JP6467694B1 - Search term evaluation device, evaluation system, evaluation method, and evaluation module production method - Google Patents

Abstract

The present invention quantitatively evaluates how an upper page of a search result of a search term as a target of SEO changes with time.
An input step in which a search word is input, search result data based on a search word when the first point is set as a searcher's position, and a second different from the first point An acquisition step of acquiring search result data based on a search term when the point is a searcher's position, and an index indicating a correlation between the search result data of the first point and the search result data of the second point And an analysis step for outputting the index as an analysis result specific to the search term.
[Selection] Figure 2

Description

  The present invention evaluates the regional dependency of a search term and a search term group (hereinafter referred to as “search market”) in advance, and performs area marketing and search engine optimization (Search Engine Optimization: hereinafter) based on the regional dependency. , SEO).

  Various technologies for supporting SEO (Search Engine Optimization) have been proposed. There is Patent Document 1 as a document disclosing this kind of technology. The web page analysis apparatus disclosed in Patent Document 1 uses a plurality of web page data in a search result for a target keyword as an analysis target web page when a certain search term is input as a target keyword, and the analysis target web page data The morpheme analysis process is performed, the number of morphemes of the same type in the morpheme group obtained by the morpheme analysis process is totaled, and the degree of contribution of each morpheme to the rank of the analysis target web page in the search results An evaluation value is obtained, and a list in which evaluation values classified by morpheme are arranged for each analysis target web page is presented as an analysis result. According to the technique of Patent Document 1, a morpheme having a high SEO effect can be found efficiently.

Patent 6164436

  By the way, some search terms used as a search query of a search engine include information such as a web page URL and content displayed in a search result which is almost the same regardless of the region where the search query is transmitted. For example, the information may vary greatly depending on which region the search query is sent from. The degree of influence of the search query source region including a certain search term on the search result of the search term on the upper page of the search result is called “region dependency of search term”. Search terms that are strongly region-dependent require more resources for area marketing and SEO, as much as individual analysis that takes into account regional characteristics and customization of the web site is required. For this reason, companies that conduct area marketing or SEO want to quantitatively evaluate the degree of regional dependency of search terms related to their products and use them as an index for their resource allocation. There is a need.

  However, the technique of Patent Document 1 cannot analyze a search result using a region different from a region where the web page analysis device is located as a search query transmission source, and is not suitable for evaluating the region dependency of a search term.

  The present invention has been made in view of such problems, and an object of the present invention is to provide a technical means for quantitatively evaluating the region dependency of a search word used for web search.

  In order to solve the above-described problems, a search word evaluation apparatus according to a preferred aspect of the present invention includes an input unit for inputting a search word, and the search word when the first point is a searcher's position. Obtaining the search result data of the search based on the search term data based on the search term when the searcher's position is a second point different from the first point; Characterized by having an analysis means for obtaining an index indicating a correlation between the search result data of the first point and the search result data of the second point, and outputting the index as an analysis result unique to the search word. Provided is an apparatus for evaluating a search term.

  In this aspect, the search result data is N (N is a natural number of 2 or more) document data in a search result when a search is performed based on the search word from a document data group described in a markup language. The analysis means sets the order of the document data and the second point as the searcher's position when the first point is set as the searcher's position for each of the N document data. The index indicating the correlation may be obtained by applying the order of the document data to a predetermined function.

Further, the analysis means is based on the rank of the i-th document data or the rank of the i-th document data of the N document data when the first point is the position of the searcher. The index is x _i , the average value of the index is xm, and the ranking of the i-th document data or the index based on the ranking when the second point is the position of the searcher is y _i , the index May be calculated according to the function shown in the following equation (1).

  In addition, information indicating the plurality of second points is input to the input unit, and each of the plurality of second points is grouped according to the value of the index ρ calculated by the analysis unit. You may further have a grouping means.

  The search result data is a snippet of N (N is a natural number of 2 or more) document data in a search result when a search based on the search word is performed from a document data group described in a markup language. Snippet analysis data obtained by performing a predetermined analysis process on the document data, and the analysis means snippet of the document data when the first point is the position of the searcher for each of the N document data The index indicating the correlation may be obtained by applying the snippet analysis data of the document data when the analysis data and the second point are set as the position of the searcher to a predetermined function.

In addition, the analyzing unit may analyze the snippet analysis data of the i-th document data or the snippet of the i-th document data of the N document data when the first point is the position of the searcher. The index based on the analysis data is x _i , the average value of the index is xm, and the snippet analysis data of the i-th document data or the snippet analysis data when the second point is the position of the searcher The index based on y _i and the average value of the index may be ym, and the index ρ indicating the correlation may be calculated according to the function shown in the following equation (1).

  A search word evaluation system according to another preferred aspect of the present invention includes a user terminal and an evaluation device communicating with the user terminal, wherein the evaluation device includes an input means for inputting a search word. The search result data based on the search word when the first point is set as the searcher's position is acquired, and the second point different from the first point is set as the searcher's position. Obtaining means for acquiring search result data of a search based on the search term, an index indicating a correlation between the search result data of the first point and the search result data of the second point is obtained, and this index is used as the search Analyzing means for outputting as an analysis result specific to the word.

  The search word evaluation method according to another preferred aspect of the present invention includes an input step in which a search word is input, and search result data of a search based on the search word when the first point is the position of the searcher. An acquisition step of acquiring search result data of a search based on the search word when a second point different from the first point is set as a searcher's position, and a search for the first point And an analysis step of obtaining an index indicating a correlation between the result data and the search result data of the second point, and outputting the index as an analysis result unique to the search word.

  According to another preferred aspect of the present invention, there is provided a search word evaluation apparatus comprising: an input means for inputting a search word; and a search result based on the search word when the first point is the searcher's position. The access analysis data of the access destination in the search result of the search based on the search term in the case where the access analysis data of the access destination is acquired and the second point different from the first point is set as the position of the searcher And obtaining an index indicating a correlation between the access analysis data at the first point and the access analysis data at the second point, and outputting the index as an analysis result specific to the search term. Means.

