehensive management and integrated display of artwork information.

[0076] The present invention realizes personalized artwork recommendation service by analyzing users' access data and interest preferences and using machine learning algorithms. At the same time, through the virtual exhibition and interactive display functions provided by the cloud platform, users can obtain a richer and more personalized artwork experience, which increases user participation and satisfaction compared to traditional static display methods.

[0077] The present invention promotes cross-cultural art communication and cooperation by establishing multi-language support and cultural interpretation functions, and collecting artwork data from different regions and cultural backgrounds. Users can appreciate artworks from all over the world on the platform, understand the artistic expressions and connotations under different cultural backgrounds, and enhance users' cultural cognition and communication experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0079] Figure 1 It is a schematic diagram of the steps of a multi-source data fusion analysis and push method based on a cloud platform of the present invention;

[0080] Figure 2 It is a system structure diagram of a multi-source data fusion analysis and push system based on a cloud platform of the present invention. DETAILED DESCRIPTION

[0081] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0082] See also Figure 1 and Figure 2 , the present invention provides a technical solution:

[0083] According to one embodiment of the present invention, Figure 1 A method for fusion analysis and push of multi-source data based on a cloud platform is shown in a schematic diagram of steps. A method for fusion analysis and push of multi-source data based on a cloud platform comprises the following steps:

[0084] S100, obtaining artwork data from a number of data sources, wherein the artwork data includes artwork display data, artist background information, artwork collection value, and artwork reviews, and integrating the data into a cloud platform, and providing virtual exhibitions and interactive displays of artworks through the cloud platform; using a natural language processing algorithm to analyze the historical background, cultural connotation, style characteristics, and artistic expression techniques of the artwork, and integrating them to form artwork features;

[0085] S200, collecting access data of different users on the cloud platform, wherein the access data includes the number of page views, dwell time, click hot spots, favorite preferences and user interaction behaviors of the exhibition page, generating the popularity and trends of different users' visits in real time, and performing visual display for the administrator;

[0086] S300, analyzing the user's access data, performing data preprocessing and feature extraction, and establishing a user profile; using a collaborative filtering algorithm, training a personalized recommendation model based on the user profile and artwork features. The goal of the model training is to predict the user's preference for artworks that have not been viewed based on the user's interests and needs, so as to achieve personalized recommendations;

[0087] S400. Based on the artwork data from different regions and cultural backgrounds, establish multi-language support functions on the cloud platform, design exhibition and interpretation services in combination with the characteristics of artworks from different regions and cultures, and provide cultural interpretation and exchange forum functions.

[0088] According to step S100, the exhibition space of the virtual exhibition is designed on the cloud platform through three-dimensional modeling, and the layout is designed according to different exhibition themes and characteristics of artworks. The exhibition space can simulate the exhibition hall of a traditional art gallery or be designed into a scene that matches the style of the artwork;

[0089] The design of the exhibition space includes: the size, shape, walls, floors, lighting, art display racks and cabinets, navigation and paths, interactive elements, sound effects and user behavior tracking of the exhibition hall; users can interact with the artwork by clicking, zooming, rotating or dragging, and set up an independent display area for each artwork in the exhibition space, providing an introduction to the artwork's historical background, cultural connotation, style characteristics and artistic expression techniques; users can freely browse the virtual exhibition space and choose exhibition areas of interest to explore and view.

[0090] According to step S100, the historical background, cultural connotation, style characteristics and artistic expression of the artwork are analyzed using a natural language processing algorithm, and text cleaning, word segmentation and part-of-speech tagging, stop word filtering, morphological restoration and stem extraction are performed through preprocessing;

[0091] In view of historical background and cultural connotation, we use the pre-trained NER model to identify and extract entities of time, place, and person in the text, use the LDA model to analyze the text, extract the main themes and keywords, analyze the sentiment tendency in the text, and use the VADER sentiment analysis model to obtain the cultural connotation of artworks and artists and the emotional color of historical background;

[0092] Based on the style characteristics and artistic expressions, TF-IDF is used to extract keywords describing the artistic style and expressions, and the syntax analysis tool SpaCy is used to parse the syntactic structure of the text to obtain the relationships in the description of the artwork;

[0093] The extracted entities, topics, emotions, keywords and syntactic structures are converted into feature vectors, and the word embedding model Word2Vec is used to convert text features into feature vectors; the feature vectors are concatenated and weighted averaged to form artwork features.

