Keyword page management system, method, device and storage medium

Abstract

The present application relates to the technical field of in-site search, and in particular to generating a keyword page based on input in-site search keywords, which includes the following steps: receiving in-site search keywords, matching product information, and obtaining a search result set about the product; classifying the product attributes of the search results to obtain a classification result set; performing feature recognition on each classification result set to obtain a feature classification subset; determining the weight of the feature classification subset according to the number of historical visits, and generating a keyword page according to the weight ratio of the feature classification subset. The present application overcomes the technical problems that the results displayed on the keyword page generated by directly inputting keywords have low relevance and are relatively random, and can help users generate more accurate search result pages.

Description

Technical Field

[0001] The present application relates to the technical field of site search, and in particular to a keyword page management system, method, device and storage medium. Background Art

[0002] The keywords of a website are important for bringing traffic to the website and benefits to the enterprise. The keyword page generated by the website theme, and then through SEO (Search Engine Optimization, i.e. search engine optimization) to improve the conversion rate, can often drive the website ranking up, thereby bringing traffic to the website and increasing the website sales.

[0003] Keyword pages can often give customers a clearer shopping direction. When customers shop, they have keywords for the products they want to buy, and keyword pages can allow customers to find the products they want from the website more quickly, providing customers with a more friendly shopping experience.

[0004] However, since the results displayed on the keyword page generated by directly entering keywords are less relevant and more random, how to help users generate more accurate search result pages is a technical problem that needs to be solved urgently. Summary of the invention

[0005] In order to improve the relevance of keyword page display results generated by keywords and generate more accurate search result pages, the present application provides a keyword page management system, method, device and storage medium.

[0006] In the first aspect, the present application provides a keyword page management method, which adopts the following technical solution:

[0007] A keyword page management method comprises the following steps:

[0008] Receive the search keywords in the site, match the product information, and obtain a set of search results about the product;

[0009] Classifying the product attributes of the search results to obtain a classification result set;

[0010] Performing feature recognition on the commodity feature set of each classification result to obtain a feature classification subset;

[0011] Collecting historical access data, and determining the weight of the feature classification subset according to the historical access data;

[0012] Generate a keyword page based on the weight ratio of the feature classification subset.

[0013] By classifying the product attributes of the search results, we can identify the product information with higher relevance, and further improve the accuracy of the search results based on feature recognition. We can also rank the product information based on the weight ratio, so that the product information with more visits is ranked higher, which helps most people find the products they want. This application has the effect of improving the relevance of the results displayed on the keyword page generated by keywords and generating a more accurate search result page.

[0014] Optionally, a keyword page management method is provided, wherein the keyword page includes a main product list and a page data information list, and the generating of the keyword page according to the weight proportion of the feature classification subset includes the following steps:

[0015] When the weight ratio of the feature classification subset is greater than a first threshold, the feature classification subset is in the expanded main product list information, and is displayed in the main product list according to the feature classification subset;

[0016] When the weight ratio of the feature classification subset is between the first threshold and the second threshold, the feature classification subset is in the folded main product list information, and corresponding feature category data is generated according to the feature classification subset and displayed in the page data information list;

[0017] The second threshold is smaller than the first threshold.

[0018] By adopting the above technical solution, the weight ratio of the feature classification subset is calculated.

[0019] Optionally, a keyword page management method, before searching for keywords in the receiving station, also includes establishing an original database, wherein the original database contains multiple product information, each of which includes product keywords, product attributes and product features.

[0020] By adopting the above technical solution, when the weight ratio of the feature classification subset is greater than the first threshold, it means that the historical access data volume is high, and the content with high historical access data volume will be displayed in the main product list first; when the historical access data volume is low, the feature classification subset will be folded and not displayed in the main product list, but the feature category data will be generated and displayed in the page data information list, and it needs to be displayed in the main product list according to the access request instruction and other methods; when the historical access data volume is low, it indicates that this part does not meet the requirements of the search results, and this part of the feature classification subset will not be displayed. By dividing the weight interval of the feature classification subset, the search results that meet the requirements are displayed first, thereby improving the relevance of the results displayed on the keyword page.

[0021] Optionally, in a keyword page management method, the specific method of collecting historical access data includes:

[0022] Obtain the length of time it takes for a user to access product information and place an order, and compare the time with the expected time. If the time is less than the expected time, determine that the time is abnormal, and remove the relevant historical access data based on the abnormal time.