  In this aspect, the access analysis data is N (N is a natural number of 2 or more) document data in a search result when a search is performed based on the search word from a document data group described in a markup language. The analysis means calculates the click-through rate of the document data and the second point when the first point is the position of the searcher for each of the N pieces of document data. The index indicating the correlation may be obtained by causing the click-through rate of the document data in the case of the searcher's position to act on a predetermined function.

In addition, the analysis unit may be configured to select a click-through rate of the i-th document data or the click for the i-th document data of the N document data when the first point is the position of the searcher. The index based on the slew rate is _xi , the average value of the index is xm, and the click-through rate or the click-through rate of the i-th document data when the second point is the position of the searcher. The index based on y _i and the average value of the index may be ym, and the index ρ indicating the correlation may be calculated according to the function shown in the following equation (1).

  In addition, information indicating the plurality of second points is input to the input unit, and each of the plurality of second points is grouped according to the value of the index ρ calculated by the analysis unit. You may further have a grouping means.

  The access analysis data is an address of N (N is a natural number of 2 or more) document data in a search result when a search is performed based on the search word from a document data group described in a markup language. The analysis means, for each of the N pieces of document data, uses the search time as the searcher's position when the first point is set as the searcher's position. The index indicating the correlation may be obtained by causing the stay time of the document data in the case of the position to act on a predetermined function.

In addition, the analysis unit may be the stay time or the stay time of the i-th document data when the first point is the position of the searcher for the i-th document data of the N document data. The index based on xi is x _i , the average value of the index is xm, and the stay time of the i-th document data or the index based on the stay time when the second point is the position of the searcher is y _i , the average value of the index may be ym, and the index ρ indicating the correlation may be calculated according to the function shown in the following equation (1).

  In addition, information indicating the plurality of second points is input to the input unit, and each of the plurality of second points is grouped according to the value of the index ρ calculated by the analysis unit. You may further have a grouping means.

  The access analysis data is an address of N (N is a natural number of 2 or more) document data in a search result when a search is performed based on the search word from a document data group described in a markup language. The analysis means, for each of the N pieces of document data, uses the first point as the searcher's position and the searcher as the searcher's leave rate and the second point. The index indicating the correlation may be obtained by applying the detachment rate of the document data in the case of the position to a predetermined function.

In addition, the analysis unit may determine whether the i-th document data out of the N pieces of document data has a detachment rate or the detachment rate of the i-th document data when the first point is the position of the searcher. The index based on xi is x _i , the average value of the index is xm, and the detachment rate of the i-th document data or the index based on the detachment rate when the second point is the position of the searcher _i , the average value of the index may be ym, and the index ρ indicating the correlation may be calculated according to the function shown in the following equation (1).

  In addition, information indicating the plurality of second points is input to the input unit, and each of the plurality of second points is grouped according to the value of the index ρ calculated by the analysis unit. You may further have a grouping means.

  The access analysis data includes the addresses of the document data (N is a natural number of 2 or more) in the search result when a search is performed based on the search word from the document data group described in the markup language. The analysis means calculates the number of inflows of the document data and the second point of each of the N pieces of document data when the first point is the position of the searcher. The index indicating the correlation may be obtained by causing the inflow number of the document data in the case of the position to act on a predetermined function.

In addition, the analysis unit may include the number of inflows of the i-th document data or the number of inflows of the i-th document data among the N pieces of document data when the first point is the position of the searcher. X _i is an index based on x, an average value of the index is x m, and an inflow number of the i-th document data or an index based on the inflow number when the second point is the position of the searcher _i , the average value of the index may be ym, and the index ρ indicating the correlation may be calculated according to the function shown in the following equation (1).

  In addition, information indicating the plurality of second points is input to the input unit, and each of the plurality of second points is grouped according to the value of the index ρ calculated by the analysis unit. You may further have a grouping means.

  The access analysis data is an address of N (N is a natural number of 2 or more) document data in a search result when a search is performed based on the search word from a document data group described in a markup language. The analysis means calculates the page number of the document data and the second point when the first point is the position of the searcher for each of the N pieces of document data. The index indicating the correlation may be obtained by causing the number of browse pages of the document data in the case of the position of the searcher to act on a predetermined function.

In addition, the analysis unit may be configured to display the number of pages viewed in the i-th document data or the browsing in the i-th document data out of the N document data when the first point is set as the position of the searcher. The index based on the number of pages is x _i , the average value of the index is xm, and the number of pages to be browsed or the number of pages to be browsed of the i-th document data when the second point is the position of the searcher. The index based on y _i and the average value of the index may be ym, and the index ρ indicating the correlation may be calculated according to the function shown in the following equation (1).

  In addition, information indicating the plurality of second points is input to the input unit, and each of the plurality of second points is grouped according to the value of the index ρ calculated by the analysis unit. You may further have a grouping means.

  A search word evaluation system according to another preferred aspect of the present invention includes a user terminal and an evaluation device communicating with the user terminal, wherein the evaluation device includes an input means for inputting a search word. The access analysis data of the access destination in the search result based on the search word when the first point is set as the searcher's position is acquired, and the second point different from the first point is searched. Acquisition means for acquiring access analysis data of the access destination in the search result based on the search term when the position is set as a person's position, access analysis data of the first point, and access analysis data of the second point And an analysis unit that obtains an index indicating a correlation with the search term and outputs the index as an analysis result unique to the search term.

  According to another preferred aspect of the present invention, there is provided a search word evaluation method comprising: an input step in which a search word is input; and a search result based on the search word when the first point is the searcher's position. The access analysis data of the access destination in the search result of the search based on the search term in the case where the access analysis data of the access destination is acquired and the second point different from the first point is set as the position of the searcher And obtaining an index indicating a correlation between the access analysis data at the first point and the access analysis data at the second point, and outputting the index as an analysis result specific to the search term. And a step.

