Classification device, classification method, and classification program

The classification device uses image processing to extract features from screen capture images, enabling accurate web page categorization based on visual appearance, addressing the limitations of conventional methods that rely on linguistic and structural information.

WO2024241444A1PCT designated stage expired Publication Date: 2024-11-28NT T INC

Patent Information

Application Number
PCT/JP2023/019020
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-22
Publication Date
2024-11-28

AI Technical Summary

Technical Problem

Conventional web page classification methods rely on linguistic information, internal structure, and URL, which can lead to inaccurate categorization due to variations in display format, internal HTML structure, and URL composition, failing to accurately classify web pages based on their display format and structure.

Method used

A classification device and method that uses image processing to extract image features from screen capture images of web pages, allowing for classification based on visual appearance independent of text information and internal structure, employing machine learning models to categorize web pages based on extracted image features.

Benefits of technology

Enables accurate classification of web pages into categories based on their visual appearance, improving categorization accuracy by disregarding text information and internal structure dependencies.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure JP2023019020_28112024_PF_FP_ABST
    Figure JP2023019020_28112024_PF_FP_ABST
Patent Text Reader

Abstract

A classification device (10) has, at least, an input unit (11) that receives input of a webpage image, which is a screen capture image when a website formed from a plurality of webpages is displayed, an image feature extraction unit (12) that applies image processing to the webpage image to extract an image feature amount of the webpage image, and a classification unit (14) that classifies the webpage indicated by the webpage image on the basis of the image feature amount.
Need to check novelty before this filing date? Find Prior Art

Description

Classification device, classification method, and classification program

[0001] The present invention relates to a classification device, a classification method, and a classification program.

[0002] Conventionally, there are techniques for classifying web site pages. For example, there is a method for classifying web pages using word n-grams contained in the text of the web page (Non-Patent Document 1). In the method described in Non-Patent Document 1, web pages are classified using a model that has learned word n-grams that tend to appear in a specific category.

[0003] There is also a method for classifying a web page by analyzing the HTML of the web page and limiting the text to be extracted (for example, see Patent Document 1). The method described in Patent Document 1 analyzes the text in a specific tag or the HTML of the web page as a tree structure. The method described in Patent Document 1 extracts text by limiting the extraction location based on the internal structure of the HTML. The method described in Patent Document 1 then uses morphological analysis to determine which category the web page belongs to using a classification model for each category, using noun phrases contained in the extracted text and their order of appearance as features.

[0004] There is also a method for classifying web pages based on the URL when the web page is accessed (for example, Patent Document 2). In the method described in Patent Document 2, the URL when accessed is treated as a character string, and each substring separated by a slash " / " or a period "." is used as a feature, and the web page is classified based on the degree of match of the substring. There is also a method for determining the category of a web page based on whether the substring matches a dictionary registered in advance.

[0005] JP 2013-196382 A JP 2004-341942 A

[0006] Koichi Hirano, Noriya Furubayashi, and Junichi Takahashi, "Automatic Classification of Japanese-speaking Blogs," Information Processing Society of Japan Technical Report, 2005.

[0007] The method described in Non-Patent Document 1, which uses text contained in web pages, uses words contained in the text as clues. Therefore, the method described in Non-Patent Document 1 has the problem that it classifies based on the linguistic information written in the text, and is unable to classify based on the display format of the web page.

[0008] For example, consider a case where there are two web pages with the function of presenting detailed product information: one that presents details about a certain piece of clothing A and another that presents details about a certain bag B. These two web pages are of the same type, presenting product details, and the display formats of the web pages, such as the layout of product images and descriptions, are similar, but the linguistic information obtained from the included text is significantly different. For this reason, if the method described in Patent Document 1 is used, the two web pages will be determined to be in different categories.

[0009] Furthermore, the method described in Patent Document 1 has a problem in that differences in the display format of the displayed Web page do not necessarily match differences in the internal structure of the HTML.

[0010] For example, if there are web pages C and D that display information of the same category, the internal HTML structures of each web page C and D may differ due to tags or tag ID attributes for purposes other than display, such as analysis, or the embedding of web beacons. Furthermore, even if the same web page C is accessed, different browser types and versions may result in analysis results that show different internal structures. In either case, even though the web page users perceive the web pages as belonging to the same category, or even if they are actually the same page, the method described in Patent Document 1 classifies the web pages as belonging to different categories.