[0094] According to step S200, the page views record the number of page views of each exhibition page, the dwell time records the time the user stays on each page, the click hotspot records the user's click behavior on the page to determine the hotspot area, the collection preference records the artworks and exhibitions collected by the user, and the interactive behavior records the user's interactive behavior in the exhibition;

[0095] When collecting data, JavaScript code is embedded in the front-end page to capture users' browsing and interaction behaviors, record user request data through server logs, and use Google Analytics to collect user behavior data.

[0096] Step S300 includes:

[0097] S301, collect user access data, analyze the user access data, perform data preprocessing and feature extraction, generate user access data feature vectors, and establish user portraits;

[0098] S302, constructing a user-artwork rating matrix, wherein each element in the matrix corresponds to a user's rating of the artwork;

[0099] S303, using cosine similarity to calculate the similarity between two users to form a user similarity matrix; based on the user similarity matrix, predicting the user's rating for the artwork that has not been visited;

[0100] S304, using cosine similarity to calculate the similarity between two artworks to form an artwork similarity matrix; based on the artwork similarity matrix, predicting the user's rating for the artwork that has not been visited;

[0101] S305, decomposing the user-artwork rating matrix into two low-rank matrices by matrix decomposition, wherein the two low-rank matrices include a user feature matrix and an artwork feature matrix;

[0102] S306, solving the user feature matrix and the artwork feature matrix by minimizing the loss function;

[0103] S307, using the gradient descent method to update the user feature matrix and the artwork feature matrix;

[0104] S308, using the trained user feature matrix and artwork feature matrix to predict the user's rating of the artwork;

[0105] S309. Use the cross-validation method to evaluate the performance of the model. The evaluation indicator used is the root mean square error. According to the predicted score, recommend the highest-rated unviewed artwork to each user.

[0106] In this embodiment, the number of users is 3 (user 1, user 2, user 3), the number of artworks is 4 (artwork A, artwork B, artwork C, artwork D), and according to step S302, a user-artwork rating matrix R is constructed, and the matrix is ​​as follows:

[0107]

[0108] The rating is calculated by weighted calculation based on the feature vectors of the user's user profile, including the number of views, dwell time, click hotspots, collection preferences, and user interaction behavior. Among them, user 1 rated artwork A 5, artwork B 3, and artwork D 1, user 2 rated artwork A 4, and artwork D 1, and user 3 rated artwork A 1, artwork B 1, and artwork D 5.

[0109] According to step S303, the similarity between users u and v is calculated using cosine similarity, and the formula is as follows:

[0110]

[0111] in, Represents the similarity between users u and v;

[0112] In this embodiment, the similarity between user 1 and user 2 is calculated:

[0113]

[0114] In this embodiment, the similarity between user 1 and user 3 is calculated:

[0115]

[0116] In this embodiment, the similarity between user 2 and user 3 is calculated:

[0117]

[0118] Then we get the following user similarity matrix:

[0119]

[0120] Based on the user similarity matrix, predict the score of user u for artwork i that has not been visited. The formula is as follows:

[0121]

[0122] in, is the predicted rating of user u on artwork i, is the average rating of user u, is a set of neighbor users similar to user u;

[0123] According to step S304, the similarity between artworks i and j is calculated using cosine similarity, and the formula is as follows:

[0124]

[0125] in, is the similarity between artworks i and j;

[0126] For example, in this embodiment, the similarity between artwork A and artwork B is calculated:

[0127]

[0128] The following artwork similarity matrix is ​​obtained, omitting some calculations:

[0129]

[0130] Based on the artwork similarity matrix, predict the score of user u for artwork i that has not been visited. The formula is as follows:

[0131]

[0132] in, is the average rating of artwork i, is the set of neighbor artworks similar to artwork i.

