A graphical display method for search engine retrieval results

Abstract

The present invention discloses a graphical display method for search engine retrieval results, which belongs to the technical field of search engines, and specifically includes: constructing a graphical knowledge graph through a graphical database, extracting advanced features by using deep learning pre-trained images, and users can initiate search requests with keywords or image samples through search engines. The system uses NLP and deep learning technologies to perform semantic analysis and image recognition on query results to achieve real-time matching with the graphical knowledge graph. The search results are displayed through a carousel and multi-dimensional information viewing is provided. The system also has an intelligent color screening function and performs graphical display in combination with shape and texture features. The system collects user feedback, optimizes the search result sorting algorithm through machine learning, and builds a graphical community, thereby achieving resource sharing, commenting and intelligent recommendation, and improving the accuracy of search.

Description

Technical Field

[0001] The invention belongs to the technical field of search engines, and in particular is a method for graphically displaying search results of search engines. Background Art

[0002] With the rapid growth of information, users' demand for search engines is not only to find relevant information, but also to understand and filter results quickly and intuitively. Graphical display can provide more intuitive and easy-to-understand search results, meeting users' needs for efficient information acquisition. Graphical display technologies, such as clustering, hyperlinking, and semantic content, make it possible to visualize search engine retrieval results. These methods display search results to users in a graphical way through different logical structures and forms of expression, improving the efficiency of users in obtaining information. The development of deep learning and natural language processing technologies enables search engines to understand user query intent more accurately, and to match and sort search results more accurately.

[0003] For example, the patent with publication number CN111339450A discloses a graphical display method of search engine search results, including: storing each piece of information in the search results in a server, and then searching for information to be displayed from each piece of information, and also including the following steps: step 1, selecting the common attributes of each piece of information to be displayed by standardizing and improving each attribute of the information to be displayed; step 2, selecting the coordinate attribute from the common attributes; step 3, displaying the graphics on the browser with the coordinate attribute as the coordinate axis. Through this technical solution, users can intuitively see the search results sorted by multiple attributes, users can quickly understand the attributes of each search result, quickly find the required search results, and in a visual interface, more search results can be displayed without turning pages or pulling down a pull-down bar or scroll bar. Thousands of search results can be displayed in a visual interface without affecting the resolution.

[0004] The above existing technologies have the following problems: 1) lack of advanced feature extraction; 2) only mention that users can view search results through a graphical interface, which limits the interaction between users and search results; 3) lack of real-time matching and intelligent recommendation. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention proposes a graphical display method for search engine retrieval results, which constructs a graphical knowledge graph through a graphical database, and uses deep learning to pre-train images to extract advanced features. Users can initiate search requests with keywords or image samples through search engines. The system uses NLP and deep learning technologies to perform semantic analysis and image recognition on query results, and realizes real-time matching with the graphical knowledge graph. The search results are displayed through a carousel and multi-dimensional information viewing is provided. The system also has an intelligent color screening function, which combines shape and texture features for graphical display. The system collects user feedback, optimizes the search result sorting algorithm through machine learning, and builds a graphical community, which realizes resource sharing, commenting and intelligent recommendation, and improves the accuracy of search.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for graphically displaying search engine retrieval results, comprising:

[0008] Step S1: Establish a graph database, build a graph knowledge graph based on the graph database, organize the images, tags, and metadata information in the graph database into a graph structure, and pre-train the images in the graph database through deep learning methods to extract high-level features of the images and build an intelligent index. Users can enter query keywords or upload image samples through the search engine to initiate a search request;

[0009] Step S2: The search engine receives the user request, uses the NLP method to perform semantic analysis on the query keywords, uses deep learning to identify the uploaded image, and matches it with the image in the graphic knowledge graph in real time, and determines the order of the search results based on the matching results and the sorting algorithm;

