Target data processing method and system

Abstract

The invention discloses a target data processing method and system, and relates to the technical field of data processing, and the method comprises the steps: determining a plurality of sub-regions according to a target image; according to the method, the multiple pieces of sub-target data are determined based on the importance levels of the multiple sub-regions, the semantic information of the multiple sub-regions and the internet database, data retrieval of the target image is achieved, and control of the sub-target data in multiple dimensions is guaranteed. Therefore, a plurality of data combinations are determined according to the plurality of sub-target data and the target image; if the combination matching coefficients of the plurality of data combinations are all lower than a preset matching coefficient threshold, determining supplementary data according to the combination matching coefficients of the plurality of data combinations and the source information of the target image, and determining target data according to the supplementary data, the plurality of data combinations and the data preference factors of the user, according to the invention, supplementary processing of multiple sub-target data is realized, and the accuracy of the target data is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method and system for processing target data. Background Art

[0002] With the development of science and technology, images are gradually applied to people's lives and presented in Internet interfaces. Users search for data on the Internet through target images to retrieve relevant data. In the existing technology, users search the Internet based on target images and obtain multiple target data associated with the target images. However, the matching coefficient between the target data and the target image is low, and the accuracy of the target data cannot be guaranteed. Summary of the Invention

[0003] The purpose of the present invention is to overcome the deficiencies of the prior art, and the present invention provides a method and system for processing target data.

[0004] An embodiment of the present invention provides a method for processing target data, including:

[0005] Determine the target image based on the user's web browsing history and the user's search key points;

[0006] determining a plurality of sub-regions according to the target image;

[0007] determining a plurality of sub-target data based on the importance levels of the plurality of sub-regions, the semantic information of the plurality of sub-regions, and an Internet database, wherein the plurality of sub-target data are Internet data of the target image at different dimensions;

[0008] determining a plurality of data combinations according to the plurality of sub-target data and the target image;

[0009] If the combined matching coefficients of multiple data combinations are all lower than the preset matching coefficient threshold, the supplementary data is determined based on the combined matching coefficients of the multiple data combinations and the source information of the target image, and the target data is determined based on the supplementary data, multiple data combinations and the user's data preference factors.

[0010] An embodiment of the present invention provides a target data processing system, which is applied to the above-mentioned target data processing method. The target data processing system includes:

[0011] A target image module is used to determine a target image based on the user's web browsing history and the user's search key points;

[0012] A sub-region module, used for determining a plurality of sub-regions according to a target image;

[0013] A sub-target data module is used to determine a plurality of sub-target data based on the importance levels of the plurality of sub-regions, the semantic information of the plurality of sub-regions, and an Internet database, wherein the plurality of sub-target data are Internet data of the target image at different dimensions;

[0014] A data combination module, configured to determine a plurality of data combinations according to a plurality of sub-target data and a target image;

[0015] The target data module is used to determine the supplementary data based on the combined matching coefficients of the multiple data combinations and the source information of the target image if the combined matching coefficients of the multiple data combinations are all lower than the preset matching coefficient threshold, and to determine the target data based on the supplementary data, the multiple data combinations and the user's data preference factors.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] In an embodiment of the present invention, through the method in the embodiment of the present invention, a target image is determined based on a user's web browsing history and the user's search points; multiple sub-regions are determined based on the target image; multiple sub-target data are determined based on the importance levels of the multiple sub-regions, the semantic information of the multiple sub-regions and an Internet database. The multiple sub-target data are respectively the Internet data of the target image in different dimensions, which are compatible with the importance levels of the multiple sub-regions, the semantic information of the multiple sub-regions and the overall consideration of the Internet database, thereby realizing data retrieval of the target image and ensuring the control of the sub-target data in multiple dimensions.

[0018] Therefore, multiple data combinations are determined based on multiple sub-target data and target images; if the combined matching coefficients of the multiple data combinations are all lower than the preset matching coefficient threshold, supplementary data are determined based on the combined matching coefficients of the multiple data combinations and the source information of the target image, and the target data is determined based on the supplementary data, the multiple data combinations and the user's data preference factors, thereby realizing supplementary processing of multiple sub-target data and improving the accuracy of the target data. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a flowchart of a method for processing target data in an embodiment of the present invention;

[0020] Figure 2 is a flow chart of step S11 in the method for processing target data in an embodiment of the present invention;

[0021] Figure 3 is a flow chart of step S12 in the method for processing target data in an embodiment of the present invention;

[0022] Figure 4 is a flow chart of step S13 in the method for processing target data in an embodiment of the present invention;

[0023] Figure 5 is a flow chart of step S14 in the method for processing target data in an embodiment of the present invention;

[0024] Figure 6 is a flow chart of step S15 in the method for processing target data in an embodiment of the present invention;

[0025] Figure 7 It is a schematic diagram of the structure of the target data processing system in an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present invention.

[0027] See also Figures 1 to 7 A target data processing method is applied to a target data processing scenario; the target data processing method includes:

[0028] Step S11: determining a target image based on the user's web browsing history and the user's search criteria;

[0029] Step S12: determining a plurality of sub-regions according to the target image;

[0030] Step S13: determining a plurality of sub-target data based on the importance levels of the plurality of sub-regions, the semantic information of the plurality of sub-regions, and an Internet database, wherein the plurality of sub-target data are Internet data of the target image at different dimensions;

[0031] Step S14: determining multiple data combinations according to the multiple sub-target data and the target image;

[0032] Step S15: If the combined matching coefficients of the multiple data combinations are all lower than the preset matching coefficient threshold, then determining supplementary data based on the combined matching coefficients of the multiple data combinations and the source information of the target image, and determining the target data based on the supplementary data, the multiple data combinations, and the user's data preference factors;

[0033] refer to Figure 2 ,In step S11, a target image is determined based on the user's web browsing history and the user's search key points;

[0034] In the specific implementation process of the present invention, the specific steps are:

[0035] S111: collecting the user's search key points, determining a search combination based on the search key points, triggering a corresponding network search based on the search combination, and generating multiple first pages;

[0036] S112: Obtaining a user's web browsing history, and determining a plurality of second pages based on the user's web browsing history and the user's browsing mark, and determining a comprehensive page based on the plurality of second pages and the plurality of first pages;

[0037] S113: The user selects a target on the comprehensive page and determines a corresponding target image.

[0038] In an embodiment of the present application, a user's search key points are collected, a search combination is determined based on the search key points, and a corresponding network search is triggered based on the search combination, and multiple first pages are generated;

[0039] At this time, users enter the information or keywords they want to query through the search box, voice assistant or other input methods; in addition to direct input, the user's click behavior, historical search records or browsing habits are also analyzed to indirectly obtain the user's search points; optionally, user Xiao Zhang enters "autumn fashion wear" as the search point in the search engine.

[0040] Based on the keywords entered by the user, use natural language processing (NLP) technology or keyword expansion tools to generate search combinations that are related to the original keywords but more specific; understand the semantics of the user input to ensure that the search combination can accurately reflect the user's query intention; optionally, based on "Autumn Fashion Outfits", the system generates the following search combinations: "2023 Autumn Fashion Outfit Trends", "Autumn Women's Clothing Recommendations", "Autumn Fashion Item Matching", etc.