[0133] According to step S305, the user-artwork rating matrix R is decomposed into two low-rank matrices P and Q by matrix decomposition, where P is the user feature matrix and Q is the artwork feature matrix, and the formula is as follows:

[0134]

[0135] in, , , k is the dimension of the eigenvector, and T represents the transpose of the matrix; in this embodiment, k=2, then the following initial matrix is ​​obtained:

[0136] , ;

[0137] According to step S306, P and Q are solved by minimizing the following loss function, and the formula is as follows:

[0138]

[0139] in, It means to find the values ​​of feature matrices P and Q so that the value of the loss function is minimized. is the feature vector of user u, is the feature vector of artwork i, is the set of rated values ​​for user u and artwork i, is the regularization parameter, which controls the influence of the regularization term and balances the prediction error and model complexity in the loss function;

[0140] According to step S307, the matrices of P and Q are updated using the gradient descent method:

[0141]

[0142] in, is the learning rate, which is set to 0.01 in this embodiment;

[0143] According to step S308, the trained P and Q matrices are used to predict the rating of user u on artwork i, and the formula is as follows:

[0144] .

[0145] Among them, the trained P and Q matrices are as follows:

[0146] , ;

[0147] For example, to calculate the predicted rating of user 1 for artwork C:

[0148]

[0149] According to step S309, the performance of the model is evaluated using the root mean square error, and the formula is as follows:

[0150]

[0151] Among them, RMSE is the root mean square error result;

[0152] Based on the predicted ratings, recommend the highest-rated unviewed artworks to each user:

[0153]

[0154] in, is the function that generates the recommendation list. Indicates that we want to find the artwork collection that user u has not visited yet, such that The largest artwork i.

[0155] For user 1, the unviewed artworks are C and D, and the predicted scores are as follows:

[0156] Artwork C: 0.93, Artwork D: 1.15, therefore, the unviewed artwork with the highest recommendation score is Artwork D.

[0157] According to step S400, a multi-language support function is established on the cloud platform, including multi-language display of the interface, exhibition introduction, artwork explanation, etc. Users can select the interface language according to their language preference, and when browsing artworks, they can select the corresponding language for explanation and understanding.

[0158] Design the exhibition content and layout according to the characteristics of artworks from different regions and cultural backgrounds. For example, for Western artworks, you can design a special exhibition to display paintings from different historical periods; for Eastern artworks, you can design an exhibition to display traditional artworks such as Chinese calligraphy and painting, and Japanese Ukiyo-e. The exhibition content should be representative and reflect the artistic style and characteristics of each region and culture.

[0159] Provide cultural interpretation services for different exhibits, including the historical background, cultural connotation, artist introduction, etc. Provide interpretation services in corresponding languages ​​for exhibits from different regions and cultures to help users better understand and appreciate artworks.

[0160] Provides a communication forum function, allowing users to discuss and exchange ideas on exhibition content, artwork themes, etc. The forum supports multilingual communication, providing a platform for users from different regions and cultural backgrounds to promote cross-cultural art communication and understanding.

[0161] A multi-source data fusion analysis and push system based on a cloud platform, comprising:

[0162] Artwork data acquisition and display module: including: data source acquisition unit, data integration unit and virtual exhibition and interactive display unit; wherein, the data source acquisition unit is responsible for acquiring artwork data through several data sources, and the artwork data includes artwork display data, artist background information, artwork collection value and artwork reviews; the data integration unit is responsible for integrating the artwork data acquired from various data sources and integrating it into the database of the cloud platform; the virtual exhibition and interactive display unit is responsible for providing virtual exhibition and interactive display functions of artworks based on the cloud platform, and users can browse artworks through the cloud platform and perform interactive operations;

[0163] User access data collection and visualization module: including: access data collection unit, real-time data analysis unit and visualization display unit; the access data collection unit is responsible for collecting access data of different users on the cloud platform, the access data includes the number of page views, dwell time, click hot spots, collection preferences and user interactive behaviors of the exhibition page, the real-time data analysis unit is responsible for real-time analysis of user access data, generating the popularity and trends of different users' visits, and the visualization display unit is responsible for displaying the analyzed user access data to the administrator in a visualized form;