[0010] Step S3: Display the results using a carousel according to the search result type. While displaying the search results, the multi-dimensional information of the image is displayed, and the user is allowed to view and switch different dimensional information through interactive operations; the multi-dimensional information of the image includes: the source of the image, the number of clicks, and the number of comments; the interactive operations include: viewing the image details, saving the image, and sharing the image by clicking and sliding;

[0011] Step S4: providing an intelligent color selector on the search interface, allowing the user to select or input a color for screening on the search interface, the search engine screening images according to the color feature vector, combining the shape and texture features of the graphics, and performing graphical display;

[0012] Step S5: Collect user feedback and suggestions, use machine learning algorithms to perform sentiment analysis on the feedback data, and iterate and optimize the graphical knowledge graph and search result sorting algorithm based on user feedback;

[0013] Step S6: Build a graphic community to allow users to share, comment and like graphic resources, introduce social graph analysis methods to obtain the correlation between users, and make intelligent recommendations based on the correlation.

[0014] Specifically, the specific steps of step S2 include:

[0015] S2.1: The search engine receives the query keywords entered by the user or the uploaded image samples;

[0016] S2.2: For query keywords, use the NLP method to encode the query keywords, generate word vectors, and use the improved semantic similarity algorithm to calculate the similarity between the word vectors and the labels or metadata in the graph knowledge graph. The formula is:

[0017]

[0018] in, represents the similarity between word vector A and word vector B, w represents the weight vector with the same dimension as A and B, represents the sparsity factor, represents the sensitivity adjustment parameter, represents the 2-norm.

[0019] Specifically, the specific steps of step S2 also include:

[0020] S2.3: For the uploaded image samples, use the deep learning model to extract features, generate the feature vector of the image, and use the graph search strategy to match the extracted image feature vector with the pre-trained image features in the graph knowledge graph to find similar images;

[0021] S2.4: According to the relevance of the matching results, the search results are sorted using a weighted sorting algorithm, and the sorted search results are returned to the user.

[0022] Specifically, the specific steps of the graph search strategy in S2.3 include:

[0023] S2.31: receiving image samples uploaded by users and performing preprocessing;

[0024] S2.32: Use the pre-trained ResNet deep learning model to extract features from the pre-processed image, and perform pooling operations on the extracted feature map to convert it into a feature vector of length h;

[0025] S2.33: Load the pre-trained image feature vector from the graph knowledge graph, and calculate the similarity between the feature vector C of the uploaded image and each pre-trained feature vector D in the graph knowledge graph according to the similarity algorithm calculation formula in S2.2 .

[0026] Specifically, the specific steps of the graph search strategy in S2.3 also include:

[0027] S2.34: Define graph search strategy, based on Search the graph knowledge graph for the image most similar to the uploaded image;

[0028] S2.35: Set the similarity threshold to H, according to Sort search results;

[0029] like , then discard;

[0030] like , then the corresponding image is output as the query result.

[0031] Specifically, the formula of the graph search strategy in S2.34 is:

[0032]

[0033] Among them, r represents the similarity coefficient between the uploaded image and the image in the graphic knowledge graph, d represents the color channel of the uploaded image, Indicates different color channels d and different spatial positions in the uploaded image The pixel weight parameter, represents the nonlinear transformation function of color channel d, Represents the template image in the graphic knowledge graph in color channel d and position The pixel value of represents the pixel mean of the template image in the color channel d in the graphic knowledge graph, Indicates the uploaded image in color channel d and position The pixel value of Represents the pixel mean of the uploaded image in color channel d, Represents the offset parameter of the template image in the color channel d in the graphic knowledge graph, Indicates the offset parameter of the uploaded image in color channel d.

[0034] Specifically, an electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of a method for graphically displaying search engine retrieval results when executing the computer program.