  Another evaluation mode production method according to the present invention includes a neural network including an input layer, an intermediate layer, and an output layer, and performs a search for a predetermined search word with a plurality of points as positions of a searcher. When the search result data group is applied to the neural network, the output signal output from the neural network or the output signal and the teacher signal is acquired, and the acquired signal is processed according to a deep learning algorithm. And a learning evaluation unit that updates weighting factors between the input layer and the intermediate layer and between the intermediate layer and the output layer based on the processing result, and the weighting factor of the learning evaluation unit An evaluation that evaluates the relationship between the search needs of the first point and the search point of the second point for the search terms for the learned model that has undergone the update And outputs as Joules.

  Another evaluation mode production method according to the present invention includes a neural network including an input layer, an intermediate layer, and an output layer, and performs a search for a predetermined search word with a plurality of points as positions of a searcher. When the access analysis data group of the access destination in the search result when applied is input to the neural network, the output signal output from the neural network, or the output signal and the teacher signal are acquired, and the acquired signal is Processing according to a deep learning algorithm, and based on the processing result, a learning evaluation unit that updates weight coefficients between the input layer and the intermediate layer and between the intermediate layer and the output layer, and The learned model that has undergone the updating of the weighting coefficient by the learning evaluation unit is used to search for the first point and the second point for the search term. And outputs as an evaluation module for evaluating the relationship between the over's.

  ADVANTAGE OF THE INVENTION According to this invention, the quantitative evaluation of the area dependence of a search market can be enabled. Therefore, according to the present invention, a company that conducts area marketing or SEO distinguishes between a product field that needs to be dealt with in consideration of regional characteristics and a product field that does not, and allocates more realistic resources. be able to.

It is a figure which shows the whole structure of the evaluation system 1 containing the evaluation apparatus 20 which is 1st Embodiment of this invention. 4 is a flowchart showing a flow of an evaluation method executed by a CPU 22 of the evaluation device 20 according to an evaluation program 26. It is a figure for demonstrating the input processing of the evaluation apparatus. It is a figure for demonstrating the division of the area in the evaluation apparatus. It is a figure for demonstrating the acquisition process and analysis process of the evaluation apparatus. It is a figure which shows the whole structure of 1 A of evaluation systems containing 20 A of evaluation apparatuses which are 2nd Embodiment of this invention. It is a figure which shows the whole structure of the evaluation system 1B containing the evaluation apparatus 20B which is 3rd Embodiment of this invention. It is a figure which shows the whole structure of 1 C of evaluation systems containing the evaluation apparatus 20C which is 4th Embodiment of this invention. It is a figure for demonstrating the acquisition process and analysis process of 20 C of evaluation apparatuses. It is a figure which shows the whole structure of evaluation system 1D containing evaluation apparatus 20D which is 5th Embodiment of this invention. It is a figure which shows the whole structure of the evaluation system 1E containing the evaluation apparatus 20E which is 6th Embodiment of this invention. It is a figure for demonstrating the acquisition process and analysis process of the evaluation apparatus. It is a figure for demonstrating the acquisition process and analysis process of 20 C of evaluation apparatuses. It is a figure for demonstrating the acquisition process and analysis process of 20 C of evaluation apparatuses. It is a figure for demonstrating the acquisition process and analysis process of 20 C of evaluation apparatuses. It is a figure for demonstrating the acquisition process and analysis process of 20 C of evaluation apparatuses. It is a figure explaining operation | movement of the evaluation apparatus which is a modification of this invention. It is a figure explaining operation | movement of the evaluation apparatus which is a modification of this invention.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
<First Embodiment>
FIG. 1 is a diagram showing an overall configuration of an evaluation system 1 including an evaluation apparatus 20 according to the first embodiment of the present invention. The evaluation system 1 shown in FIG. 1 is a group of document data (data described in a markup language such as HTML (Hyper Text Markup Language)) scattered on the Internet 90 as a web site, web page, or web content. Therefore, it is possible to quantitatively evaluate the region dependence of the search result when a search based on a certain search word is performed.

  As shown in FIG. 1, the evaluation system 1 includes a user terminal 10 and an evaluation device 20. The user terminal 10 and the evaluation device 20 are connected via the Internet 90. A search engine server device 50 is connected to the Internet 90.

  The search engine server device 50 is a device that plays a role of providing a search engine service. The search engine server device 50 circulates the Internet 90 and performs a patrol process for indexing information obtained from the web site, web page, and web content of the patrol destination into the index database, and a search word and position information from the searcher's computer. Of an HTTP (Hyper Text Transfer Protocol) request (search query) including, and a web site, a web page, and a web content obtained when a web search is performed using the search word at the position indicated by the position information. A search process for returning search results in which a set of title, URL (Uniform Resource Locator), and snippet (Snippet) is arranged in order from the top (highest) is performed. Here, in the search process, the search engine server device 50 responds to the position of the searcher indicated by the position information given together with the search word, in the search result, the web site, the web page, or the web content in the search result. To change. Although only one search engine server device 50 is illustrated in FIG. 1, the number of search engine server devices 50 may be plural.

  The user terminal 10 is a personal computer. The user of the user terminal 10 is given a unique ID and password. The user accesses the evaluation device 20 from his / her user terminal 10 to perform an authentication procedure, and uses a service provided by the evaluation device 20. Although only one user terminal 10 is illustrated in FIG. 1, the number of user terminals 10 in the evaluation system 1 may be plural.

  The evaluation device 20 is a device that plays a role of providing a region-dependent quantitative evaluation service for a search term. The region-dependent quantitative evaluation service for search terms receives a search term and two different points (hereinafter referred to as a first point and a second point) that serve as a reference when evaluating the region dependency. When the search is performed based on the search word with the first point as the position of the searcher, the ranking of the web site, the web page, and the web content in the search result and the search point with the second point as the position of the searcher This is a service for obtaining an index indicating a correlation with the ranking of a web site, a web page, and a web content in a search result when a search based on is performed, and presenting this index to a user as an analysis result unique to the search term.

  As shown in FIG. 1, the evaluation apparatus 20 includes a communication interface 21, a CPU (Central Processing Unit) 22, a RAM (Random Access Memory) 23, a ROM (Read Only Memory) 24, and a hard disk 25. The communication interface 21 transmits and receives data to and from devices connected to the Internet 90. The CPU 22 executes various programs stored in the ROM 24 and the hard disk 25 while using the RAM 23 as a work area. The ROM 24 stores an IPL (Initial Program Loader) and the like. The hard disk 25 stores an evaluation program 26 having functions specific to the present embodiment.