[0023] By adopting the above technical solution, it is possible to determine whether a seller has engaged in fake order behavior by judging the length of time it takes for a user to access product information and place an order. Usually, the length of time taken by sellers who engage in fake order behavior is shorter. Therefore, if the length of time is less than the time estimate, it can be considered that the historical access data is invalid and can be eliminated, thereby making the determination of the weight of the feature classification subset more accurate.

[0024] Optionally, in a keyword page management method, the specific method of collecting historical access data includes:

[0025] Obtain the user's comment pictures for a single product, and identify all the comment pictures under the single product to determine the similarity between the comment pictures. If there are at least two comment pictures whose similarity is greater than the duplicate index threshold, the comment pictures are determined to be abnormal, and the relevant historical access data is removed based on the abnormal comment pictures.

[0026] By adopting the above technical solution, it is determined whether the seller has engaged in fake orders by judging the similarity between the comment pictures. Usually, sellers who engage in fake orders will steal or reuse the same comment pictures, and may also slightly modify the comment pictures and then steal or reuse the pictures. Therefore, if the similarity of the comment pictures is greater than the duplicate index threshold, it can be considered that this historical access data is invalid and can be eliminated, thereby making the determination of the weights of the feature classification subsets more accurate.

[0027] Optionally, in a keyword page management method, the specific method of determining the similarity between comment pictures includes:

[0028] All the review images under a single product are input into the recognition detection model to match the similarities between the review images;

[0029] If the same elements between the comment images exceed the duplicate index threshold, the comment images are determined to be abnormal.

[0030] By adopting the above technical solution, whether the comment pictures are abnormal can be determined by judging whether the same elements between the comment pictures exceed the duplicate index threshold, so as to more accurately determine whether there is any seller fake order behavior in the historical data.

[0031] Optionally, in a keyword page management method, the specific method of feature identification includes:

[0032] Obtain the product images to be identified in each classification result set, put the product images to be identified into the pre-trained convolutional neural network model, and obtain the color category set and / or component category set of the product images.

[0033] By adopting the above technical solution, product images are identified in the classification result set, and the products in the classification result set are classified according to the color or component category of the product images, which can increase the dimension of the search result page and help users generate more accurate search result pages.

[0034] In the second aspect, the present application provides a keyword page management system, which adopts the following technical solution:

[0035] A keyword page management system, comprising:

[0036] The search module is used to receive the search keywords in the site, match the product information, and obtain the search result set about the product;

[0037] A classification module is used to classify the product attributes of the search results to obtain a classification result set;

[0038] The recognition module is used to perform feature recognition on each classification result set to obtain a feature classification subset;

[0039] The page generation module is used to determine the weight of the feature classification subset according to the historical access quantity, and generate a keyword page according to the weight ratio of the feature classification subset.

[0040] By adopting the above technical solution, the search results are classified by product attributes, and product information with higher relevance is divided. Then, the accuracy of the search results is further improved based on feature recognition, and the product information is ranked according to the weight ratio, so that product information with more visits is ranked higher, which helps most people find the products they want. This application has the effect of improving the relevance of the results displayed on the keyword page generated by keywords and generating a more accurate search result page.

[0041] In a third aspect, the present application provides a keyword page management device, which adopts the following technical solution:

[0042] A keyword page management device includes a memory and a processor, wherein the processor can be loaded with and execute any one of the keyword page management methods described above.

[0043] In a fourth aspect, the present application provides a computer-readable storage medium, which adopts the following technical solution:

[0044] A computer-readable storage medium stores a keyword page management method as described in any one of the above which can be loaded and executed by a processor. DETAILED DESCRIPTION

[0045] The embodiment of the present application discloses a keyword page management method.

[0046] A keyword page management method comprises the following steps:

[0047] S0. Establish an original database, enter commodity information of commodities in the original database, so that there are multiple commodity information in the original database, and each commodity information includes commodity keywords, commodity attributes and commodity features.

[0048] In a specific embodiment, if the product is a laptop computer, the product keyword may be "notebook", "laptop computer", "ultrabook", "entertainment notebook", "gaming notebook", and the keyword may also be accompanied by the corresponding brand name, series name, memory, size, CPU series, etc. For example, the keyword may be "Zhongwei Notebook MateBook 24-inch 2K HD Eight-core Standard Pressure Business Gaming i7 Fingerprint Metal 2024 New Laptop"; the product attribute is "laptop computer"; the product features may include brand "Zhongwei", size "24 inches", hard disk capacity "512G", memory "16G", graphics card type "independent graphics card", "fingerprint recognition", "face unlocking", "wireless charging", etc.

[0049] S1. Receive the search keywords in the site, match the product information, and obtain a set of search results about the product.