[0035] Specifically, a computer-readable storage medium stores computer instructions, which, when executed, execute the steps of a method for graphically displaying search engine retrieval results.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] 1. The present invention proposes a graphical display method for search engine retrieval results. By organizing images, labels and metadata through a graph structure, the search engine can quickly locate relevant image resources, and the combination of NLP and deep learning makes the search of keywords and image samples more intelligent, thereby improving the accuracy of the search. The multi-dimensional information display method of a carousel image can intuitively understand the detailed information of the search results, such as the labels and metadata of the image. At the same time, the intelligent color selector allows users to filter according to color features, and combined with the shape and texture features of the graphics, the search results are more in line with the user's personalized needs.

[0038] 2. The present invention proposes a graphical display method for search engine retrieval results. By collecting user feedback and suggestions and using machine learning algorithms to perform sentiment analysis, the search result sorting algorithm and graphic knowledge graph are optimized. The construction of a graphic community allows users to share, comment and like graphic resources, thereby promoting communication and interaction between users. At the same time, by introducing a social graph analysis method, the system can obtain the correlation between users and make intelligent recommendations based on the correlation. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A schematic diagram of a method for graphically displaying search engine retrieval results according to the present invention;

[0040] Figure 2 A flowchart of a method for graphically displaying search engine retrieval results according to the present invention;

[0041] Figure 3 A flowchart of a method for graphically displaying search results of a search engine according to the present invention;

[0042] Figure 4 The present invention is a flowchart of a method for graphically displaying search engine retrieval results and a search strategy implementation diagram. DETAILED DESCRIPTION

[0043] Example 1

[0044] See also Figure 1-Figure 2 The present invention provides an embodiment: a method for graphically displaying search engine retrieval results, comprising the following steps:

[0045] Step S1: Establish a graph database, construct a graph knowledge graph based on the graph database, organize the images, tags, and metadata information in the graph database into a graph structure, and pre-train the images in the graph database through a deep learning method, extract high-level features of the images, and build an intelligent index. The user enters a query keyword or uploads an image sample through a search engine to initiate a search request. Among them, pre-training images using a deep learning method is a prior art content in this field, and is not an inventive solution of the present application, and will not be elaborated here;

[0046] The specific steps to build a graph database include:

[0047] (1) Use image processing and computer vision technology, such as the SURF algorithm, to extract key features of the image, such as color, texture, and shape, from the Internet or local resources. The SURF algorithm is a prior art in the art and is not an inventive solution of the present application, and is not described in detail here;

[0048] (2) These feature information and image metadata, such as title, description, and tags, are stored in a graph database to form a structured image data warehouse, where the database uses Neo4j.

[0049] The specific steps to build a graphical knowledge graph include:

[0050] (1) Determine the goals and scope of the knowledge graph;

[0051] (2) First, design the ontology construction layer, including defining entities, relationships, attributes, and their types, and then extract knowledge from structured data sources based on these designs;

[0052] (3) Use rule-based information extraction methods to identify named entities from text and extract the associations between entities from the text to form a network knowledge structure;

[0053] (4) Perform knowledge fusion on the extracted information and use a graph database to store the knowledge graph in the form of a graph;

[0054] (5) Regularly update and expand the knowledge graph as new data sources and knowledge are added.

[0055] Step S2: The search engine receives the user request, uses the NLP method to perform semantic analysis on the query keywords, uses deep learning to identify the uploaded image, and matches it with the image in the graphic knowledge graph in real time, and determines the order of the search results based on the matching results and the sorting algorithm;

[0056] Step S3: Display the results using a carousel according to the search result type. While displaying the search results, the multi-dimensional information of the image is displayed, and the user is allowed to view and switch different dimensional information through interactive operations. The multi-dimensional information includes: source, click volume, number of comments, and interactive operations include: click, slide to view image details, save the image, and share the image;

[0057] Step S4: providing an intelligent color selector on the search interface, allowing the user to select or input a color for screening on the search interface, the search engine screening images according to the color feature vector, combining the shape and texture features of the graphics, and performing graphical display;