[0164] Personalized recommendation module: including: data preprocessing and feature extraction unit, collaborative filtering algorithm training unit and personalized recommendation generation unit; among them, the data preprocessing and feature extraction unit is responsible for preprocessing and feature extraction of user access data and establishing user portraits; the collaborative filtering algorithm training unit is responsible for training personalized recommendation models based on user portraits and artwork features; the personalized recommendation generation unit is responsible for predicting the user's preference for artworks that have not been browsed based on the user's interests and needs, using the trained personalized recommendation model, and thus generating personalized artwork recommendations;

[0165] Cross-cultural art exchange module: includes: cross-cultural artwork data collection unit, multilingual support and cultural interpretation unit and exchange forum and community function unit. Among them, the cross-cultural artwork data collection unit is responsible for acquiring artwork data from different regions and cultural backgrounds, the multilingual support and cultural interpretation unit is responsible for establishing multilingual support functions on the cloud platform to provide users with multilingual exhibition and interpretation services, and the exchange forum and community function unit is responsible for providing cultural interpretation and exchange forum functions.

[0166] In this embodiment, a language switching option is added to the platform interface to support at least ten major languages ​​(such as English, Chinese, Spanish, French, German, Japanese, Korean, Arabic, Russian, and Portuguese). All artwork descriptions, exhibition information, and cultural interpretation content are translated into multiple languages ​​using a combination of machine translation and manual proofreading. A professional team of art history and cultural experts is formed to write and review detailed interpretation content of artworks, including historical background, cultural connotation, artistic style, and expression techniques. Audio and video interpretation is used to provide multimedia interpretation services so that users can understand artworks more intuitively by listening and watching. An interactive exhibition function is designed so that users can click on artworks to obtain detailed information and interpretation. A virtual tour function is provided, where users can choose a language and be led by a virtual tour guide to visit and explain.

[0167] Establish a themed discussion section and create discussion topics based on different types of artworks, regional cultures, artistic styles, etc. Provide a multilingual forum where users can choose their own language to post and reply, and provide real-time translation functions to help cross-language communication. After users register and log in, they can join different art interest groups (such as Impressionist art, niche cultural art, etc.). Each interest group has its own activity page, where users can post activities, share artworks, and organize online and offline art exchange activities. Users can like, comment on, and share artworks and interpretation content to increase interactivity. Set up a feedback mechanism so that users can make suggestions for improvements to the platform's multilingual support and cultural interpretation content, and the platform will optimize based on user feedback.

[0168] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0169] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A multi-source data fusion analysis and push method based on a cloud platform, characterized in that: The method comprises the following steps: S100, obtaining artwork data through several data sources, wherein the artwork data includes artwork display data, artist background information, artwork collection value and artwork reviews, and integrating the data into a cloud platform, and providing virtual exhibitions and interactive displays of artworks through the cloud platform; using a natural language processing algorithm to analyze the historical background, cultural connotation, style characteristics and artistic expression techniques of the artwork, and integrating them to form artwork features; displaying artwork images to users losslessly and completely, so that users can locate, mark and search for artwork images; According to step S100, the historical background, cultural connotation, style characteristics and artistic expression of the artwork are analyzed using a natural language processing algorithm, and text cleaning, word segmentation and part-of-speech tagging, stop word filtering, morphological restoration and stem extraction are performed through preprocessing; In view of historical background and cultural connotation, we use the pre-trained NER model to identify and extract entities of time, place, and person in the text, use the LDA model to analyze the text, extract the main themes and keywords, analyze the sentiment tendency in the text, and use the VADER sentiment analysis model to obtain the cultural connotation of artworks and artists and the emotional color of historical background; Based on the style characteristics and artistic expressions, TF-IDF is used to extract keywords describing the artistic style and expressions, and the syntax analysis tool SpaCy is used to parse the syntactic structure of the text to obtain the relationships in the description of the artwork; The extracted entities, topics, sentiments, keywords and syntactic structures are converted into feature vectors, and the word embedding model Word2Vec is used to convert text features into feature vectors; the feature vectors are concatenated and weighted averaged to form artwork features; S200, collecting access data of different users on the cloud platform, wherein the access data includes the number of page views, dwell time, click hot spots, favorite preferences and user interaction behaviors of the exhibition page, generating the popularity and trends of different users' visits in real time, and performing visual display for the administrator; S300, analyzing the user's access data, performing data preprocessing and feature extraction, and establishing a user profile; using a collaborative filtering algorithm, training a personalized recommendation model based on the user profile and artwork features. The goal of the model training is to predict the user's preference for artworks that have not been viewed based on the user's interests and needs, so as to achieve personalized recommendations; Step S300 includes: S301, collect user access data, analyze the user access data, perform data preprocessing and feature extraction, generate user access data feature vectors, and establish user portraits; S302, constructing a user-artwork rating matrix, wherein each element in the matrix corresponds to a user's rating of the artwork; S303, using cosine similarity to calculate the similarity between two users to form a user similarity matrix; based on the user similarity matrix, predicting the user's rating for the artwork that has not been visited; S304, using cosine similarity to calculate the similarity between two artworks to form an artwork similarity matrix; based on the artwork similarity matrix, predicting the user's rating for the artwork that has not been visited; S305, decomposing the user-artwork rating matrix into two low-rank matrices by matrix decomposition, wherein the two low-rank matrices include a user feature matrix and an artwork feature matrix; S306, solving the user feature matrix and the artwork feature matrix by minimizing the loss function; S307, using the gradient descent method to update the user feature matrix and the artwork feature matrix; S308, using the trained user feature matrix and artwork feature matrix to predict the user's rating of the artwork; S309, using a cross-validation method to evaluate the performance of the model, using the root mean square error as the evaluation indicator, and recommending the highest-rated unviewed artwork to each user based on the predicted score; S400. Based on the artwork data from different regions and cultural backgrounds, establish multi-language support functions on the cloud platform, design exhibition and interpretation services in combination with the characteristics of artworks from different regions and cultures, and provide cultural interpretation and exchange forum functions.