[0011] Furthermore, the method described in Patent Document 2 has a problem in that differences in the display format of the displayed web page do not necessarily correspond to differences in the URL. An example of this case is a URL string such as "https: / / www.example.com / sid3a5ebc" in which the path portion is composed of a character string such as an internal website management ID. Furthermore, there are cases in which the URL includes information indicating the currently displayed web page and other information, such as page transitions and operations leading up to the display of that page (for example, the part after "ref" in "https: / / www.example.com / products.php?&ref=foo&bar&baz").

[0012] Therefore, with the conventional method, there are cases where it is not possible to accurately categorize a web page due to the dependency on the text information, internal structure, and URL of the web page.

[0013] The present invention has been made in view of the above, and an object of the present invention is to provide a classification device, a classification method, and a classification program that can accurately classify categories of web pages.

[0014] In order to solve the above-mentioned problems and achieve the objectives, the classification device is characterized by having at least an input unit that accepts input of a web page image, which is a screen capture image when a website consisting of multiple web pages is displayed, an extraction unit that performs image processing on the web page image and extracts image features of the web page image, and a classification unit that classifies the web page indicated by the web page image based on the image features.

[0015] According to the present invention, it is possible to accurately classify categories of web pages.

[0016] FIG. 1 is a diagram illustrating an example of the configuration of a classification device according to Embodiment 1. FIG. 2 is a diagram illustrating an example of the data configuration of a web page image. FIG. 3 is a diagram illustrating an example of the data configuration of a web page image. FIG. 4 is a diagram illustrating an example of a classification result output by the classification device illustrated in FIG. 1. FIG. 5 is a diagram illustrating an example of a computer that executes a classification program. FIG. 6 is a diagram illustrating an example of the configuration of a classification device according to Embodiment 2. FIG. 7 is a flowchart illustrating the procedure for classification processing according to Embodiment 2. FIG. 8 is a diagram illustrating an example of the configuration of a classification device according to Embodiment 3. FIG. 9 is a diagram illustrating an example of the data configuration of a web page image and associated information. FIG. 10 is a flowchart illustrating the procedure for classification processing according to Embodiment 3. FIG. 11 is a diagram illustrating an example of the configuration of a classification device according to Modification 2 of Embodiment 3. FIG. 12 is a flowchart illustrating the procedure for classification processing according to Modification 2 of Embodiment 3. FIG. 13 is a diagram illustrating an example of a computer that implements a classification device by executing a program.

[0017] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a classification device, a classification method, and a classification program according to the present invention will be described in detail with reference to the accompanying drawings. However, the present invention is not limited to the following embodiments.

[0018] [First Embodiment] In the first embodiment, the pages of a website are classified into categories using image features of a screen capture image (hereinafter referred to as a web page image) acquired when the page of the website is displayed. In the first embodiment, image processing is performed on the web page image, which is a screen capture image acquired when a website consisting of multiple web pages is displayed, to extract image features of the web page image, and to classify the web page indicated by the web page image based on the extracted image features.

[0019] [Classification Device] First, a description will be given of a classification device according to embodiment 1. Fig. 1 is a diagram showing an example of the configuration of a classification device according to embodiment 1.

[0020] The classification device 10 according to the first embodiment is implemented by loading a predetermined program into a computer or the like including a read-only memory (ROM), a random access memory (RAM), a central processing unit (CPU), etc., and causing the CPU to execute the predetermined program. The classification device 10 also has a communication interface for transmitting and receiving various types of information to and from other devices connected via a network or the like.

[0021] As shown in FIG. 1, the classification devices 10 , 210 , and 310 include an input unit 11 , an image feature extraction unit 12 (extraction unit), a category feature storage unit 13 , a classification unit 14 , and an output unit 15 .

[0022] The input unit 11 receives an input of a web page image 20 to be classified. The web page image 20 includes a screen capture image of a page of a website consisting of multiple pages when the page is displayed. The input unit 11 receives an input of the web page image 20 transmitted from another device connected via a network, for example.