Next, the operation of this embodiment will be described.
FIG. 2 is a flowchart showing the flow of the evaluation method executed by the CPU 22 of the evaluation apparatus 20 according to the evaluation program 26. The CPU 22 functions as an input unit that executes the input process (step SA100), an acquisition unit that executes the acquisition process (step SA110), and an analysis unit that executes the analysis process (step SA120) by executing the evaluation program 26. To do.

  In the input process of step SA100 in FIG. 2, the CPU 22 transmits web page data in which the structure of the input screen is described in HTML, as shown in FIG. 3, to the user terminal 10 that has completed the authentication procedure. When the user terminal 10 receives the web page data, the user terminal 10 displays an input screen. On the input screen, a character string “Please enter a search term” is displayed. Below that, there is a search term input field 51. In the search word input field 51, the search word to be evaluated is input by the user. Under the search word input field 51, characters “Please specify the first point” are displayed. Below that, there is a point input field 52. In the point input field 52, one of two points serving as a reference when evaluating the region dependency of the search word is input by the user as information indicating the first point. A character string “Please specify a second point” is displayed below the point input field 52. Below that, there is a point input field 53. In the point input field 53, as the second point, the other one of the two points serving as a reference when evaluating the regional dependency of the search word is input by the user.

  The input procedure in the point input fields 52 and 53 will be described in more detail. In this embodiment, as shown in FIG. 4, Japan is divided into five regions (“Hokkaido”, “Tohoku”, “Kanto”, “Kansai”, For each of the representative points Rk (k = 1 to n) of each region when divided into five regions of “Kyushu”), a character string indicating the position information of the representative point Rk and the region name of the representative point Rk (“ A table in which “Hokkaido”, “Tohoku”, “Kanto”, “Kansai”, and “Kyushu” are stored in association with each other is embedded in the web page data. As specific examples of the position information, geographical information indicating the latitude and longitude of the government office of the prefecture with the largest population in the area indicated by the representative point Rk, and information that can be converted into the geographical information (for example, Or a character string indicating the name of the government office of the prefecture).

  When the user terminal 10 is operated by moving the mouse pointer to the point input field 52 and left-clicking, the pull-down list in which the above five region names are arranged appears to prompt the user to select a region name. When any of the region names is selected, the selected region name is displayed in the point input field 52, and the position information of the representative point Rk stored in the table in association with the region name is acquired. In addition, when the user terminal 10 is operated by moving the mouse pointer to the point input field 53 and left-clicking, the user terminal 10 prompts the user to select a region name by causing a pull-down list in which the above five region names are arranged to appear. When any region name in the list is selected, the selected region name is displayed in the point input field 53, and the position information of the representative point Rk stored in the table in association with the region name is acquired. To do.

  When the same position information as that of the first point is input as the position information indicating the second point, a message prompting the user to input again may be output to the user terminal 10. Also, in order to avoid the same position information as the first point being input, a pull-down list excluding the area name specified as the first point when the pull-down list for specifying the second point is displayed. May be displayed.

  A character string “Please enter the analysis range” is displayed below the point input field 53. The analysis range is an analysis of the upper rank in the search result when a web search based on the search word in the search word input field 51 is performed using the position indicated by the position information in the point input field 52 as the position of the searcher. It means to target. Below the character string “Please enter the analysis range”, there is an analysis range input field 54. In the analysis range input field 54, a number N (N is a natural number of 2 or more) indicating the analysis range is input by the user. The analysis range input field 54 may select one of a plurality of options such as the top 10 place, the top 20 place, and the top 30 place. Below the analysis range input field 54 is a send button 55.

  The user inputs a search word in the search word input field 51 of the input screen, inputs information indicating the first point in the point input field 52, inputs information indicating the second point in the point input field 53, The number N indicating the analysis range is input to the analysis range input field 54 and the transmission button 55 is pressed. When the send button 55 is pressed, the user terminal 10 uses the information input in the search word input field 51, the point input field 52, the point input field 53, and the analysis range input field 54 as input information. It transmits to the evaluation apparatus 20. When the CPU 22 receives input information from the user terminal 10 via the Internet 90, the CPU 22 writes the received input information in the RAM 23.

Next, the CPU 22 executes an acquisition process (step SA110). As shown in FIG. 2, the acquisition process includes Step SA110a and Step SA110b. In step SA110a, the CPU 22 performs the search based on the search word in the input information at the first point (the position indicated by the information in the point input field 52), the top N-level web site in the search result, the web page. Or Web content as an analysis target, and the search result data A _i (i = 1 to N, i is a document to be analyzed) of the order of N web sites, web pages, or web content to be analyzed As an index indicating each piece of data).

In step SA110b, the CPU 22 sets the ranking of the N web sites, web pages, or web contents to be analyzed at the second point (the position indicated by the information in the point input field 52) as search result data for the second point. Obtained as B _i (i = 1 to N). Here, the web site, web page, or web content that was within the top N positions in the search result of the first point search may be out of range in the search result of the second point search. Such web site, web page, or web content search result data _Bi may be a sufficiently large value (for example, 1000).

In the analysis process of step SA120, the CPU 22 indicates the order indicated by the first location search result data A _i (i = 1 to N) and the second location search result data B _i (i = 1 to N). By applying the rank to the function of the following expression (1), an index ρ indicating the correlation between the two is obtained, and this index ρ is output as an analysis result specific to the search word in the input information.

Here, x _i in the expression (1) is an index associated with the search result data A _i in the table TBL 1 of the ROM 24. The table TBL1 stores each pair of ranking and index. The index of the table TBL1 has a larger value as it corresponds to a higher rank. xm is an average value (additional average) of the indices x _i corresponding to each of the search result data A _i (i = 1 to N). Y _i is an index associated with the search result data B _i in the table TBL1. ym is an average value (additional average) of the indices y _i corresponding to each of the search result data B _i (i = 1 to N).