2. According to claim 1, a multi-source data fusion analysis and push method based on a cloud platform is characterized in that: According to step S100, the exhibition space of the virtual exhibition is designed on the cloud platform through three-dimensional modeling, and the layout is designed according to different exhibition themes and characteristics of artworks. The exhibition space can simulate the exhibition hall of a traditional art gallery or be designed into a scene that matches the style of the artwork; The design of the exhibition space includes: the size, shape, walls, floors, lighting, art display racks and display cabinets, navigation and paths, interactive elements, sound effects and user behavior tracking of the exhibition hall; Users can interact with artworks by clicking, zooming, rotating or dragging, and set up independent display areas for each artwork in the exhibition space, providing an introduction to the artwork's historical background, cultural connotations, style characteristics and artistic expression techniques; users can freely browse the virtual exhibition space and select exhibition areas of interest to explore and view.

3. The multi-source data fusion analysis and push method based on a cloud platform according to claim 1 is characterized in that: According to step S200, the page views record the number of page views of each exhibition page, the dwell time records the time the user stays on each page, the click hotspot records the user's click behavior on the page to determine the hotspot area, the collection preference records the artworks and exhibitions collected by the user, and the interactive behavior records the user's interactive behavior in the exhibition; When collecting data, JavaScript code is embedded in the front-end page to capture users' browsing and interaction behaviors, record user request data through server logs, and use Google Analytics to collect user behavior data.

4. The multi-source data fusion analysis and push method based on a cloud platform according to claim 1 is characterized in that: According to step S302, a user-artwork rating matrix R is constructed, in which each element r ui represents the rating of user u on artwork i, which is weighted based on the feature vectors of browsing volume, dwell time, click hotspots, collection preferences, and user interaction behavior in the user profile. The user-artwork rating matrix R is expressed as follows: Where n represents the number of users and m represents the number of artworks.

5. The multi-source data fusion analysis and push method based on a cloud platform according to claim 1 is characterized in that: According to step S303, the similarity between users u and v is calculated using cosine similarity, and the formula is as follows: Among them, sim(u,v) represents the similarity between users u and v; Based on the user similarity matrix, predict the score of user u for artwork i that has not been visited. The formula is as follows: in, is the predicted rating of user u on artwork i, is the average rating of user u, N(u) is the set of neighboring users similar to user u; According to step S304, the similarity between artworks i and j is calculated using cosine similarity, and the formula is as follows: Among them, sim(i,j) is the similarity between artworks i and j; Based on the artwork similarity matrix, predict the score of user u for artwork i that has not been visited. The formula is as follows: in, is the average rating of artwork i, and N(i) is the set of neighboring artworks similar to artwork i.