[0023] 2 and 3 are diagrams illustrating an example of the data structure of a web page image 20. Specifically, the input unit 11 accepts input of a list 21 (FIG. 2) and an actual image file 22 (FIG. 3) as the web page image 20. The list 21 is composed of an ID that identifies the web page image and a path where the web page image is located. The image file 22 is a screen capture image of the web page when the page of the website is displayed, and corresponds to the web page image itself.

[0024] The image feature extraction unit 12 performs image processing on the web page image 20 (screen capture image) to extract image features as first feature amounts. For example, the image feature extraction unit 12 obtains a three-dimensional array consisting of RGB values ​​of each pixel of the web page image 20, such as the array [[[83, 109, 109], [82, 108, 108], [85, 109, 107], ..., [100, 141, 104], [100, 143, 106], [100, 144, 105]], ...], and sets this as the first feature amount.

[0025] The first feature amount may be any image feature amount that represents the appearance of the web page. For example, the first feature amount may be whether or not a specific partial image designated in advance is included.

[0026] The category feature storage unit 13 stores a plurality of models 131 (first machine learning models). Each model 131 is a machine learning model provided for each category of web pages. When a first feature of a web page image is input, each model 131 classifies the web page image as belonging to a category corresponding to the model 131.

[0027] The model 131 has already been trained to classify web pages into categories based on image features extracted from training web page images. The model 131 is trained using training data of positive examples and training data of negative examples extracted from the training web page images.

[0028] The training data of positive examples are image features extracted from web page images that belong to the category to be classified. The training data of negative examples are image features extracted from web page images that do not belong to the category to be classified. In training the model 131, for example, a machine learning algorithm such as a support vector machine (SVM) can be used.

[0029] Specifically, when each model 131 receives input of the first feature amount of the web page image 20 output from the image feature extraction unit 12 , it outputs a score value for the category of the model 131 .

[0030] For example, if classification into 10 categories is desired, a model 131 is created in advance for each of the 10 categories and stored in the category feature storage unit 13 .

[0031] For example, in the case of an EC (Electronic Commerce) site, the categories of web page images are classified according to the web page function, such as top page, search page, search result list page, individual product page, shop information page, etc. Furthermore, the categories of web page images may be classified according to the display pattern, such as the presence or absence of images or the arrangement of images, or the color tone of the web page, as long as the classification can be determined from the display appearance.

[0032] The classification unit 14 classifies, for each input web page image 20, the category to which the web page indicated by this web page image 20 belongs, based on the model 131 of each category and the first feature of the web page image 20 extracted by the image feature extraction unit 12.

[0033] Specifically, when classifying into 10 categories, the classification unit 14 calculates the score value of the web page image for each category using each model 131 corresponding to each of the 10 categories. Then, the classification unit 14 classifies the category with the largest score value as the category to which the web page indicated by the web page image to be classified (hereinafter referred to as the web page to be classified) belongs.

[0034] Furthermore, the classification unit 14 may assign a category to the web page to be classified, not limited to the highest score value, but for example, multiple categories with scores equal to or greater than a threshold value.

[0035] The output unit 15 outputs various information including the classification results. The output unit 15 is, for example, a liquid crystal display, a printer, or the like, and may also output various information to other devices connected via a network, etc.

[0036] Fig. 4 is a diagram showing an example of a classification result output by the classification device 10 shown in Fig. 1. As shown in Fig. 4, the output unit 15 outputs a combination of the input web page image and the category as a web page category 30, which is the classification result.

[0037] [Classification Processing] Next, a description will be given of the processing procedure of the classification processing according to Embodiment 1. Fig. 5 is a flowchart showing the processing procedure of the classification processing according to Embodiment 1.

[0038] As shown in FIG. 5, when the classification device 10 receives input of a web page image 20 to be classified (step S1), the image feature extraction unit 12 performs an image feature extraction process to extract image features of the web page image 20 as first features (step S2).

[0039] The classification unit 14 performs a classification process to classify the category to which the web page indicated by the web page image 20 belongs based on the model 131 of each category and the first feature extracted in step S2 (step S3), and outputs the classification result from the output unit 15 (step S4).