  The acquisition process and the analysis process will be described in more detail. As shown in the example of FIG. 5, the CPU 22 determines that the information in the search word input field 51 on the input screen is “insurance” and the information in the point input field 52 is “Kanto ", If the information in the point input field 53 is" Hokkaido "and the information in the analysis range input field 54 is" 10 ", the search based on the search term" insurance "when Tokyo is the position of the searcher The search based on the same search term “insurance” when Hokkaido is the position of the searcher is executed, and the top 10 web sites, web pages, and web contents in the search results in Tokyo are analyzed.

The CPU 22 sets the ranking of the 10 web sites, web pages, and web contents to be analyzed in the search results in Tokyo as search result data A ₁ , A ₂ ... A _10, and the analysis target in the search results in Hokkaido. The rankings of the ten web sites, web pages, and web contents are set as search result data B ₁ , B ₂ ... B ₁₀ . The CPU 22 retrieves the index x ₁ , x ₂ ... X ₁₀ corresponding to the search result data A ₁ , A ₂ ... A ₁₀ and the search result data B ₁ , B _{2. The} indices y ₁ , y ₂ ... Y ₁₀ corresponding to ₁₀ are taken out. Furthermore, CPU 22 obtains an average ym average xm, and the index _{y 1, y} 2 ··· _{y 10} of the index _{x 1, x 2 ··· x 10} .

Then, the CPU 22 obtains the index ρ by inputting the indices x ₁ , x ₂ ... X ₁₀ and the average mx and the indices y ₁ , y ₂ ... Y ₁₀ and the average ym into the equation (1). The index ρ is transmitted to the user terminal 10 as an analysis result unique to the search term “insurance”.

  The user of the user terminal 10 that is the transmission source of the input information can quantitatively evaluate the regional dependence of the search term to be evaluated using the index ρ returned from the evaluation device 20. For example, for a certain search word X, the representative point Rk (any one of k = 1 to 5) in FIG. 4 is the first point, and the representative point Rm (m = 1 to 5: where m ≠ k) is the first point. If the average of the index ρkm calculated in accordance with the above equation (1) is ρm in the case of 2 points, the strength of region dependence can be quantitatively evaluated by the magnitude of the value of ρm. Specifically, if the value of ρm is close to 1, it means that the web page face obtained as a search result by the search word X and the search order of each web page are substantially constant regardless of the region. X's regional dependency is evaluated as low. Conversely, the smaller the value of ρm (for example, close to 0), the web page face obtained as a search result by the search word X and the search order of each web page mean that it is different for each region. The region dependency of the search word X is evaluated to be high.

  Also, the value of the index ρ calculated for each of the plurality of second points is compared, and the point where the value of the index ρ is close to 1, that is, the web page face included in the search result and the search order of each web page By extracting a point similar to the first point, it is possible to estimate a trade area that will have the same product needs as the first point. For example, if the indices ρ13 and ρ23 in FIG. 4 are close to 1 and ρ34 and ρ35 are small values, it is estimated that the products that will be searched by the search word X have a common need in eastern Japan. can do.

  In addition, as shown in FIG. 4, each of the representative points Rk (k = 1 to 5) of each region of “Hokkaido”, “Tohoku”, “Kanto”, “Kansai”, and “Kyushu” is used as a node, and an index The network is modeled by a weighted undirected graph obtained by connecting each node with a link weighted by ρkm, and existing community extraction algorithms such as random walk, greedy method, and spin glass method are applied to this network model. It is also possible to perform clustering based on the similarity of search ranks in search terms having region dependency, thereby clustering trade areas having common needs for products and the like.

  As described above, according to the evaluation device 20 of the present embodiment, it is possible to quantitatively evaluate the region dependence of the search market. Therefore, according to this embodiment, a company that performs area marketing or SEO distinguishes the product field that is required to respond more strongly considering regional characteristics and the product field that does not, and allocates more realistic resources. It can be carried out.

Second Embodiment
FIG. 6 is a diagram showing an overall configuration of an evaluation system 1A including an evaluation apparatus 20A that is the second embodiment of the present invention. The evaluation system 1A uses a search result area when a search is performed based on a search word from a group of document data collected in the document management server device 30 from a user terminal 10 in a corporate LAN (Local Area Network) 80. It enables quantitative evaluation of dependence.

  In FIG. 6, the user terminal 10 and the evaluation device 20 </ b> A of the evaluation system 1 </ b> A are connected to the document management server device 30 via the LAN 80. The document management server device 30 is a device that plays a role of providing a document management service. The document management server device 30 receives document data from the user terminal 10 in the LAN 80, stores the received document data in the document database 31, and receives a search query from the user terminal 10 at various positions. Then, the document data related to the search term in the search query is searched from the document database 31, and a search result in which the set of title, URL, and snippet (snippet) of the searched document data is arranged in order from the top (highest ranking) And search processing to reply.

The evaluation device 20A of the evaluation system 1A executes the input process, the acquisition process, and the analysis process shown in FIG. 2 by the function of the evaluation program 26 in the hard disk 25. The contents of these processes are as follows: the first location search result data A _i (i = 1 to N) and the second location search result data B _i (i = 1 to N) are acquired from the document management server device. The third embodiment is the same as the first embodiment except for 30 document databases 31. According to this embodiment, the same effect as the first embodiment can be obtained.

<Third Embodiment>
FIG. 7 is a diagram showing an overall configuration of an evaluation system 1B including an evaluation apparatus 20B that is the third embodiment of the present invention. Similar to the evaluation system 1 of the first embodiment, the evaluation system 1C can quantitatively evaluate the regional dependence of the search result when a search is performed based on a search term from a group of document data scattered on the Internet 90. It is to make.

  In FIG. 7, the user terminal 10 and the evaluation device 20 </ b> A of the evaluation system 1 </ b> A are connected to the search result collection server device 40 via the LAN 80. The search result collection server device 40 is a device that plays a role of providing a search result collection service. The search result collection server device 40 transmits a search query of various search terms having various points as searcher positions to the search engine server device 50 at predetermined time intervals (for example, every 10 minutes). A search result is received from the engine server device 50, and a search term, a search date, and a set of search results are stored in the search result database 41.