[0058] Step S5: Collect user feedback and suggestions, use machine learning algorithms to perform sentiment analysis on the feedback data, and iterate and optimize the graphic knowledge graph and search result sorting algorithm based on user feedback. In using the machine learning algorithm to perform sentiment analysis on the feedback data, the machine learning algorithm uses a naive Bayes algorithm, which is a prior art in this field and is not an inventive solution of the present application, and will not be described in detail here;

[0059] Furthermore, the specific steps of iterating and optimizing the graphical knowledge graph and search result ranking algorithm include:

[0060] (1) Count and analyze the results of sentiment analysis to understand users’ satisfaction and opinions on the graphical knowledge graph and search results;

[0061] (2) Determine the optimization direction of the graphical knowledge graph and search result ranking algorithm based on user feedback;

[0062] (3) Use the entity relationship extraction method of the knowledge graph to iterate and optimize the construction algorithm of the graphical knowledge graph and the search result ranking algorithm to improve the accuracy and completeness of the knowledge graph. The entity relationship extraction method of the knowledge graph is the prior art content in this field and is not the inventive solution of this application, so it will not be described in detail here;

[0063] (4) Using a feature selection algorithm to optimize the search result sorting algorithm to improve the accuracy of the search results. The feature selection algorithm is a prior art in this field and is not an inventive solution of the present application, and is not described in detail here.

[0064] (5) Use user feedback as an evaluation metric to test and evaluate the optimized graph knowledge graph and search result ranking algorithm;

[0065] (6) Continuously iterate and optimize the graphical knowledge graph and search result sorting algorithm based on user feedback and test results.

[0066] Step S6: Build a graphic community to allow users to share, comment and like graphic resources, introduce social graph analysis methods to obtain the correlation between users, and make intelligent recommendations based on the correlation.

[0067] Furthermore, the specific steps to build a graphical community include:

[0068] (1) Determine the main purpose of the community, such as sharing graphic design resources, exchanging knowledge graph construction experience, and defining core functions; (2) Design the community structure and interface; (3) Develop the community platform; (4) Collect user behavior data, including user browsing history, likes, and comments; (5) Build a social network between users based on user behavior data; (6) Use a graph database to store user social network data and design query and analysis methods; (7) According to the characteristics of the community and user needs, integrate the results of social graph analysis into the collaborative filtering recommendation algorithm, and continuously optimize and improve the collaborative filtering recommendation algorithm based on user feedback and data analysis results; (8) Conduct a comprehensive test of the functions of the community platform and collect user feedback through questionnaires and user interviews.

[0069] Example 2

[0070] See also Figure 3-Figure 4 In this embodiment, the specific steps of step S2 include:

[0071] S2.1: The search engine receives the query keywords entered by the user or the uploaded image samples;

[0072] S2.2: For query keywords, use the NLP method to encode the query keywords, generate word vectors, and use the improved semantic similarity algorithm to calculate the similarity between the word vectors and the labels or metadata in the graph knowledge graph. The formula is:

[0073]

[0074] in, Represents the similarity between word vector A and word vector B, w represents the weight vector with the same dimensions as A and B, which is used to adjust the importance of different dimensions. Represents the sparsity factor, which is used to deal with the situation where the vector is sparse. When the vector is very sparse, that is, contains many zero elements, the denominator will become very small, resulting in an abnormally high similarity value. By adding a non-zero value, you can avoid this situation, Represents the sensitivity adjustment parameter, which is used to control the sensitivity of similarity. When it is larger, the similarity value is more sensitive to the change of vector length. When it is small, the similarity value is insensitive to the change of vector length. represents the 2-norm, and , The configuration mode is determined by those skilled in the art through a large number of experiments.

[0075] Furthermore, NLP preprocessing of keywords mainly adopts a word segmentation method based on dictionary matching to divide the text into individual words or phrases, and uses predefined part-of-speech tagging rules to annotate the words according to their context or form. Among them, the word segmentation method based on dictionary matching and the part-of-speech tagging rules are the prior art content in this field, and are not the inventive solution of this application, and will not be elaborated here.