[0050] In a specific embodiment, after receiving the in-site search keyword "notebook", the in-site search keyword "notebook" is compared with a large amount of product information in the database to generate all search results containing the in-site search keyword "notebook".

[0051] S2. Classify the product attributes of the search results to obtain a classification result set.

[0052] According to the product attributes of the search results, the search results can be classified into "laptops" and "notebook stationery". By classifying the product attributes, irrelevant search results can be removed, the scope of product search can be narrowed, and the relevance of product search can be improved.

[0053] S3. Feature recognition is performed on the commodity features of each classification result set to obtain a feature classification subset. In a specific embodiment, feature classification can be divided into "16G", "8G", "4G", etc. from the memory dimension, and the graphics card type can be divided into "independent graphics card" and "integrated graphics card". A notebook can belong to the "16G" and "independent graphics card" categories.

[0054] The specific method of feature recognition described above may also include obtaining a product image to be identified in each classification result set, placing the product image to be identified into a pre-trained convolutional neural network model, and obtaining at least one of a color matching category set and a component category set of the product image. Since the convolutional neural network model is an existing technology, it will not be described in detail here.

[0055] S4, collecting historical access data, and determining the weight of the feature classification subset according to the historical access data. The specific method of collecting historical access data includes:

[0056] Obtain the length of time it takes for a user to access product information and place an order, and compare the length of time with the time estimate. If the length of time is less than the time estimate, the length of time is determined to be abnormal, and the relevant historical access data is removed based on the abnormal length of time. Sellers who engage in fake ordering usually have shorter lengths of time, so if the length of time is less than the time estimate, the historical access data can be considered invalid and can be removed. By determining the length of time it takes for a user to access product information and place an order, it is determined whether the seller has engaged in fake ordering, making the determination of the weight of the feature classification subset more accurate.

[0057] The specific method for collecting historical access data includes obtaining the user's comment pictures for a single product, identifying all comment pictures under a single product, and determining the similarity between the comment pictures. If there are at least two comment pictures whose similarity is greater than the duplicate index threshold, the comment picture is determined to be abnormal, and the relevant historical access data is eliminated based on the abnormal comment pictures.

[0058] The specific method for determining the similarity between the review images includes inputting all review images under a single product into a recognition detection model to match the similarity between the review images;

[0059] If the same elements between the comment images exceed the duplicate index threshold, the comment image is judged to be abnormal.

[0060] S5. Generate a keyword page according to the weight ratio of the feature classification subset.

[0061] The keyword page includes a main product list and a page data information list. Generating a keyword page according to the weight ratio of the feature classification subset includes the following steps: when the weight ratio of the feature classification subset is greater than the first threshold, the feature classification subset is in the expanded main product list information, and is displayed in the main product list according to the feature classification subset. When the weight ratio of the feature classification subset is between the first threshold and the second threshold, the feature classification subset is in the collapsed main product list information, and the corresponding feature category data is generated according to the feature classification subset and displayed in the page data information list. Among them, the second threshold is less than the first threshold.

[0062] The implementation principle of a keyword page management method in the embodiment of the present application is: by classifying the product attributes of the search results, the product information with higher relevance is divided, and then the accuracy of the search results is further improved according to feature recognition, and the product information is ranked according to the weight ratio, so that the product information with more visits is ranked higher, thereby helping most people to find the desired products. The present application has the effect of improving the relevance of the keyword page display results generated by the keywords and generating a more accurate search result page.

[0063] A keyword page management system, comprising:

[0064] The search module is used to receive the search keywords in the site, match the product information, and obtain the search result set about the product;

[0065] A classification module is used to classify the product attributes of the search results to obtain a classification result set;

[0066] The recognition module is used to perform feature recognition on each classification result set to obtain a feature classification subset;

[0067] The page generation module is used to determine the weight of the feature classification subset according to the historical access quantity, and generate a keyword page according to the weight ratio of the feature classification subset.

[0068] The implementation principle of a keyword page management method in the embodiment of the present application is: by classifying the product attributes of the search results, the product information with higher relevance is divided, and then the accuracy of the search results is further improved according to feature recognition, and the product information is ranked according to the weight ratio, so that the product information with more visits is ranked higher, thereby helping most people to find the desired products. The present application has the effect of improving the relevance of the keyword page display results generated by the keywords and generating a more accurate search result page.