[0041] Use the search engine API or built-in search engine to perform network searches based on the generated search combinations; in order to improve the diversity and accuracy of the search results, retrieve information from multiple search engines or data sources; optionally, the system uses the Google search engine API to perform searches based on search combinations such as "2023 Autumn Fashion Wear Trends", "Autumn Women's Wear Recommendations" and "Autumn Fashion Item Matching".

[0042] Furthermore, the pages that are most relevant to the user's query intention and have the highest quality are filtered out from the search results; the filtered pages are sorted according to factors such as the relevance, authority, and update frequency of the pages; the top N pages after sorting (N is a preset value) are used as the first page for the user to browse or further process; optionally, the system filters out the top 10 pages related to "autumn fashion wear" from the search results, and these pages come from fashion magazine websites, e-commerce platforms, social media, etc.; the system sorts these pages from high to low according to relevance, and generates a first page list containing the top 10 pages.

[0043] Specifically, suppose user Xiao Zhang wants to know the fashion wear trends in the fall of 2023. He enters "autumn fashion wear" as the search key in the search engine; the system generates multiple search combinations based on this keyword, such as "2023 autumn fashion wear trends", "autumn women's clothing recommendations", etc.; then, the system uses the search engine API to perform network searches for these search combinations respectively, and filters out the top 10 most relevant and highest-quality pages from the search results; these pages are sorted and generate the first page list for Xiao Zhang to browse; Xiao Zhang obtains information about the latest trends, recommended items, and matching suggestions for the fall of 2023 from these pages.

[0044] Furthermore, the user's web browsing history is obtained, and multiple second pages are determined based on the user's web browsing history and the user's browsing marks. A comprehensive page is determined based on the multiple second pages and the multiple first pages, which is compatible with the overall consideration of the multiple second pages and the multiple first pages, ensuring the accuracy of the comprehensive page.

[0045] At this time, users' web browsing records usually come from their browser history, cookies, website analysis tools or user behavior tracking systems; the system needs to collect users' browsing records from these data sources regularly or in real time, including the URLs of the web pages they visited, the time of visit, the length of stay, and other information; optionally, user Xiao Zhang visited multiple fashion websites in the past week, including the official website of a well-known fashion magazine, the personal websites of several fashion bloggers, and the fashion sections of several large e-commerce platforms; the system collected the URLs of these websites and the length of time Xiao Zhang stayed on each page and the visit time.

[0046] A second page is introduced, and multiple second pages are determined based on the user's web browsing history and the user's browsing marks. At this time, the user's browsing marks include bookmarks, collections, likes, comments or shares, which reflect the user's interest and preference in the page content; based on the user's browsing history and browsing marks, the system selects the page that the user is most interested in or most relevant to as the second page; this is achieved by calculating indicators such as page visit frequency, stay time, user behavior (such as likes, comments), etc.

[0047] Optionally, in Xiao Zhang's browsing history, the system found that he spent the longest time on the autumn outfit special page of a fashion magazine's official website, and liked and commented on the page; therefore, the system identified this page as one of the second pages; at the same time, the system also filtered out the pages related to autumn outfit that Xiao Zhang had browsed on other websites, and these pages were also regarded as the second pages.

[0048] Analyze the content of the second page and the first page to extract key information, themes, and relevance. Based on factors such as content similarity, complementarity, and user interests, integrate the information of the second page and the first page to generate one or more comprehensive pages. These pages should contain the information that users are most interested in and be presented in a way that is easy to understand and browse.

[0049] Specifically, the system performs content analysis on Xiao Zhang's second page (such as a fashion magazine's autumn outfit special page) and the first page (such as a page about autumn fashion outfit trends generated based on retrieval points); by comparing the themes, keywords, pictures and other information of these pages, the system finds that there is a high correlation between them; therefore, the system integrates the information of these pages to generate a comprehensive autumn outfit guide page; this page contains the latest autumn fashion outfit trends, recommended items, matching suggestions and specific content that users (such as Xiao Zhang) are interested in (such as a special article in a fashion magazine).

[0050] Therefore, the user selects a target on the comprehensive page and determines a corresponding target image, thereby introducing the target image.

[0051] At this point, the system presents the generated comprehensive page to the user, the page content includes text, pictures, videos and other forms of information; users interact with the page by browsing, scrolling, clicking, etc. to find the information they are interested in or need; users identify the target content they want to learn more about or operate from the page based on their needs and interests; this is a specific product, an article, a picture or a video, etc.

[0052] Optionally, user Xiao Zhang browses information about autumn fashion on a comprehensive page; the page contains multiple articles about autumn fashion, recommended pictures of multiple fashion items, and videos with matching suggestions; Xiao Zhang is particularly interested in a picture of autumn jacket matching in one of the articles, and he clicks on the picture to view more details.

[0053] When a user selects a target content (such as an article, product, or video), the system needs to extract the corresponding image from this content; this is usually achieved by parsing HTML / CSS, using image recognition technology, or accessing the API provided by the content provider; the extracted image needs to match the user's selection intent, that is, to ensure that the image is indeed part of the target content that the user wants to view or operate; this is achieved by comparing the image's metadata, tags, or user feedback; once the image is extracted and verified, it is determined to be the corresponding target image and is ready for subsequent processing or display.

[0054] Specifically, after Xiao Zhang clicked on the picture of autumn jacket combinations, the system extracted the URL of the picture from the HTML element of the picture; the system also checked the alt text of the picture and the surrounding text content to ensure that this picture is indeed the autumn jacket combination picture that Xiao Zhang wanted to see; once the verification is passed, the picture is determined as the corresponding target image and is ready to be displayed in Xiao Zhang's browser or used for other purposes (such as saving to the user's favorites, sharing to social media, etc.); through this process, users can select the target content they are interested in from the comprehensive page and determine the corresponding target image; this provides users with an intuitive and interactive way to obtain the information they want, while also providing the system with valuable feedback about user interests and needs.

[0055] In one embodiment of the present application, assume that there is a comprehensive page containing multiple articles and images related to autumn fashion. A user selection matching table is created to record the relationship between the user's selection and the corresponding target image. The user selection matching table is shown in Table 1:

[0056] Table 1 User selection matching table

[0057]

[0058] In this user selection matching table, each row represents a user selection and the corresponding target image; for example, when a user selects the description in article A, the system will find the corresponding target image (a picture of a new autumn windbreaker paired with jeans) according to the matching table and extract its URL for subsequent use.

[0059] refer to Figure 3 , in step S12, a plurality of sub-regions are determined according to the target image;

[0060] In the specific implementation process of the present invention, the specific steps are:

[0061] S121: collecting target information of the user, determining multiple keywords based on keyword screening of the target information, and determining reference weights of the multiple keywords based on the multiple keywords and a reference weight relationship table;

[0062] S122: Acquire a target image, mark multiple semantic parts according to the analysis of the target image, determine multiple semantic combinations according to reference weights of the multiple semantic parts and multiple keywords, and trigger segmentation of the target image according to the multiple voice combinations to determine multiple sub-regions;

[0063] In an embodiment of the present application, the user's target information is collected, multiple keywords are determined based on keyword screening of the target information, and reference weights of the multiple keywords are determined for the multiple keywords and a reference weight relationship table, thereby introducing reference weights of multiple keywords.