[0040] [Effects of Embodiment 1] As described above, the classification device 10 according to Embodiment 1 classifies a category to which a web page indicated by a web page image belongs, using, as a first feature, an image feature extracted from the web page image, which is a screen capture image acquired when a page of a website is displayed. Therefore, the classification device 10 performs classification based on the display appearance of the web page, and can classify a category to which a web page to be classified belongs without relying on the text information, internal structure, and URL of the web page.

[0041] [Embodiment 2] Next, a description will be given of embodiment 2. Fig. 6 is a diagram showing an example of the configuration of a classification device according to embodiment 2.

[0042] A classification device 210 according to the second embodiment has a classification unit 214 instead of the classification unit 14 of the classification device 10 shown in Fig. 1. Note that the classification device 210 can omit the category feature storage unit 13.

[0043] The classification unit 214 groups the plurality of web page images 20 based on the similarity of the first feature amounts of the web page images 20 extracted by the image feature extraction unit 12, and classifies each group into one category. For example, the classification unit 214 assigns a category to each group according to a predetermined rule.

[0044] The grouping method based on the similarity of the first feature amount can be, for example, a clustering algorithm such as k-means or x-means. The output unit 215 outputs the combination of the input web page image and category as a web page category 230.

[0045] [Classification Processing] Fig. 7 is a flowchart showing the processing procedure of the classification processing according to embodiment 2. Steps S11 and S12 shown in Fig. 7 are the same processes as steps S1 and S2 shown in Fig. 5.

[0046] In the classification device 210, the classification unit 214 performs a classification process in which the multiple web page images are grouped based on the similarity of the first feature extracted in step S2, and each group is classified into a single category (step S13). Step S14 is the same process as step S3 shown in FIG. 5.

[0047] [Effects of Embodiment 2] As in this embodiment 2, even when grouping and categorizing web page images 20 based on the similarity of the first feature of the web page images 20, the web pages can be categorized without relying on the text information, internal structure, and URL of the web page.

[0048] [Third Embodiment] Next, a third embodiment will be described. In the third embodiment, classification is performed using not only the web page image but also the attached information when the web page image is displayed.

[0049] 8 is a diagram showing an example of the configuration of a classification device according to embodiment 3. A classification device 310 according to embodiment 3 has the same functions as the classification device 10. The classification device 310 has an input unit 11, a feature extraction unit 312 (extraction unit), a category feature storage unit 313, a classification unit 314, and an output unit 15.

[0050] The input unit 11 accepts input of a web page image and associated information 320. The associated information is information related to the web page indicated by the web page image. Specifically, the associated information is the URL when the web page image is displayed and the HTML source code of the web page indicated by the web page image.

[0051] 9 is a diagram illustrating an example of the data configuration of the web page image and attached information 320. For example, the web page image and attached information 320 includes a list 321 (FIG. 9) and the actual image file 22 (FIG. 3).

[0052] 9, the list 321 is composed of an ID for identifying a web page image, a path where the web page image is located, a URL when the web page image is displayed, and an HTML source code of the web page. Note that the attached information may be information other than the URL when the web page image is displayed and the HTML source code of the web page, as long as it is information acquired when the web page is displayed.

[0053] The feature extraction unit 312 performs image processing on the web page image to extract image features of the web page image as first feature amounts, and also extracts feature amounts from the attached information. The feature extraction unit 312 includes an image feature extraction unit 12, a URL feature extraction unit 3122, an HTML feature extraction unit 3123, and a feature synthesis unit 3124.

[0054] The URL feature extraction unit 3122 extracts features of the URL when the web page image to be classified is displayed, and sets the extracted features as the second feature. For example, the URL feature extraction unit 3122 regards the URL as a character string, replaces slashes " / ", periods ".", and the digits "0" to "9" with spaces, and extracts the arrangement of the character string that has undergone stem extraction processing as the second feature.

[0055] Specifically, the URL feature extraction unit 3122 extracts "www, hostc, com, product, product" from the URL "http: / / www.hostC.com / products&ref=procuts" as the second feature. The extraction process of the URL feature extraction unit 3122 may be limited to extracting a character string sequence as a feature from only the path portion of the URL.

[0056] The HTML feature extraction unit 3123 extracts features of the HTML source code of the web page to be classified and sets the extracted features as the third feature amount. For example, the HTML feature extraction unit 3123 sets the result of morphological analysis of the text displayed on the web page as the third feature amount.