The evaluation device 20B of the evaluation system 1B executes the input process, the acquisition process, and the analysis process shown in FIG. 2 by the function of the evaluation program 26 in the hard disk 25. The content of these processes is that the search result data A _i (i = 1 to N) of the first point and the search result data B _i (i = 1 to N) of the second point are the search result collection server. Except for the fact that it is the search result database 41 of the device 40, it is the same as in the first embodiment. According to this embodiment, the same effect as the first and second embodiments can be obtained.

<Fourth embodiment>
FIG. 8 is a diagram showing an overall configuration of an evaluation system 1C including an evaluation apparatus 20C that is the fourth embodiment of the present invention. Similar to the evaluation system 1 of the first embodiment, the evaluation system 1C can quantitatively evaluate the regional dependence of the search result when a search is performed based on a search term from a group of document data scattered on the Internet 90. It is to make.

  In FIG. 8, a search engine server device 50 and an access analysis server device 60 are connected to the Internet 90. The access analysis server device 60 is a device that plays a role of providing an access analysis service. The access analysis server device 60 receives an HTTP request including the URL of the document data subject to access analysis on the Internet 90 from the user terminal 10, and the click-through rate, stay time, and withdrawal rate of the URL of the document data subject to access analysis Information such as the number of inflows and the number of browsed pages is returned as an access analysis result. Here, when the access analysis server device 60 receives an HTTP request including the URL of the document data to be analyzed for access and the location of the searcher, it returns an access analysis result at that location.

  The evaluation device 20C of the evaluation system 1C executes the input process, the acquisition process, and the analysis process shown in FIG. 2 by the function of the evaluation program 26 in the hard disk 25. Among these processes, the contents of the input process are the same as in the first embodiment.

In the present embodiment, in the acquisition process, the CPU 22 of the evaluation device 20C includes a search result when a search is performed based on a search word in the input information at the first point (the position indicated by the information in the point input field 52). The top N web sites, web pages, or web contents are targeted for analysis, and the click-through rate of the N web sites, web pages, or web contents to be analyzed is acquired from the access analysis server device 60, and these clicks are made. The slew rate is set as access analysis data C _i (i = 1 to N) of the access destination in the search result of the search for the first point. In addition, the CPU 22 acquires, from the access analysis server device 60, the click-through rate of N web sites, web pages, or web contents to be analyzed at the second point (the position indicated by the information in the point input field 52). These click-through rates are set as access analysis data D _i (i = 1 to N) of the access destination in the search result of the second point search.

In the analysis processing, the CPU 22 clicks the click-through rate indicated by the first point access analysis data C _i (i = 1 to N) and the second point access analysis data D _i (i = 1 to N). By causing the slew rate to act on the function of the above equation (1), an index ρ indicating the correlation between the two is obtained, and this index ρ is output as an analysis result specific to the search word in the input information.

  The acquisition process and the analysis process of this embodiment will be described in more detail. As shown in the example of FIG. 9, the CPU 22 indicates that the information in the search word input field 51 of the input screen is “insurance” and the point input field 52 If the information is “Kanto”, the information in the point input field 53 is “Hokkaido”, and the information in the analysis range input field 54 is “10”, the search term “insurance” when Tokyo is the position of the searcher. And the search based on the same search term “insurance” when Hokkaido is the position of the searcher, and the top 10 web sites, web pages, and web content in the search results of Tokyo search Is the analysis target.

When the position of the searcher is Tokyo, the CPU 22 sets the click-through rates of the 10 web sites, web pages, and web contents to be analyzed in the search result as access analysis data C ₁ , C ₂ ... C _10. , The click-through rates of the 10 web sites, web pages, and web contents to be analyzed in the search results when Hokkaido is the position of the searcher are set as access analysis data D ₁ , D ₂ ... D ₁₀ . . The CPU 22 reads indexes x ₁ , x ₂ ... X ₁₀ corresponding to the access analysis data C ₁ , C ₂ ... C ₁₀ and access analysis data D ₁ , D _{2. The} indices y ₁ , y ₂ ... Y ₁₀ corresponding to ₁₀ are taken out. Furthermore, CPU 22 obtains an average ym average xm, and the index _{y 1, y} 2 ··· _{y 10} of the index _{x 1, x 2 ··· x 10} .

Then, the CPU 22 obtains the index ρ by inputting the indices x ₁ , x ₂ ... X ₁₀ and the average mx and the indices y ₁ , y ₂ ... Y ₁₀ and the average ym into the equation (1). The index ρ is transmitted to the user terminal 10 as an analysis result specific to the search term “insurance”. According to this embodiment, the same effect as the first to third embodiments can be obtained.

<Fifth Embodiment>
FIG. 10 is a diagram showing an overall configuration of an evaluation system 1D including an evaluation apparatus 20D that is the fifth embodiment of the present invention. In FIG. 10, the user terminal 10 and the evaluation device 20 </ b> D of the evaluation system 1 </ b> D are connected to the document management server device 30 via the LAN 80. The document management server device 30 receives a request including the storage address of the document data subject to access analysis in the document database 31 from the user terminals 10 at various positions, and the click-through rate of the document data subject to access analysis. Information such as stay time, departure rate, number of inflows, number of browsed pages, etc. is returned as an access analysis result.

The evaluation device 20D of the evaluation system 1D executes the input process, the acquisition process, and the analysis process shown in FIG. 2 by the function of the evaluation program 26 in the hard disk 25. The contents of these processes are as follows: the first location access analysis data C _i (i = 1 to N) and the second location access analysis data D _i (i = 1 to N) are acquired from the document management server device. Except for the point of 30, it is the same as the fourth embodiment. According to this embodiment, the same effect as the first to fourth embodiments can be obtained.

<Sixth Embodiment>
FIG. 11 is a diagram showing an overall configuration of an evaluation system 1E including an evaluation apparatus 20E that is the sixth embodiment of the present invention. In FIG. 11, the user terminal 10 and the evaluation device 20E of the evaluation system 1E are connected to the access analysis result collection server device 45 via the LAN 80.