[0064] At this point, the user's target information comes from a variety of channels, such as query terms entered into the search box, form content filled out by the user, conversation records between the user and the system, or user intent derived from user behavior analysis. This information is in the form of text, voice, image, or other multimedia. Before further processing, the system must perform necessary preprocessing on the information, such as text cleaning, voice-to-text conversion, and image recognition. Alternatively, suppose user Xiao Zhang enters "new autumn windbreaker for men" into the search box of an e-commerce platform. This is his target information.

[0065] By using natural language processing (NLP) technology, such as word segmentation and part-of-speech tagging, keywords are extracted from the user's target information; keywords most closely related to the user's target are screened out from the extracted keywords; operations such as keyword frequency statistics and semantic similarity calculation are introduced. Optionally, from the user Xiao Zhang's input "new autumn windbreaker for men", the system extracts the keywords "autumn", "new style", "windbreaker" and "men's style"; these keywords are all highly relevant to the user's target information.

[0066] The reference weight relationship table is a predefined table used to store the relative importance of different keywords in specific scenarios or tasks; these weights are determined based on multiple factors such as domain knowledge, user feedback, historical data, etc.; according to the reference weight relationship table, a reference weight is assigned to each filtered keyword; this weight reflects the importance of the keyword in the user's goal; optionally, assume that there is a reference weight relationship table, in which the weight of "windbreaker" is 0.5, the weight of "autumn" and "new style" is 0.25, and the weight of "men's style" is 0.1 (these weight values are preset based on historical data, user preferences and other factors); for user Xiao Zhang's keywords "autumn", "new style", "windbreaker" and "men's style", the system assigns them corresponding weights according to the reference weight relationship table.

[0067] Specifically, suppose user Xiao Zhang enters "new autumn windbreaker for men" in the search box of the e-commerce platform. The system captures Xiao Zhang's input "new autumn windbreaker for men"; the system uses NLP technology to extract the keywords "autumn", "new style", "windbreaker" and "men's style"; the system assigns weights to these keywords according to the preset reference weight relationship table; for example, the weight of "windbreaker" is 0.5, the weights of "autumn" and "new style" are 0.25, and the weight of "men's style" is 0.1; through such a processing flow, the system can accurately understand the user's target information and provide strong support for subsequent search, recommendation or analysis tasks.

[0068] Furthermore, a target image is collected, multiple semantic parts are marked according to the analysis of the target image, and multiple semantic combinations are determined according to the reference weights of the multiple semantic parts and multiple keywords, and the division of the target image is triggered according to the multiple voice combinations to determine multiple sub-regions, which is compatible with the overall consideration of the reference weights of multiple semantic parts and multiple keywords to ensure the accuracy of multiple semantic combinations.

[0069] At this time, the target image is collected. The target image is usually stored in JPEG, PNG, BMP and other formats. The target image is further analyzed and analyzed based on target detection. According to the analysis results, the image is divided into different semantic parts, such as objects, backgrounds, color areas, etc., and labels or tags are assigned to these parts. Optionally, suppose that user Xiao Zhang uploads a picture containing a new autumn windbreaker for men. This is the target image. The system analyzes the picture uploaded by Xiao Zhang, identifies the windbreaker, background (such as trees, sky, etc.), and specific color areas on the windbreaker (such as brown, khaki, etc.) as semantic parts, and assigns corresponding labels to them.

[0070] Match previously determined keywords (such as "autumn", "new style", "windbreaker", "men's style") with the parsed semantic parts; determine multiple semantic combinations based on the reference weights of the keywords and the importance of the semantic parts; these combinations reflect the strength of association between different parts of the image and the user's target keywords; optionally, the system matches the keyword "windbreaker" with the parsed windbreaker part, and associates "autumn" with autumn elements in the background (such as fallen leaves, sky color, etc.), while considering the degree of matching between "new style" and "men's style" and the style, color and other features of the windbreaker; based on these matching results and the reference weights of the keywords, the system determines multiple semantic combinations, such as "windbreaker-new style-brown", "background-autumn-sky", etc.

[0071] Based on the determined semantic combinations, the target image is further divided to determine multiple sub-regions; each sub-region corresponds to one or more semantic combinations, reflecting a specific part of the image that is closely related to the user's target keyword; optionally, the system divides the picture uploaded by Xiao Zhang into multiple sub-regions based on the determined semantic combinations; for example, one sub-region focuses on the style and color of the windbreaker (corresponding to the "windbreaker-new model-brown" semantic combination), and another sub-region displays the autumn elements in the background (corresponding to the "background-autumn-sky" semantic combination).

[0072] Specifically, suppose user Xiao Zhang uploads a picture of a new men's autumn trench coat. The system captures the picture uploaded by Xiao Zhang; the system parses the picture, identifies semantic parts such as the trench coat, background, and specific color areas on the trench coat, and assigns labels to them; the system matches the keywords "trench coat", "autumn", "new style", and "men's style" with the parsed semantic parts, and determines multiple semantic combinations based on the reference weights of the keywords, such as "trench coat-new style-brown" and "background-autumn-sky"; the system divides the picture into multiple sub-areas based on the determined semantic combinations, each sub-area corresponds to one or more semantic combinations, reflecting the specific part of the image that is closely related to the user's target keyword; through such a processing flow, the system can understand the content of the target image more deeply and provide strong support for subsequent image analysis, search or recommendation tasks.

[0073] In one embodiment of the present application, a semantic combination matching table is used to record the relationship between semantic parts, keywords and their reference weights, thereby determining multiple semantic combinations and triggering the segmentation of the target image; the semantic combination matching table is shown in Table 2:

[0074] Table 2 Semantic combination matching table

[0075] Semantic part Keywords Reference weight Semantic Combination Windbreaker area trench coat 0.6 Show the style, color and other features of the windbreaker Background area autumn 0.4 Display autumn elements such as fallen leaves, sky, etc. Button / zipper area New 0.3 Highlight new design details, such as special buttons / zippers Men's feature areas Men's 0.2 Showcases men's-specific design elements, such as pocket placement

[0076] Sub-area 1: Windbreaker area, mainly showing the styles and colors of windbreakers, with a weight of 0.6 (determined by the reference weight of the keyword "windbreaker");

[0077] Sub-region 2: Background region, containing autumn elements such as fallen leaves and sky, with a weight of 0.4 (determined by the reference weight of the keyword "autumn");

[0078] Sub-region 3 (optional): button / zipper area, highlighting the new design details, with a weight of 0.3 (determined by the reference weight of the keyword "new style"; if this area is prominent in the image, it is included);

[0079] Sub-area 4 (optional): Men's style feature area, showing design elements unique to men's style, with a weight of 0.2 (determined by the reference weight of the keyword "men's style"; if this area is prominent in the image and meets the characteristics of men's style, it is included).

[0080] refer to Figure 4 In step S13, a plurality of sub-target data are determined based on the importance levels of the plurality of sub-regions, the semantic information of the plurality of sub-regions, and an Internet database, where the plurality of sub-target data are Internet data of the target image at different dimensions;

[0081] In the specific implementation process of the present invention, the specific steps are:

[0082] S131: The user weights the multiple sub-regions to determine weights of the multiple sub-regions, and determines importance levels of the multiple sub-regions based on the weights of the multiple sub-regions and the user's preference.