[0076] The specific steps of step S2 also include:

[0077] S2.3: For the uploaded image samples, use the deep learning model to extract features, generate the feature vector of the image, and use the graph search strategy to match the extracted image feature vector with the pre-trained image features in the graph knowledge graph to find similar images;

[0078] S2.4: According to the relevance of the matching results, the search results are sorted using a weighted sorting algorithm, and the sorted search results are returned to the user.

[0079] The specific formula of the weighted sorting algorithm is:

[0080]

[0081] in, M k , k ∈ [ 1 , n ] represents the kth correlation measure, w k , k ∈ [ 1 , n ] express M k , k ∈ [ 1 , n ] The corresponding weights can be adjusted according to the actual situation to reflect the importance of different metrics to the final ranking results.

[0082] The specific steps of S2.3 Chinese Library Search Strategy include:

[0083] S2.31: receiving image samples uploaded by users and performing preprocessing;

[0084] S2.32: Use the pre-trained ResNet deep learning model to extract features from the pre-processed image, and perform pooling operations on the extracted feature map to convert it into a feature vector of length h;

[0085] S2.33: Load the pre-trained image feature vector from the graph knowledge graph, and calculate the similarity between the feature vector C of the uploaded image and each pre-trained feature vector D in the graph knowledge graph according to the similarity algorithm calculation formula in S2.2 .

[0086] The specific steps of S2.3 Chinese Library Search Strategy also include:

[0087] S2.34: Define graph search strategy, based on Search the graph knowledge graph for the image most similar to the uploaded image;

[0088] S2.35: Set the similarity threshold to H, according to Sort search results;

[0089] like , then discard;

[0090] like , then the corresponding image is output as the query result.

[0091] The formula for the graph search strategy in S2.34 is:

[0092]

[0093] Among them, r represents the similarity coefficient between the uploaded image and the image in the graphic knowledge graph, d represents the color channel of the uploaded image, Indicates different color channels d and different spatial positions in the uploaded image The pixel weight parameter, represents the nonlinear transformation function of color channel d, Represents the template image in the graphic knowledge graph in color channel d and position The pixel value of represents the pixel mean of the template image in the color channel d in the graphic knowledge graph, Indicates the uploaded image in color channel d and position The pixel value of Represents the pixel mean of the uploaded image in color channel d, Represents the offset parameter of the template image in the color channel d in the graphic knowledge graph, Indicates the offset parameter of the uploaded image in color channel d.

[0094] Example 3

[0095] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of a method for graphically displaying search engine retrieval results when executing the computer program.

[0096] A computer-readable storage medium stores computer instructions, which, when executed, execute the steps of a method for graphically displaying search engine retrieval results.

[0097] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in the field may also change, modify, replace and modify the above-mentioned embodiments without departing from the purpose and scope of protection of the present invention, and all of these are within the protection of the present invention.

Claims

1. A method for graphically displaying search engine retrieval results, characterized in that: include: Step S1: Establish a graph database, build a graph knowledge graph based on the graph database, organize the images, tags, and metadata information in the graph database into a graph structure, and pre-train the images in the graph database through deep learning methods to extract high-level features of the images and build an intelligent index. Users can enter query keywords or upload image samples through the search engine to initiate a search request; Step S2: The search engine receives the user request, uses the NLP method to perform semantic analysis on the query keywords, uses deep learning to identify the uploaded image, and matches it with the image in the graphic knowledge graph in real time, and determines the order of the search results based on the matching results and the sorting algorithm; Step S3: displaying the results using a carousel image according to the search result type, displaying the multi-dimensional information of the image while displaying the search results, and allowing the user to view and switch different dimensional information through interactive operations; The multi-dimensional information of the image includes: the source of the image, the number of clicks, and the number of comments; the interactive operations include: viewing the image details, saving the image, and sharing the image by clicking or sliding; Step S4: providing an intelligent color selector on the search interface, allowing the user to select or input a color for screening on the search interface, the search engine screening images according to the color feature vector, combining the shape and texture features of the graphics, and performing graphical display; Step S5: Collect user feedback and suggestions, use machine learning algorithms to perform sentiment analysis on the feedback data, and iterate and optimize the graphical knowledge graph and search result sorting algorithm based on user feedback; Step S6: Build a graphic community to allow users to share, comment and like graphic resources, introduce social graph analysis methods to obtain the correlation between users, and make intelligent recommendations based on the correlation.