  The access analysis result collection server device 45 is a device that plays a role of providing an access analysis result collection service. The access analysis result collection server device 45 sends HTTP requests including URLs of various document data having various points as searchers to the access analysis server device 60 at predetermined time intervals (for example, every 10 minutes). And receives an access analysis result (information such as a click-through rate, a staying time, a leaving rate, an inflow number, and a number of browse pages) from the access analysis server device 60, and sets the URL of the document data, the analysis date and time, and the analysis result Is stored in the access analysis result database 46.

The evaluation device 20E of the evaluation system 1E executes the input process, the acquisition process, and the analysis process shown in FIG. 2 by the function of the evaluation program 26 in the hard disk 25. The contents of these processes are that the access analysis data C _i (i = 1 to N) of the first point and the access analysis data D _i (i = 1 to N) of the second point are collected by the access analysis result. Except for the fact that it is the access analysis result database 46 of the server device 45, it is the same as in the fourth embodiment. According to this embodiment, the same effect as the first to fifth embodiments can be obtained.

<Modification>
Although the first to sixth embodiments of the present invention have been described above, the following modifications may be added to this embodiment.

(1) In the first to sixth embodiments, the user designates only one second point for one first point. However, the user designates a plurality of second points. The index ρ may be calculated for each of the designated second points. In this case, the evaluation device 20 may be provided with grouping means for grouping a plurality of second points according to the value of the index ρ. Specific examples of such grouping means include grouping the group into a group in which the value of the index ρ exceeds a predetermined threshold (for example, 0.5) and a group in which the value of the index ρ is equal to or less than the threshold, A means for grouping a plurality of second points using the clustering described above can be mentioned.

(2) In the above first to sixth embodiments, two points that serve as the criteria for evaluation of region dependency for the search term to be evaluated are specified in units of regions in Japan. It may be specified in units of towns and villages. For example, the first point is specified in units of municipalities, and as the second point, the municipality adjacent to the first point and the other municipalities further adjacent to the adjacent municipality are sequentially specified. Thus, by calculating the index ρ, it is possible to quantitatively evaluate to what extent the needs for products and the like are spread in common with the first point, centering on the first point. In addition, the designation of the two points that serve as the criteria for evaluating the regional dependence of the search terms to be evaluated is not limited to designation in administrative units such as regions and prefectures in Japan. It may be specified in units. Further, each of the first and second points may be designated by latitude and longitude geographic information.

(3) In the first to sixth embodiments, the user of the user terminal 10 is designated as both the first point and the second point. However, either the first point or the second point is specified. Is the location of the user terminal 10 (that is, the current position of the searcher), the other designation is sufficient. That is, it is not essential for the user to designate both the first and second points, and any mode is acceptable as long as at least one of the first point and the second point is designated by the user.

(4) In the first to third embodiments, the CPU 22 of the evaluation device 20, 20A, 20B, in the acquisition process (step SA110 in FIG. 2), N document data to be analyzed (web site, web page, Or web content), the ranking of each of the N pieces of document data when the first point is set as the position of the searcher is acquired as search result data A _i (i = 1 to N) of the first point, The ranking of the N pieces of document data when the second point is the searcher's position is acquired as the search result data B _i (i = 1 to N) of the second point. However, for N document data to be analyzed, the first snippet analysis data obtained by performing a predetermined analysis process on each snippet of the N document data when the first point is the position of the searcher is the first. Obtained as search result data A _i (i = 1 to N) at the point of, and obtained by subjecting each snippet of N document data to a predetermined analysis process when the second point is the position of the searcher. The snippet analysis data may be acquired as search result data B _i (i = 1 to N) of the second point. As shown in the example of FIG. 12, in this modified example, in the analysis process, the table TBL3 in which the snippet analysis data in the ROM 24 is associated with the index is referred to, and the search result data A _i is converted into the index x _i and searched. It is preferable to perform conversion from the result data B _i to the index y _i . The contents of the input process, the acquisition process, and the analysis process in this modified example are the first to the first except that the snippet analysis data at the first point and the second point are set as the search result data A _i and B _i . It may be the same as in the third embodiment.

(5) In the fourth to sixth embodiments, the CPU 22 of the evaluation devices 20C, 20D, and 20E performs N pieces of document data (web site, web page, web page) to be analyzed in the acquisition process (step SA110 in FIG. 2). Or the web content), the click-through rate of each URL of the N pieces of document data when the first point is the position of the searcher is the access analysis data C _i (i = 1 to N) of the first point. And the click-through rate of each URL of the N document data when the second point is the searcher's position is acquired as the access analysis data D _i (i = 1 to N) of the second point. did. However, for the N pieces of document data to be analyzed, the stay time of each URL of the N pieces of document data when the first place is the position of the searcher is the access analysis data C _i (i of the first place). = 1 to N), and the stay time of each URL of the N document data when the second point is the searcher's position is the access analysis data D _i (i = 1 to 1) of the second point. N) may be obtained. As shown in the example of FIG. 13, in this modified example, in the analysis process, the table TBL4 in which the stay time in the ROM 24 is associated with the index is referred to, and the conversion from the access analysis data C _i to the index x _i and the access analysis are performed. Conversion from the data D _i to the index y _i may be performed. The contents of the input process, the acquisition process, and the analysis process in this modified example are the fourth to sixth except that the stay time at the first point and the second point is the access analysis data C _i and D _i . It may be similar to the embodiment.