[0083] S132: performing semantic analysis on the multiple sub-regions, and introducing semantic information of the multiple sub-regions through the semantic analysis. At this time, presenting semantic contents corresponding to the multiple sub-regions through the semantic information of the multiple sub-regions;

[0084] S133: Determine the content to be retrieved based on the importance levels of the multiple sub-areas and the voice contents corresponding to the multiple sub-areas, and determine multiple sub-target data based on the content to be retrieved and the Internet database. The multiple sub-target data are Internet data of the target image in different dimensions. At the same time, the content to be retrieved is retrieved in the Internet database based on different dimensions, and the different dimensions include regional dimension, time dimension and object type dimension.

[0085] In an embodiment of the present application, the user weights multiple sub-regions to determine the weights of the multiple sub-regions, and determines the importance levels of the multiple sub-regions based on the weights of the multiple sub-regions and the user's preference, which is compatible with the overall consideration of the weights of the multiple sub-regions and the user's preference, and ensures the accuracy of the importance levels of the multiple sub-regions.

[0086] At this point, the user is asked to mark each sub-area with a weight, which reflects the user's interest or attention in the sub-area; the system then determines the importance level of each sub-area based on these weights and the user's overall preference; the user interacts with the system through the interface to mark each sub-area with a weight; the weight marking method is a slider, star rating, numerical input, etc., depending on how the system is implemented; the user assigns a weight value to each sub-area based on his or her interest and attention, and this value is usually within a preset range, such as 0-100, 1-5 stars, etc.

[0087] Optionally, assume that the user is weighting an image containing multiple sub-regions; the sub-regions in the image include: the main body of the windbreaker (sub-region A), the trees in the background (sub-region B), and the button details of the windbreaker (sub-region C); the user believes that the main body of the windbreaker is the most important, so it is assigned the highest weight (such as 5 stars); the trees in the background are second and are assigned a medium weight (such as 3 stars); the button details are relatively less important and are assigned a lower weight (such as 1 star).

[0088] The system collects the weight values assigned by users to each sub-area. In addition to the weight values, the system also considers the overall preference of users, which is obtained through questionnaires, historical behavior analysis, etc. The system comprehensively determines the importance level of each sub-area based on the weight values and the user's preference level. The importance level is discrete (such as high, medium, and low) and continuous (such as a specific numerical range).

[0089] Optionally, after the system collects the user's weight labels for sub-areas A, B, and C, it combines the user's overall preference (such as preference for windbreaker styles); the system integrates this information and determines that sub-area A is high-level (because it has the highest weight and the user likes windbreaker styles very much), sub-area B is medium-level (medium weight and the user has a certain interest in it), and sub-area C is low-level (the lowest weight and the user has little interest in it).

[0090] Specifically, suppose the user is weighting and importance-grading an image containing a windbreaker, trees, and button details; the user assigns a 5-star weight to the main body of the windbreaker (sub-region A); the user assigns a 3-star weight to the trees in the background (sub-region B); and the user assigns a 1-star weight to the button details of the windbreaker (sub-region C).

[0091] The system combines the user's weight tag and the degree of preference for the windbreaker style (assuming the user likes this windbreaker very much); the system determines that sub-area A (the main body of the windbreaker) is a high level; the system determines that sub-area B (the trees in the background) is a medium level, because although the user is not as interested in it as the main body of the windbreaker, it still has a certain degree of attention; the system determines that sub-area C (button details) is a low level because the user is least interested in it; through such steps, the system can divide each sub-area into important levels based on the user's weight tag and preference, providing a basis for subsequent processing and analysis.

[0092] Furthermore, semantic analysis is performed on the multiple sub-regions, and semantic information of the multiple sub-regions is introduced under the semantic analysis. At this time, the semantic content corresponding to the multiple sub-regions is presented through the semantic information of the multiple sub-regions, ensuring the accuracy of the semantic information of the multiple sub-regions.

[0093] At this point, the system will perform semantic analysis on multiple sub-regions, which means using natural language processing (NLP) or computer vision technology to understand and extract key information or labels in the sub-regions; then, the system will introduce the semantic information of these sub-regions and use this information to present the semantic content corresponding to each sub-region.

[0094] The system uses advanced algorithms and technologies to perform semantic analysis on each sub-region; this involves multiple technologies such as image recognition, text analysis, and object detection; the goal is to extract key information in the sub-region, such as object name, attributes, relationships, etc.; optionally, suppose there is an image containing multiple sub-regions, one of which is a red car; the system uses image recognition technology to identify key information in this sub-region, such as "car", "red", etc.

[0095] The system integrates the parsed semantic information into the description of each sub-region; this information includes the name, category, color, shape, size and other attributes of the object; the system also uses external resources such as knowledge graphs and databases to enrich this information; optionally, for the red car sub-region mentioned above, the system integrates its semantic information into: "This is a red car;" the system further uses the knowledge graph to add additional information such as the brand, model, and year of the car.

[0096] The system presents this semantic information to the user in the form of text, labels, icons or other visual forms; the purpose is to help users better understand the content of each sub-area and the relationship between sub-areas; the presentation method varies depending on the system design and user needs; optionally, for the red car sub-area, the system presents its semantic content in the form of a text label: "red car"; in a more complex interface, the system presents it in an image + text format, including an image of the car and a descriptive text below.

[0097] Specifically, suppose there is an image containing three sub-regions: a red car (sub-region A), a person wearing a blue coat (sub-region B), and a green grass (sub-region C); the system recognizes the key information in sub-region A: "car" and "red"; the system recognizes the key information in sub-region B: "person" and "blue coat"; the system recognizes the key information in sub-region C: "grass" and "green".

[0098] By introducing semantic information, for sub-region A, the system integrates the semantic information as: "This is a red car;" for sub-region B, the system integrates the semantic information as: "This is a person wearing a blue coat;" for sub-region C, the system integrates the semantic information as: "This is a green grass;".

[0099] The system presents the semantic content of each sub-region in the form of a text label. On the image, a text label appears next to sub-region A: "red car"; a text label appears next to sub-region B: "person in blue coat"; and a text label appears next to sub-region C: "green grass." Through these steps, the system can extract, integrate, and present the semantic information of each sub-region, helping users better understand the image content.

[0100] Therefore, the content to be retrieved is determined based on the importance levels of multiple sub-areas and the voice contents corresponding to the multiple sub-areas, and multiple sub-target data are determined based on the content to be retrieved and the Internet database. The multiple sub-target data are the Internet data of the target image in different dimensions. At the same time, the content to be retrieved is retrieved in the Internet database based on different dimensions, including regional dimension, time dimension and object type dimension, thereby ensuring the accuracy of multiple sub-target data.

[0101] At this point, the system will determine the content to be retrieved based on the importance levels and corresponding semantic content of the multiple sub-areas previously determined; subsequently, these contents to be retrieved will be used to search for relevant sub-target data in the Internet database; these sub-target data represent the Internet data of the target image in different dimensions, and the retrieval process will cover the regional dimension, time dimension, and object type dimension.

[0102] The system will first refer to the importance levels of multiple sub-areas and give priority to the semantic content corresponding to high-level sub-areas; these semantic contents include the names, attributes, relationships, etc. of objects, which will be integrated into keywords or phrases to be retrieved; optionally, assuming there are three sub-areas: a red car (high level), a person wearing a blue coat (medium level), and a green grass (low level); the system will mainly determine the content to be retrieved based on the semantic content of the red car, such as "red car".