[0057] The HTML feature extraction unit 3123 may also employ an extraction algorithm other than morphological analysis, such as an algorithm for extracting word N-grams.The HTML feature extraction unit 3123 may also extract features by limiting the extraction range to, for example, the page title (the text enclosed by tags).

[0058] The feature combining unit 3124 combines the first feature, the second feature, and the third feature, and outputs the combined feature to the classification unit 314. The feature combining unit 3124 may combine the features using various methods. For example, the feature combining unit 3124 may combine the individual feature quantities to form a single feature quantity, or may combine the individual feature quantities and then select and discard the individual feature quantities using Maxpooling.

[0059] The category feature storage unit 313 stores a plurality of models 3131 (second machine learning models). The models 3131 are machine learning models that have already been trained.

[0060] The model 3131 has learned to categorize web pages through machine learning based on features extracted from training web page images and the attached information of the training web page images.

[0061] The model 3131 is a model trained using composite features (positive example training data) extracted from web page images belonging to a category to be classified and auxiliary information of learning web page images, and composite features (negative example training data) extracted from web page images not belonging to a category to be classified and auxiliary information. The model 3131 is provided for each category to be classified.

[0062] The classification unit 314 classifies the category to which the web page indicated by the web page image belongs, using the model 3131 and the first feature and the feature extracted from the attached information. Specifically, the classification unit 314 classifies the category to which the web page indicated by the web page image belongs, based on the model 3131 and the combined feature combined by the feature combination unit 3124.

[0063] The classification unit 314, like the classification unit 14, calculates a score value for each category using each model 131 corresponding to each category. The classification unit 14 then classifies the category with the largest score value as the category to which the web page to be classified belongs. The output unit 15 outputs the combination of the input web page image and the category as the classification result, web page category 30 (see, for example, FIG. 4 ).

[0064] [Classification Processing] Next, a description will be given of the processing procedure of the classification processing according to the embodiment 3. Fig. 9 is a flowchart showing the processing procedure of the classification processing according to the embodiment 3.

[0065] When the classification device 310 receives input of the web page image and the attached information 320 to be classified (step S21), the feature extraction unit 312 extracts first to third feature amounts.

[0066] Specifically, the image feature extraction unit 12 extracts image features of the web page image 20 as first features (step S22). The URL feature extraction unit 3122 extracts URL features of the web page image 20 from the ancillary information as second features (step S23). The HTML feature extraction unit 3123 extracts HTML source code features of the web page to be classified as third features (step S24).

[0067] The feature amount combining unit 3124 combines the first feature amount, the second feature amount, and the third feature amount (step S25), and outputs the combined feature amount to the classifying unit 314.

[0068] The classification unit 314 classifies the category to which the web page indicated by the web page image belongs based on the model 3131 and the combined features combined by the feature combination unit 3124 (step S26), and outputs the classification result from the output unit 15 (step S27).

[0069] [Effects of Embodiment 3] In this way, in embodiment 3, in addition to the web page image, attached information when the web page image is displayed is also used to classify the web page into categories, thereby improving classification accuracy compared to embodiment 1.

[0070] [First Modification of Third Embodiment] Category classification may be performed based on each of the first to third feature amounts, and the results obtained based on the first to third feature amounts may be integrated. In this case, a model corresponding to each of the first to third feature amounts is provided for each category.

[0071] For example, in addition to model 131 that performs category classification based on a first feature, a second model that performs category classification based on a second feature and a third model that performs category classification based on a third feature are provided.

[0072] The second model is a model trained using URL features (positive example training data) of training web page images that belong to the category to be classified and URL features (negative example training data) of training web page images that do not belong to the category to be classified. The third model is a model trained using HTML source code features (positive example training data) of training web page images that belong to the category to be classified and HTML source code features (negative example training data) of training web page images that do not belong to the category to be classified.

[0073] The classification unit 314 uses the model 131, the second model, and the third model to perform category classification for each web page for each of the first to third feature amounts extracted from the web page image and the attached information 320. The classification unit 314 then integrates the results of the category classifications and performs an integrated determination of the category of the web page to be classified.