[0069] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A keyword page management method, characterized in that: The following steps are involved: Receive the search keywords in the site, match the product information, and obtain a set of search results about the product; Classify the product attributes of the search results to obtain a classification result set; Perform feature recognition on the commodity feature set of each classification result to obtain a feature classification subset; The specific method of feature recognition includes: obtaining a product image to be recognized in each classification result set, putting the product image to be recognized into a pre-trained convolutional neural network model, and obtaining a color category set and / or a component category set of the product image; Collecting historical access data, and determining the weight of the feature classification subset according to the historical access data; The specific method for collecting historical access data includes: obtaining the length of time it takes for a user to access product information and place an order, comparing the length of time with a time threshold, and if the length of time is less than the time threshold, determining that the length of time is abnormal, and eliminating relevant historical access data based on the abnormal length of time. The length of time used by sellers who engage in fake order behavior is shorter, so if the length of time is less than the time threshold, the historical access data is considered invalid and is eliminated; the specific method for collecting historical access data also includes obtaining a user's comment image on a single product, and identifying all comment images under a single product, determining the similarity between the comment images, and if there are at least two comment images whose similarity is greater than a duplicate index threshold, determining that the comment image is abnormal, and eliminating relevant historical access data based on the abnormal comment images. The specific method for determining the similarity between comment images includes inputting all comment images under a single product into a recognition detection model, matching the similarity between comment images, and determining that the comment images are abnormal if the same elements between the comment images exceed the duplicate index threshold; Generate a keyword page according to the weight proportion of the feature classification subset; The keyword page includes a main product list and a page data information list, and the step of generating a keyword page according to the weight proportion of the feature classification subset includes the following steps: When the weight ratio of the feature classification subset is greater than a first threshold, the feature classification subset is in the expanded main product list information, and is displayed in the main product list according to the feature classification subset; When the weight ratio of the feature classification subset is between the first threshold and the second threshold, the feature classification subset is in the folded main product list information, and corresponding feature category data is generated according to the feature classification subset and displayed in the page data information list; The second threshold is smaller than the first threshold.

2. A keyword page management method according to claim 1, characterized in that: Before searching keywords in the receiving station, an original database is also established. The original database contains a plurality of commodity information, each of which includes commodity keywords, commodity attributes and commodity features.

3. A keyword page management system, characterized in that: include: The search module is used to receive the search keywords in the site, match the product information, and obtain the search result set about the product; A classification module is used to classify the product attributes of the search results to obtain a classification result set; The recognition module is used to perform feature recognition on the commodity feature set of each classification result to obtain a feature classification subset; The specific method of feature recognition includes: obtaining product images to be identified in each classification result set, putting the product images to be identified into a pre-trained convolutional neural network model to obtain a color category set and / or a component category set of the product images; collecting historical access data, and determining the weight of the feature classification subset according to the historical access data; the specific method of collecting historical access data includes: obtaining the length of time it takes for a user to access product information and place an order, comparing the length of time with a time threshold, and if the length of time is less than the time threshold, determining that the length of time is abnormal, and eliminating relevant historical access data according to the abnormal length of time. The length of time used by sellers who engage in fake orders is shorter, so if the length of time is less than the time threshold, the length of time used is abnormal. If the historical access data exceeds the threshold, the historical access data is considered invalid and is removed; the specific method for collecting historical access data also includes obtaining the user's comment pictures on a single product, and identifying all the comment pictures under the single product, determining the similarity between the comment pictures, and if there are at least two comment pictures whose similarity is greater than the duplicate index threshold, the comment pictures are determined to be abnormal, and the relevant historical access data are removed according to the abnormal comment pictures. The specific method for determining the similarity between the comment pictures includes inputting all the comment pictures under the single product into the recognition detection model, matching the similarity between the comment pictures, and if the same elements between the comment pictures exceed the duplicate index threshold, the comment pictures are determined to be abnormal; A page generation module is used to generate a keyword page according to the weight ratio of the feature classification subset; the keyword page includes a main product list and a page data information list, and the generation of the keyword page according to the weight ratio of the feature classification subset includes the following steps: when the weight ratio of the feature classification subset is greater than a first threshold, the feature classification subset is in the expanded main product list information, and is displayed in the main product list according to the feature classification subset; when the weight ratio of the feature classification subset is between the first threshold and the second threshold, the feature classification subset is in the folded main product list information, corresponding feature category data is generated according to the feature classification subset, and is displayed in the page data information list; wherein the second threshold is less than the first threshold.

4. A keyword page management device, comprising a memory and a processor, characterized in that: The processor can be loaded and execute the keyword page management method as described in any one of claims 1-2.

5. A computer-readable storage medium, characterized in that: The method for managing keyword pages according to any one of claims 1 to 2 is stored and can be loaded and executed by a processor.