[0103] The system inputs the content to be retrieved into the Internet database and conducts an extensive search; the purpose of the search is to find data related to the target image, which comes from different websites, social media, e-commerce platforms, etc.; the system will collect these related data as sub-target data; optionally, for the content to be retrieved "red car", the system will search the Internet database for related images, descriptions, prices, brands and other information; the collected sub-target data includes high-definition pictures of red cars, detailed product descriptions, quotations from different merchants, and brand information, etc.

[0104] During the retrieval process, the system will consider different dimensions to ensure the comprehensiveness and diversity of the retrieval results; these dimensions include regional dimensions (such as searching for red cars in a specific region), time dimensions (such as searching for the latest or specific year's red cars), and object type dimensions (such as searching for different styles or configurations of red cars); optionally, in the regional dimension, the system searches and presents red car data from around the world or specific regions (such as North America, Europe); in the time dimension, the system searches and displays the latest released red car styles, or styles of a specific year specified by the user; in the object type dimension, the system searches and compares red cars of different brands, models, and configurations to meet the diverse needs of users.

[0105] Specifically, suppose the user is looking for a red car, and they are particularly interested in the North American market, the latest released models, and SUV-type cars; the system determines the content to be retrieved based on the user's previous weight marking and semantic analysis of the sub-region: "red car"; the system searches for data related to "red car" in the Internet database and collects a large amount of sub-target data, including red car information from different brands, models, and configurations; in the regional dimension, the system pays special attention to red car data in the North American market and filters out models that meet user needs; in the time dimension, the system searches for the latest released red car models to ensure that users can see the latest product information; in the object type dimension, the system focuses on SUV-type red cars, providing users with a wealth of choices and comparisons; through such steps, the system can retrieve relevant sub-target data in the Internet database based on user preferences and needs, and provide users with comprehensive, diverse and personalized information support.

[0106] refer to Figure 5 , in step S14, a plurality of data combinations are determined according to the plurality of sub-target data and the target image;

[0107] In the specific implementation process of the present invention, the specific steps are:

[0108] S141: collecting a plurality of sub-target data, matching the plurality of sub-target data with a target image, and determining matching coefficients of the plurality of sub-target data based on matching of the plurality of sub-target data with respect to the target image;

[0109] S142: Divide the matching levels according to the matching coefficients of the multiple sub-target data, and determine the multiple target data of the same matching level as a data combination to collect multiple data combinations.

[0110] In an embodiment of the present application, multiple sub-target data are collected and matched with the target image. The matching coefficients of the multiple sub-target data are determined based on the matching of the multiple sub-target data relative to the target image, which is compatible with the overall consideration of the matching of the multiple sub-target data relative to the target image and ensures the accuracy of the matching coefficients of the multiple sub-target data.

[0111] At this time, multiple sub-target data are collected and matched with the target image to determine the degree of matching between them; based on the matching results, a matching coefficient is calculated for each sub-target data, which reflects the similarity or matching degree between the sub-target data and the target image.

[0112] Collect multiple sub-target data, which are images, text descriptions, metadata or other forms of information; ensure that the collected sub-target data are diverse enough to cover various aspects of the target image; optionally, assume that the target image is a picture depicting a red car; retrieve multiple pictures related to red cars from an Internet database, which show red cars of different brands, models, and angles; also obtain text descriptions related to these pictures, such as model information, price, user reviews, etc.

[0113] Each collected sub-target data is matched with the target image; the matching process involves the extraction of image features (such as color, shape, texture, etc.), the calculation of similarity (such as cosine similarity, Euclidean distance, etc.), and the formulation of matching strategies (such as best match, top N matches, etc.); the matching result will be a numerical value or score representing the similarity or matching degree between the sub-target data and the target image; optionally, for each red car picture retrieved from the Internet database, an image matching algorithm (such as SIFT, SURF or a deep learning model) is used to extract its features and compare them with the features of the target image; by calculating the similarity between the features (such as using cosine similarity), a numerical value representing the matching degree of each picture with the target image is obtained.

[0114] Based on the matching results, a matching coefficient will be determined for each sub-target data; this coefficient is a value between 0 and 1, where 0 indicates a complete mismatch and 1 indicates a complete match; the determination of the matching coefficient is based on methods such as direct conversion of similarity, threshold division, and normalization processing; optionally, assuming that the calculated similarity value range is 0 to 1, this value is directly used as the matching coefficient; or, a threshold is set (such as 0.5), and sub-target data with a similarity greater than the threshold is marked as a high match (matching coefficient is 1), and sub-target data with a similarity less than or equal to the threshold is marked as a low match (matching coefficient is 0 or other smaller values); in addition, in order to make the matching coefficients between different sub-target data comparable, the similarity is also normalized so that the values of all matching coefficients are between 0 and 1.

[0115] Specifically, assuming the target image is a picture depicting a red car, three related red car pictures (Picture A, Picture B, and Picture C) and their text descriptions are retrieved from the Internet database; Picture A: a red SUV, different from the target image model but the same color; Picture B: a red car, the same model and color as the target image; Picture C: a red sports car, different from the target image model but the same color; Text description: model information, price, etc. related to each picture.

[0116] Use the image matching algorithm to calculate the similarity between each image and the target image; assume that the similarities obtained are: Image A (0.6), Image B (0.9), Image C (0.5); directly use the similarity as the matching coefficient to obtain: Image A (matching coefficient 0.6), Image B (matching coefficient 0.9), Image C (matching coefficient 0.5); or, set a threshold (such as 0.7), mark Image A and Image C as low matches (matching coefficient is 0 or other smaller values), and mark Image B as high match (matching coefficient is 1); through such steps and examples, we can clearly see how to collect multiple sub-target data and match them with the target image, as well as how to determine the matching coefficient of each sub-target data; this will provide strong support for subsequent analysis and processing.

[0117] Furthermore, matching levels are divided according to matching coefficients of the plurality of sub-target data, and the plurality of target data of the same matching level are determined as a data combination, so as to collect the plurality of data combinations;

[0118] At this point, multiple sub-target data are divided into matching levels based on their matching coefficients calculated previously; then, multiple sub-target data of the same matching level are combined together to form multiple data combinations; this helps to better understand and analyze these data, and provides a basis for subsequent processing and decision-making.

[0119] According to the value of the matching coefficient, the sub-target data are divided into different matching levels; the matching level is determined based on the range or threshold of the matching coefficient; for example, the sub-target data with a matching coefficient above 0.8 is divided into a high matching level, the sub-target data with a matching coefficient between 0.5 and 0.8 is divided into a medium matching level, and the sub-target data with a matching coefficient below 0.5 is divided into a low matching level; the purpose of dividing the matching level is to classify the sub-target data with high similarity into one category for subsequent analysis and processing; optionally, assuming there are 10 sub-target data, their matching The coefficients are 0.9, 0.85, 0.78, 0.65, 0.52, 0.48, 0.35, 0.29, 0.22 and 0.15 respectively; according to the set threshold (such as 0.8 is a high match, 0.5 to 0.8 is a medium match, and below 0.5 is a low match), these sub-target data are divided into three matching levels: high matching level (including two sub-target data with matching coefficients of 0.9 and 0.85), medium matching level (including two sub-target data with matching coefficients of 0.78 and 0.65) and low matching level (including the remaining six sub-target data).