[0074] For example, as a method of integrated judgment, a method is adopted in which the results of category classification for the first to third feature amounts and their scores are calculated, and when the sum is taken, the category with the highest score is judged to be the category of the web page to be classified.

[0075] For example, the following example will be described: the result of the category classification based on the first feature amount is category A with a score value of "0.7," the result of the classification based on the second feature amount is category B with a score value of "0.8," and the result of the classification based on the third feature amount is category A with a score value of "0.5." In this case, the classification unit 314 determines that this webpage belongs to category A, where the sum of the score values ​​based on the first feature amount and the third feature amount is "1.2."

[0076] Alternatively, the integrated category determination method may involve assigning weights to the score values ​​of the first to third feature amounts and performing an integrated category determination based on corrected scores obtained by multiplying the score values ​​of the feature amounts by the weights. Alternatively, the integrated category determination method may involve assigning priorities to the category determinations of the first to third feature amounts and employing a category with a lower priority if no results are obtained from a category determination with a higher priority.

[0077] [Second Modification of Third Embodiment] The classification device 210 according to the second embodiment may also perform classification based on the web page images and the attached information 320. Fig. 11 is a diagram showing an example of the configuration of a classification device according to a second modification of the third embodiment.

[0078] A classification device 310A according to the second modification of the third embodiment includes an input unit 11, a feature extraction unit 312, a classification unit 314A, and an output unit 15.

[0079] The classification unit 314A groups the multiple web page images and classifies each group into a category based on the similarity between the multiple image features and the multiple features extracted from the auxiliary information by the feature extraction unit 312. The classification unit 314A groups the multiple web page images and classifies each group into a category based on the similarity between the multiple combined features obtained by combining the first feature, the second feature, and the third feature.

[0080] [Classification Processing] Next, a description will be given of a processing procedure of classification processing according to Modification 2 of Embodiment 3. Fig. 12 is a flowchart showing the processing procedure of classification processing according to Modification 2 of Embodiment 3.

[0081] Steps S31 to S35 in Fig. 12 are the same processes as steps S21 to S25 in Fig. 10. The classification unit 314A groups the multiple web page images based on the similarity of the multiple combined feature amounts, and classifies each group into a single category (step S36). Step S37 is the same process as step S3 shown in Fig. 5.

[0082] In this way, even when grouping and categorizing web page images 20, the classification accuracy can be improved by using additional information when the web page images are displayed in addition to the web page images themselves.

[0083] [System Configuration of the Embodiments] The components of the classification devices 10, 210, and 310 are conceptual functional components and do not necessarily need to be physically configured as shown in the drawings. In other words, the specific forms of distribution and integration of the functions of the classification devices 10, 210, and 310 are not limited to those shown in the drawings, and all or part of the components can be functionally or physically distributed or integrated in any unit depending on various loads, usage conditions, etc.

[0084] Furthermore, all or any part of the processes performed by the classification devices 10, 210, and 310 may be realized by a CPU and a program analyzed and executed by the CPU. Furthermore, each process performed by the classification devices 10, 210, and 310 may be realized as hardware using wired logic.

[0085] Furthermore, among the processes described in the embodiments, all or part of the processes described as being performed automatically can be performed manually. Alternatively, all or part of the processes described as being performed manually can be performed automatically using a known method. In addition, the processing procedures, control procedures, specific names, and information including various data and parameters described above and illustrated can be changed as appropriate unless otherwise specified.

[0086] 11 is a diagram showing an example of a computer in which the classification devices 10, 210, and 310 are realized by executing a program. The computer 1000 includes, for example, a memory 1010 and a CPU 1020. The computer 1000 also includes a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, and a network interface 1070. These components are connected by a bus 1080.

[0087] The memory 1010 includes a ROM 1011 and a RAM 1012. The ROM 1011 stores a boot program such as a BIOS (Basic Input Output System). The hard disk drive interface 1030 is connected to a hard disk drive 1090. The disk drive interface 1040 is connected to a disk drive 1100. A removable storage medium such as a magnetic disk or optical disk is inserted into the disk drive 1100. The serial port interface 1050 is connected to a mouse 1110 and a keyboard 1120, for example. The video adapter 1060 is connected to a display 1130, for example.