6. The multi-source data fusion analysis and push method based on a cloud platform according to claim 1 is characterized in that: According to step S305, the user-artwork rating matrix R is decomposed into two low-rank matrices P and Q by matrix decomposition, where P is the user feature matrix and Q is the artwork feature matrix, and the formula is as follows: R≈PQ T Where P∈R n*k , Q∈R m*k , k is the dimension of the eigenvector, and T represents the transpose of the matrix; According to step S306, P and Q are solved by minimizing the following loss function, and the formula is as follows: Among them, min P,Q It means to find the values ​​of feature matrices P and Q so that the value of the loss function is minimized, p u is the feature vector of user u, q i is the feature vector of artwork i, κ is the rated set of user u and artwork i, and λ is the regularization parameter, which controls the influence of the regularization term and balances the prediction error and model complexity in the loss function; According to step S307, the matrices of P and Q are updated using the gradient descent method: Where γ is the learning rate; According to step S308, the trained P and Q matrices are used to predict the rating of user u on artwork i, and the formula is as follows:

7. The multi-source data fusion analysis and push method based on a cloud platform according to claim 1 is characterized in that: According to step S309, the performance of the model is evaluated using the root mean square error, and the formula is as follows: Among them, RMSE is the root mean square error result; Based on the predicted ratings, recommend the highest-rated unviewed artworks to each user: Among them, Recommendation(u) is the function of generating the recommendation list. Indicates that we want to find the artwork collection that user u has not visited yet, such that The largest artwork i.

8. A multi-source data fusion analysis push system based on a cloud platform, using a multi-source data fusion analysis push method based on a cloud platform according to any one of claims 1 to 7, characterized in that: include: Artwork data acquisition and display module: including: data source acquisition unit, data integration unit, virtual exhibition and interactive display unit and artwork image processing unit; wherein, the data source acquisition unit is responsible for acquiring artwork data through several data sources, and the artwork data includes artwork display data, artist background information, artwork collection value and artwork reviews; the data integration unit is responsible for integrating the artwork data acquired from various data sources and integrating it into the database of the cloud platform; the virtual exhibition and interactive display unit is responsible for providing virtual exhibition and interactive display functions of artworks based on the cloud platform, and users can browse artworks and perform interactive operations through the cloud platform; the artwork image processing unit is responsible for displaying artwork images to users losslessly and completely, and users can locate, mark and retrieve artwork images; User access data collection and visualization module: including: access data collection unit, real-time data analysis unit and visualization display unit; the access data collection unit is responsible for collecting access data of different users on the cloud platform, the access data includes the number of page views, dwell time, click hot spots, collection preferences and user interactive behaviors of the exhibition page, the real-time data analysis unit is responsible for real-time analysis of user access data, generating the popularity and trends of different users' visits, and the visualization display unit is responsible for displaying the analyzed user access data to the administrator in a visualized form; Personalized recommendation module: including: data preprocessing and feature extraction unit, collaborative filtering algorithm training unit and personalized recommendation generation unit; among them, the data preprocessing and feature extraction unit is responsible for preprocessing and feature extraction of user access data and establishing user portraits; the collaborative filtering algorithm training unit is responsible for training personalized recommendation models based on user portraits and artwork features; the personalized recommendation generation unit is responsible for predicting the user's preference for artworks that have not been browsed based on the user's interests and needs, using the trained personalized recommendation model, and thus generating personalized artwork recommendations; Cross-cultural art exchange module: includes: cross-cultural artwork data collection unit, multilingual support and cultural interpretation unit and exchange forum and community function unit. Among them, the cross-cultural artwork data collection unit is responsible for acquiring artwork data from different regions and cultural backgrounds, the multilingual support and cultural interpretation unit is responsible for establishing multilingual support functions on the cloud platform to provide users with multilingual exhibition and interpretation services, and the exchange forum and community function unit is responsible for providing cultural interpretation and exchange forum functions.

Images