[0120] Combine multiple sub-target data of the same matching level to form a data combination; the data combination contains the same or similar sub-target data, which have similar values in the matching coefficient; by forming a data combination, the distribution of sub-target data at different matching levels can be seen more intuitively, and it facilitates subsequent analysis and processing; optionally, for the three matching levels obtained in the previous step, combine the sub-target data in them into three data combinations: data combination 1 (high matching level, containing two sub-target data), data combination 2 (middle matching level, containing two sub-target data) and data combination 3 (low matching level, containing six sub-target data).

[0121] The formed data combinations are collected and organized; the collected data combinations are stored in a database, file or other data storage medium for subsequent analysis and processing; by collecting multiple data combinations, a comprehensive data set is constructed, which contains multiple sub-target data at different matching levels, providing rich information for subsequent decision-making and reasoning; optionally, the three data combinations obtained in the previous step are stored in a database, with each data combination as an independent record or table item; the database contains information such as the identification of the data combination, the matching level, the sub-target data included, etc.; in this way, it is convenient to query, analyze and process these data combinations.

[0122] Specifically, assume that a target image is a picture depicting a red car, and 10 pictures related to red cars (Picture 1 to Picture 10) and their matching coefficients (0.9, 0.85, 0.78, 0.65, 0.52, 0.48, 0.35, 0.29, 0.22 and 0.15, respectively) have been retrieved from the Internet database.

[0123] According to the set threshold (0.8 is a high match, 0.5 to 0.8 is a medium match, and below 0.5 is a low match), the 10 images are divided into three matching levels: high matching level (images 1 and 2), medium matching level (images 3 and 4), and low matching level (images 5 to 10); for each matching level, the images are combined into a data combination: data combination 1 (high matching level, including images 1 and 2), data combination 2 (medium matching level, including images 3 and 4), and data combination 3 (low matching level, including images 5 to 10).

[0124] These three data combinations are stored in a database, with each data combination as an independent record or table item; the database contains the identifiers of the data combinations (such as "data combination 1", "data combination 2" and "data combination 3"), matching levels (such as "high match", "medium match" and "low match"), and included image information (such as the image file name, URL or storage path), etc.; through such steps and examples, we can clearly see how to divide the matching levels according to the matching coefficients of multiple sub-target data, and combine multiple sub-target data of the same matching level to form multiple data combinations; this will provide strong support for subsequent analysis and processing.

[0125] refer to Figure 6 In step S15, if the combined matching coefficients of the multiple data combinations are all lower than the preset matching coefficient threshold, the supplementary data is determined based on the combined matching coefficients of the multiple data combinations and the source information of the target image, and the target data is determined based on the supplementary data, the multiple data combinations, and the user's data preference factors;

[0126] In the specific implementation process of the present invention, the specific steps are:

[0127] S151: Matching multiple data combinations, determining combined matching coefficients of the multiple data combinations based on the matching of the multiple data combinations, determining a preset matching coefficient threshold based on the target image and a preset image matching table, and comparing the combined matching coefficients of the multiple data combinations with the preset matching coefficient threshold;

[0128] S152: If the combined matching coefficients of the multiple data combinations are all lower than the preset matching coefficient threshold, source information of the target image is collected, and multiple traceability data are determined based on the tracing of the source information of the target image. The retrieval of supplementary data is triggered based on the combined content of the multiple traceability data and the multiple data combinations to determine the supplementary data;

[0129] S153: Collect the user's data preference factors, and determine the preference data combination based on the supplementary data and the user's data preference factors, and determine the target data by combining the preference data combination and multiple data combinations.

[0130] In an embodiment of the present application, multiple data combinations are matched, and the combined matching coefficients of the multiple data combinations are determined based on the matching of the multiple data combinations. A preset matching coefficient threshold is determined based on the target image and a preset image matching table. The combined matching coefficients of the multiple data combinations are compared with the preset matching coefficient threshold, which is compatible with the overall consideration of the target image and the preset image matching table, and ensures the accuracy of the preset matching coefficient threshold.

[0131] At this time, multiple data combinations are matched and the combined matching coefficients between them are determined; then, a preset matching coefficient threshold is determined based on the target image and the preset image matching table; finally, the combined matching coefficients of the multiple data combinations are compared with this preset threshold to determine whether they meet the matching requirements.

[0132] Match multiple data combinations with each other; the matching process involves comparing sub-target data, features, labels or other relevant information in the data combinations; the purpose of matching is to find the similarity or correlation between the data combinations, and provide a basis for the subsequent calculation of the combination matching coefficient; optionally, assume that there are three data combinations: combination A (containing sub-target data 1 and sub-target data 2), combination B (containing sub-target data 3 and sub-target data 4) and combination C (containing sub-target data 5); find the similarity or correlation between them by comparing the sub-target data in these combinations; for example, compare the color, shape, texture and other features of the sub-target data, or compare their labels, descriptions and other information.

[0133] The combination matching coefficient is calculated based on the matching results of the data combination; the combination matching coefficient is a value between 0 and 1, which is used to represent the similarity or correlation between the data combinations; the method for calculating the combination matching coefficient includes calculating the cosine similarity, Euclidean distance, Manhattan distance, etc. of the feature vector, or scoring based on a certain rule or algorithm; optionally, it is assumed that cosine similarity is used to calculate the combination matching coefficient; for combination A and combination B, the cosine similarity of each sub-target data between them is calculated respectively, and then the average value is taken as the combination matching coefficient of combination A and combination B; assuming that the combination matching coefficient of combination A and combination B is 0.75, it means that there is a high similarity or correlation between them.

[0134] A preset matching coefficient threshold is determined based on the target image and a preset image matching table; this threshold is used to judge whether the degree of matching between the data combination and the target image is high enough; the preset image matching table contains information such as the characteristics, labels, and historical matching data of the target image, which helps determine a reasonable matching coefficient threshold; optionally, assuming that the target image is a picture depicting a red car, and the preset image matching table contains multiple pictures similar to the target image and their corresponding matching coefficients; a reasonable preset matching coefficient threshold is determined by analyzing the matching coefficients in the preset image matching table; for example, setting the threshold to 0.7 means that the data combination is considered to meet the matching requirements only when the matching coefficient between the data combination and the target image is greater than or equal to 0.7.

[0135] The combined matching coefficients of multiple data combinations are compared with the preset matching coefficient threshold; the purpose of the comparison is to determine whether the data combination meets the matching requirements; if the combined matching coefficient of the data combination is greater than or equal to the preset matching coefficient threshold, the data combination is considered to meet the matching requirements; otherwise, the data combination is considered to not meet the matching requirements; optionally, assume that there are three data combinations: combination A, combination B and combination C, and their combined matching coefficients are 0.75, 0.65 and 0.55 respectively; compare these three combined matching coefficients with the preset matching coefficient threshold of 0.7; the results show that the combined matching coefficient of combination A is greater than the threshold (0.75>0.7), so it meets the matching requirements; while the combined matching coefficients of combination B and combination C are less than the threshold (0.65<0.7, 0.55<0.7), so they do not meet the matching requirements.