2. A method for graphically displaying search engine retrieval results as claimed in claim 1, characterized in that: The specific steps of step S2 include: S2.1: The search engine receives the query keywords entered by the user or the uploaded image samples; S2.2: For query keywords, use the NLP method to encode the query keywords, generate word vectors, and use the improved semantic similarity algorithm to calculate the similarity between the word vectors and the labels or metadata in the graph knowledge graph. The formula is: ； in, represents the similarity between word vector A and word vector B, w represents the weight vector with the same dimension as A and B, represents the sparsity factor, represents the sensitivity adjustment parameter, represents the 2-norm.

3. A method for graphically displaying search engine retrieval results as claimed in claim 2, characterized in that: The specific steps of step S2 also include: S2.3: For the uploaded image samples, use the deep learning model to extract features, generate the feature vector of the image, and use the graph search strategy to match the extracted image feature vector with the pre-trained image features in the graph knowledge graph to find similar images; S2.4: According to the relevance of the matching results, the search results are sorted using a weighted sorting algorithm, and the sorted search results are returned to the user.

4. A method for graphically displaying search engine retrieval results as claimed in claim 3, characterized in that: The specific steps of the graph search strategy in S2.3 include: S2.31: receiving image samples uploaded by users and performing preprocessing; S2.32: Use the pre-trained ResNet deep learning model to extract features from the pre-processed image, and perform pooling operations on the extracted feature map to convert it into a feature vector of length h; S2.33: Load the pre-trained image feature vector from the graph knowledge graph, and calculate the similarity between the feature vector C of the uploaded image and each pre-trained feature vector D in the graph knowledge graph according to the similarity algorithm calculation formula in S2.2 .

5. A method for graphically displaying search engine retrieval results as claimed in claim 4, characterized in that: The specific steps of the graph search strategy in S2.3 also include: S2.34: Define graph search strategy, based on Search the graph knowledge graph for the image most similar to the uploaded image; S2.35: Set the similarity threshold to H, according to Sort search results; like , then discard; like , then the corresponding image is output as the query result.

6. A method for graphically displaying search engine retrieval results as claimed in claim 5, characterized in that: The formula of the graph search strategy in S2.34 is: ； Among them, r represents the similarity coefficient between the uploaded image and the image in the graphic knowledge graph, d represents the color channel of the uploaded image, Indicates different color channels d and different spatial positions in the uploaded image The pixel weight parameter, represents the nonlinear transformation function of color channel d, Represents the template image in the graphic knowledge graph in color channel d and position The pixel value of represents the pixel mean of the template image in the color channel d in the graphic knowledge graph, Indicates the uploaded image in color channel d and position The pixel value of Represents the pixel mean of the uploaded image in color channel d, Represents the offset parameter of the template image in the color channel d in the graphic knowledge graph, Indicates the offset parameter of the uploaded image in color channel d.

7. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for graphically displaying search engine retrieval results as described in any one of claims 1 to 6 are implemented.

8. A computer-readable storage medium, characterized in that: Computer instructions are stored thereon, and when the computer instructions are executed, the steps of a method for graphically displaying search engine retrieval results as described in any one of claims 1 to 6 are executed.

Images