[0088] The hard disk drive 1090 stores, for example, an OS 1091, an application program 1092, a program module 1093, and program data 1094. That is, the programs that define the processes of the classification devices 10, 210, and 310 are implemented as program modules 1093 in which code executable by the computer 1000 is written. The program modules 1093 are stored, for example, in the hard disk drive 1090. For example, the program modules 1093 for executing processes similar to the functional configurations of the classification devices 10, 210, and 310 are stored in the hard disk drive 1090. The hard disk drive 1090 may be replaced with an SSD (Solid State Drive).

[0089] Furthermore, setting data used in the processing of the above-described embodiment is stored as program data 1094, for example, in memory 1010 or hard disk drive 1090. Then, CPU 1020 reads out program module 1093 or program data 1094 stored in memory 1010 or hard disk drive 1090 into RAM 1012 as necessary and executes them.

[0090] The program module 1093 and program data 1094 may not necessarily be stored in the hard disk drive 1090, but may also be stored in a removable storage medium and read by the CPU 1020 via the disk drive 1100 or the like. Alternatively, the program module 1093 and program data 1094 may be stored in another computer connected via a network (such as a local area network (LAN) or a wide area network (WAN)). The program module 1093 and program data 1094 may then be read by the CPU 1020 from the other computer via the network interface 1070.

[0091] Although the present invention has been described above as an embodiment, the present invention is not limited to the description and drawings that form part of the disclosure of the present invention. In other words, other embodiments, examples, and operational techniques that can be made by those skilled in the art based on the present invention are all included in the scope of the present invention.

[0092] 10, 210, 310, 310A Classification device 11 Input unit 12 Image feature extraction unit 13 Category feature storage unit 14, 214, 314, 314A Classification unit 15 Output unit 20 Web page image 30 Web page category 312 Feature extraction unit 3122 URL feature extraction unit 3123 HTML feature extraction unit

Claims

1. A classification device comprising at least an input unit that accepts input of a web page image, which is a screen capture image when a website consisting of multiple web pages is displayed; an extraction unit that performs image processing on the web page image and extracts image features of the web page image; and a classification unit that classifies the web page indicated by the web page image based on the image features.

2. The classification device described in claim 1, characterized in that the classification unit classifies the category to which the web page indicated by the web page image belongs based on a first machine learning model that has been trained to classify web pages based on image features extracted from a learning web page image, and the image features.

3. The classification device described in claim 1, characterized in that the classification unit groups the multiple web page images based on the similarity of the multiple image features extracted by the extraction unit, and classifies each group into one category.

4. The classification device described in claim 2, characterized in that the input unit accepts input of the web page image along with ancillary information related to the web page indicated by the web page image, the extraction unit performs image processing on the web page image to extract image features of the web page image and extracts features from the ancillary information that are different from the image features, and the classification unit classifies the category to which the web page indicated by the web page image belongs based on a second machine learning model that has learned to classify web pages based on the image features extracted from the learning web page image and the features extracted from the ancillary information of the learning web page image, the image features, and the features extracted from the ancillary information by the extraction unit.

5. The classification device described in claim 3, characterized in that the input unit accepts input of the Web page image along with auxiliary information related to the Web page indicated by the Web page image, the extraction unit performs image processing on the Web page image to extract image features of the Web page image and extracts features from the auxiliary information that are different from the image features, and the classification unit groups the multiple Web page images based on the similarity between the multiple image features and the multiple features extracted from the auxiliary information by the extraction unit, and classifies each group into a single category.

6. A classification method executed by a classification device, comprising at least the steps of: accepting input of a web page image, which is a screen capture image of a website consisting of multiple web pages when it is displayed; performing image processing on the web page image to extract image features of the web page image; and classifying the web page indicated by the web page image based on the image features.

7. A classification program for causing a computer to execute at least the steps of: accepting input of a web page image, which is a screen capture image when a website consisting of multiple web pages is displayed; performing image processing on the web page image to extract image features of the web page image; and classifying the web page indicated by the web page image based on the image features.

Citation Information

Patent Citations

  • Image retrieving apparatus and method, image retrieving program and recording medium

    JP2006059036A

  • Extraction device, extraction method, and extraction program

    WO2020240718A1

Cited By

  • Method and system for automated website generation

    KR102962461B1