[0136] Specifically, suppose there is a target image that is a picture of a red car, and three data combinations have been retrieved from the database: combination A (containing two pictures of red cars), combination B (containing a picture of a red car and a blue SUV), and combination C (containing two pictures of scenery); by comparing the sub-target data in these three data combinations, it is found that both pictures in combination A are similar to the target image, while only one picture in combination B is similar to the target image, and the picture in combination C is unrelated to the target image.

[0137] Cosine similarity was used to calculate the combination matching coefficient, and the combination matching coefficient of combination A was 0.75, the combination matching coefficient of combination B was 0.65, and the combination matching coefficient of combination C was 0.55; the matching coefficients in the preset image matching table were analyzed, and the preset matching coefficient threshold was determined to be 0.7; the combination matching coefficients of combination A, combination B, and combination C were compared with the preset matching coefficient threshold of 0.7, and it was found that only combination A met the matching requirements (0.75>0.7), while combination B and combination C did not meet the matching requirements (0.65<0.7, 0.55<0.7); combination A was determined to be the data combination that met the matching requirements, while combination B and combination C needed further processing or analysis.

[0138] Furthermore, if the combined matching coefficients of multiple data combinations are all lower than the preset matching coefficient threshold, the source information of the target image is collected, and multiple traceability data are determined based on the tracing of the source information of the target image. The retrieval of supplementary data is triggered based on the combined content of the multiple traceability data and the multiple data combinations to determine the supplementary data.

[0139] At this time, when the combined matching coefficients of multiple data combinations are all lower than the preset matching coefficient threshold, a series of measures are taken to retrieve and determine the supplementary data; these measures include collecting the source information of the target image, tracing and determining multiple traceability data, and triggering the retrieval of supplementary data based on the combination of traceability data and existing data.

[0140] When the combined matching coefficients for all data combinations fall below a preset threshold, a deeper understanding of the target image's background is necessary. This typically involves collecting source information for the target image, including the image's shooting location, time, equipment, photographer, and copyright information. Collecting this source information involves communicating with the image database administrator, photographer, copyright holder, or image sharing platform. Alternatively, assuming the target image is a landscape photo shared on social media, source information, such as the shooting location, time, and uploader, can be obtained through the social media platform's API or by contacting the platform administrator.

[0141] After obtaining the source information of the target image, start tracing other data related to the image; this data includes other images taken at the same place, time or situation as the image, image description information, or other images or information related to the image uploader; determining the traceability data involves querying the image database, searching the social media platform, or further communication with the photographer or copyright owner; optionally, assuming that the shooting location and time of the target image are obtained through the social media platform, search for other images taken within the same place and time range on the platform, or contact the uploader to inquire whether there is more relevant information.

[0142] Once the traceability data is available, it is used to trigger the retrieval of supplementary data; this involves searching for images similar to the traceability data in a larger database or image collection, or using the descriptive information or tags in the traceability data to search for related images; the retrieval of supplementary data involves complex image matching algorithms, tag search, or content-based image retrieval technology; optionally, assuming that the traceability data shows that the target image was taken at a famous tourist attraction, other images related to the attraction are searched in a database containing a large number of tourist attraction images; at the same time, the descriptive information in the target image and the traceability data (such as the attraction name, weather conditions, etc.) is also used to further refine the search conditions.

[0143] Specifically, suppose you are an image analyst at a travel company and your task is to select high-quality images for a set of travel brochures. You already have an image database containing multiple data combinations, but unfortunately, none of these data combinations match the target brochure theme (such as the scenery of a specific tourist attraction) very well.

[0144] We noticed that the target image was a photo of the scenic spot shared on social media. We obtained the image's shooting location, time, and uploader information by communicating with the social media platform. Using the shooting location and time information, we searched for multiple images of the same scenic spot taken during the same time period on the social media platform. We also contacted the uploader, who shared some additional information and background stories about the shooting of the scenic spot.

[0145] These traceability data (including additional information and background stories) were input into a database containing a large number of tourist attraction images, and content-based image retrieval was performed; descriptive information in the traceability data (such as attraction name, weather conditions, season, etc.) was also used to further refine the search criteria; after retrieving a large amount of supplementary data, it was manually reviewed and compared with the existing data combination; finally, several images that were highly relevant to the target brochure theme and of high quality were selected as supplementary data.

[0146] Therefore, the user's data preference factors are collected, and the preference data combination is determined based on the supplementary data and the user's data preference factors, and the target data is determined by combining the preference data combination and multiple data combinations, which is compatible with the overall consideration of the supplementary data and the user's data preference factors, and ensures the accuracy of the preference data combination.

[0147] At this point, a preference data combination is determined based on the user's data preference factors and combined with the previously retrieved supplementary data; then, this preference data combination is combined with the original multiple data combinations to jointly determine the final target data; this step is intended to ensure that the selected data not only meets the technical matching standards, but also meets the user's personalized needs.

[0148] In this step, it is necessary to comprehensively collect and analyze user preferences for data; this includes user preferences for specific colors, styles, themes or content, as well as their preferences for data presentation (for example, prefer images or text descriptions); methods for collecting user preference factors include questionnaires, user behavior analysis, personalized recommendation system feedback, etc.; optionally, assume that images are being selected for the product display page of an e-commerce platform; in order to understand user preferences, an online questionnaire is used to collect user preferences for product image styles (such as real photos, hand-drawn illustrations, cartoon images, etc.), and whether they prefer to view images with detailed text descriptions or concise and clear images.

[0149] After obtaining the user's preference factors, this information is compared and analyzed with the supplementary data to find the data that best matches the user's preferences; this involves classifying and labeling the data, or using machine learning algorithms to make personalized recommendations; the goal is to screen out data that meets both technical matching requirements (such as a high correlation with the target image) and the user's personalized needs, to form a preference data combination; optionally, assuming that in the example of an e-commerce platform, through analysis of user questionnaires, it is learned that most users prefer product images in a real photo style and tend to view images with short descriptions; therefore, images that meet these conditions are screened out from the supplementary data to form a preference data combination.

[0150] In this step, the preference data combination is combined with the original multiple data combinations, and the final target data is determined by comprehensively considering the data matching degree, user preferences, and business goals (such as improving click-through rate, conversion rate, etc.); this process involves weighting and sorting the data, or using a certain decision algorithm to select the optimal data combination; optionally, in the example of an e-commerce platform, the preference data combination (product images in a real photo style) is combined with the original multiple data combinations (including images in different styles such as hand-drawn illustrations and cartoon images); then, according to factors such as user preferences, image quality, and relevance to the product, the images are sorted and screened, and finally the target data for the product display page is determined.

[0151] Specifically, suppose you are selecting images to showcase specific tourist destinations for an online travel booking platform. You already have an image library containing multiple data combinations, but you want to further optimize the image selection based on user preferences. Through online questionnaires and user behavior analysis, you find that users prefer a combination of natural scenery and cultural landscapes for tourist destination images, and they also prefer images with bright colors and beautiful composition.

[0152] Images that meet user preferences are screened out from the supplementary data, namely those that show a combination of natural scenery and cultural landscapes, with bright colors and beautiful composition; these images form a preference data combination; the preference data combination is combined with the original multiple data combinations, and the image matching degree (i.e., relevance to a specific tourist destination), user preferences (such as color, composition, etc.), and business goals (such as increasing user click-through rate and booking conversion rate) are comprehensively considered; through the comprehensive application of these factors, the target data for displaying specific tourist destinations is finally determined. These images not only have a high degree of technical matching, but can also effectively attract users' attention and stimulate their interest in travel.

[0153] See also Figure 7 , Figure 7: is a schematic diagram of the structure of a target data processing system according to an embodiment of the present invention; the target data processing system includes:

[0154] A target image module 21 is configured to determine a target image based on a user's web browsing history and a user's search criteria;

[0155] A sub-region module 22 is configured to determine a plurality of sub-regions according to a target image;

[0156] A sub-target data module 23 is configured to determine a plurality of sub-target data based on the importance levels of the plurality of sub-regions, the semantic information of the plurality of sub-regions, and an Internet database, wherein the plurality of sub-target data are Internet data of the target image at different dimensions;

[0157] A data combination module 24 is configured to determine a plurality of data combinations based on the plurality of sub-target data and the target image;

[0158] The target data module 25 is used to determine the supplementary data based on the combined matching coefficients of the multiple data combinations and the source information of the target image if the combined matching coefficients of the multiple data combinations are all lower than the preset matching coefficient threshold, and determine the target data based on the supplementary data, the multiple data combinations and the user's data preference factors.

[0159] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for processing target data, characterized in that: include: Determine the target image based on the user's web browsing history and the user's search key points; determining a plurality of sub-regions according to the target image; determining a plurality of sub-target data based on the importance levels of the plurality of sub-regions, the semantic information of the plurality of sub-regions, and an Internet database, wherein the plurality of sub-target data are Internet data of the target image at different dimensions; determining a plurality of data combinations according to the plurality of sub-target data and the target image; If the combined matching coefficients of multiple data combinations are all lower than the preset matching coefficient threshold, the supplementary data is determined based on the combined matching coefficients of the multiple data combinations and the source information of the target image, and the target data is determined based on the supplementary data, multiple data combinations and the user's data preference factors.

2. The target data processing method according to claim 1, characterized in that: The determining of the target image based on the user's web browsing history and the user's search key points includes: Collecting the user's search key points, determining a search combination based on the search key points, triggering a corresponding network search based on the search combination, and generating multiple first pages; Obtaining a user's web browsing history, and determining a plurality of second pages based on the user's web browsing history and the user's browsing marks, and determining a comprehensive page based on the plurality of second pages and the plurality of first pages; The user selects a target on the comprehensive page and determines the corresponding target image.

3. The target data processing method according to claim 1, characterized in that: The determining of the plurality of sub-regions according to the target image includes: Collecting target information of the user, determining multiple keywords based on keyword screening of the target information, and determining reference weights of the multiple keywords based on the multiple keywords and a reference weight relationship table; A target image is captured, multiple semantic parts are marked according to the analysis of the target image, multiple semantic combinations are determined according to reference weights of the multiple semantic parts and multiple keywords, and the target image is divided according to the multiple voice combinations to determine multiple sub-areas.

4. The method for processing target data according to claim 1, wherein: The plurality of sub-target data are determined based on the importance levels of the plurality of sub-regions, the semantic information of the plurality of sub-regions, and an Internet database. The plurality of sub-target data are Internet data of the target image at different dimensions, including: The user weights the multiple sub-regions to determine the weights of the multiple sub-regions, and determines the importance levels of the multiple sub-regions based on the weights of the multiple sub-regions and the user's preference; Semantic parsing is performed on the multiple sub-regions, and semantic information of the multiple sub-regions is introduced during the semantic parsing. At this time, semantic contents corresponding to the multiple sub-regions are presented through the semantic information of the multiple sub-regions.

5. The method for processing target data according to claim 4, characterized in that: Based on the importance levels of the multiple sub-regions, the semantic information of the multiple sub-regions, and the Internet database, multiple sub-target data are determined. The multiple sub-target data are Internet data of the target image at different dimensions, and also include: The content to be retrieved is determined based on the importance levels of the multiple sub-areas and the voice contents corresponding to the multiple sub-areas. The multiple sub-target data are determined based on the content to be retrieved and an Internet database. The multiple sub-target data are Internet data of the target image in different dimensions. At the same time, the content to be retrieved is retrieved in the Internet database based on different dimensions, including regional dimension, time dimension and object type dimension.

6. The method for processing target data according to claim 1, wherein: The determining of multiple data combinations according to the multiple sub-target data and the target image includes: A plurality of sub-target data are collected, and the plurality of sub-target data are matched with a target image, and matching coefficients of the plurality of sub-target data are determined based on matching of the plurality of sub-target data with respect to the target image.

7. The method for processing target data according to claim 6, wherein: The determining of the plurality of data combinations according to the plurality of sub-target data and the target image further includes: The matching levels are divided according to the matching coefficients of the multiple sub-target data, and the multiple target data of the same matching level are determined as a data combination to collect multiple data combinations.

8. The method for processing target data according to claim 1, wherein: If the combined matching coefficients of the multiple data combinations are all lower than the preset matching coefficient threshold, then determining the supplementary data based on the combined matching coefficients of the multiple data combinations and the source information of the target image, and determining the target data based on the supplementary data, the multiple data combinations, and the user's data preference factors, including: Match multiple data combinations, determine the combined matching coefficients of the multiple data combinations based on the matching of the multiple data combinations, determine a preset matching coefficient threshold based on the target image and a preset image matching table, and compare the combined matching coefficients of the multiple data combinations with the preset matching coefficient threshold.

9. The target data processing method according to claim 8, characterized in that: If the combined matching coefficients of the plurality of data combinations are all lower than a preset matching coefficient threshold, determining supplementary data based on the combined matching coefficients of the plurality of data combinations and source information of the target image, and determining target data based on the supplementary data, the plurality of data combinations, and the user's data preference factors, further comprising: If the combined matching coefficients of the multiple data combinations are all lower than the preset matching coefficient threshold, the source information of the target image is collected, and multiple traceability data are determined based on the tracing of the source information of the target image. The retrieval of supplementary data is triggered based on the combined content of the multiple traceability data and the multiple data combinations to determine the supplementary data; Collect the user's data preference factors, determine the preference data combination based on the supplementary data and the user's data preference factors, and determine the target data by combining the preference data combination and multiple data combinations.

10. A target data processing system, characterized in that: The target data processing system is applied to the target data processing method according to any one of claims 1 to 9, and the target data processing system includes: A target image module is used to determine a target image based on the user's web browsing history and the user's search key points; A sub-region module, used for determining a plurality of sub-regions according to a target image; A sub-target data module is used to determine a plurality of sub-target data based on the importance levels of the plurality of sub-regions, the semantic information of the plurality of sub-regions, and an Internet database, wherein the plurality of sub-target data are Internet data of the target image at different dimensions; A data combination module, configured to determine a plurality of data combinations according to a plurality of sub-target data and a target image; The target data module is used to determine the supplementary data based on the combined matching coefficients of the multiple data combinations and the source information of the target image if the combined matching coefficients of the multiple data combinations are all lower than the preset matching coefficient threshold, and to determine the target data based on the supplementary data, the multiple data combinations and the user's data preference factors.

Images