(6) In the fourth to sixth embodiments, the CPU 22 of the evaluation devices 20C, 20D, and 20E performs N pieces of document data (web site, web page, analysis target) in the acquisition process (step SA110 in FIG. 2). Or the web content), the click-through rate of each URL of the N pieces of document data when the first point is the position of the searcher is the access analysis data C _i (i = 1 to N) of the first point. And the click-through rate of each URL of the N document data when the second point is the searcher's position is acquired as the access analysis data D _i (i = 1 to N) of the second point. did. However, for the N document data to be analyzed, when the first point is the position of the searcher, the URL withdrawal rate of each of the N pieces of document data is expressed as the first point access analysis data C _i (i = 1 to N), and when the second point is set as the position of the searcher, the URL departure rate of each of the N pieces of document data is obtained as the second point access analysis data D _i (i = 1 to 1). N) may be obtained. As shown in the example of FIG. 14, in this modification, in the analysis process, the table TBL5 in which the withdrawal rate in the ROM 24 is associated with the index is referred to, conversion from the access analysis data C _i to the index x _i and access analysis Conversion from the data D _i to the index y _i may be performed. The contents of the input process, the acquisition process, and the analysis process in this modified example are the fourth to sixth except that the access analysis data C _i and D _i are the leaving rates of the first point and the second point. It may be similar to the embodiment.

(7) In the fourth to sixth embodiments, the CPU 22 of the evaluation devices 20C, 20D, and 20E performs N pieces of document data (web site, web page, analysis target) in the acquisition process (step SA110 in FIG. 2). Or the web content), the click-through rate of each URL of the N pieces of document data when the first point is the position of the searcher is the access analysis data C _i (i = 1 to N) of the first point. And the click-through rate of each URL of the N document data when the second point is the searcher's position is acquired as the access analysis data D _i (i = 1 to N) of the second point. did. However, for the N pieces of document data to be analyzed, the number of URL inflows of each of the N pieces of document data when the first place is the position of the searcher is the access analysis data C _i (i = 1 to N), and the number of URL inflows of each of the N document data when the second point is the searcher's position is the access analysis data D _i (i = 1 to 1) of the second point. N) may be obtained. As shown in the example of FIG. 15, in this modified example, in the analysis process, the table TBL6 in which the number of inflows in the ROM 24 is associated with the index is referred to, conversion from the access analysis data C _i to the index x _i and access analysis Conversion from the data D _i to the index y _i may be performed. The contents of the input process, the acquisition process, and the analysis process in this modification are the fourth to sixth except that the numbers of inflows at the first point and the second point are the access analysis data C _i and D _i . It may be similar to the embodiment.

(8) In the fourth to sixth embodiments, the CPU 22 of the evaluation devices 20C, 20D, and 20E performs N document data (web site, web page, analysis target) in the acquisition process (step SA110 in FIG. 2). Or the web content), the click-through rate of each URL of the N pieces of document data when the first point is the position of the searcher is the access analysis data C _i (i = 1 to N) of the first point. And the click-through rate of each URL of the N document data when the second point is the searcher's position is acquired as the access analysis data D _i (i = 1 to N) of the second point. did. However, for the N document data to be analyzed, the number of browse pages of each URL of the N document data when the first point is the position of the searcher is the access analysis data C _i ( i = 1 to N), and the number of browse pages of each URL of the N document data when the second point is the position of the searcher is the access analysis data D _i (i = 1-N). As shown in the example of FIG. 16, in this modification, in the analysis process, the table TBL6 in which the number of browse pages in the ROM 24 is associated with the index is referred to, and conversion and access from the access analysis data C _i to the index x _i are performed. The analysis data D _i may be converted into the index y _i . The contents of the input process, the acquisition process, and the analysis process in this modified example are the fourth to fourth except that the number of browsing pages at the first point and the second point is the access analysis data C _i and D _i . It may be the same as in the sixth embodiment.

(9) In the above first to third embodiments, each of the rank itself of N document data in the case where the first point to the searcher's location, the index x _{i (i} = 1 based on the ranking N), and the ranking of each of the N pieces of document data when the second point is the position of the searcher may be an index y _i (i = 1 to N) based on the ranking. Similarly, when the snippet analysis data is set as the search result data A _i and B _i , the snippet analysis data of the first point and the second point of each of the N pieces of document data are based on the snippet analysis data. The indices x _i (i = 1 to N) and y _i (i = 1 to N) may be used.

In addition, when the first point is the position of the searcher, the number obtained by converting the rank of each of the N pieces of document data into a logarithm (log) or a reciprocal (1 / rank) is an index x _i ( i = 1 to N), and the number obtained by converting the rank of each of the N pieces of document data into a logarithm (log) or a reciprocal (1 / rank) when the second point is the position of the searcher. It is good also as parameter | index _yi based on (i = 1-N). Similarly, when the snippet analysis data is set as the search result data A _i and B _i , the snippet analysis data at the first point and the second point of each of the N pieces of document data are logarithmically (log) or reciprocal (1 / the number of converted into snippets analysis data), or as an index based on the snippet analysis data _x i (i = _1~N) and _y i (i = 1~N).

Also, each of the number obtained by multiplying a weighting factor to each of the ranking of the N of the N document data in the case where the first point to the searcher's location, the index based on the rank x _{i (i} = 1~N ) And the number obtained by multiplying each rank of the N document data by N weighting factors when the second point is the position of the searcher, is an index y _i (i = 1 to N) based on the rank. ). In this case, the higher the ranking, the higher the weighting factor, or the higher the ranking, the lower the weighting factor. Similarly, when the snippet analysis data is used as the search result data A _i and B _i , the number obtained by multiplying the snippet analysis data at the first point and the second point of each of the N document data by N weighting factors. May be used as indices x _i (i = 1 to N) and y _i (i = 1 to N) based on the snippet analysis data.

(10) In the fourth to sixth embodiments, the click-through rate of each of the N pieces of document data when the first point is set as the position of the searcher is an index x _i based on the click-through rate. i = 1 to N), and the click-through rate of each of the N pieces of document data when the second point is the position of the searcher is an index y _i (i = 1 to N) based on the click-through rate. ). Similarly, when the stay times are the access analysis data C _i and D _i , the stay times of the first point and the second point of the N pieces of document data themselves are used as indices x _i (i = 1 to N) and y _i (i = 1 to N). Similarly, when the leaving rate is set as the access analysis data C _i and D _i , similarly, the leaving rate itself of the first point and the second point of each of the N document data is used as an index x _i based on the leaving rate. (I = 1 to N) and y _i (i = 1 to N) may be used. Similarly, when the flowing speed and the access analysis data C _i and D _i, the first point and the inlet number itself second point of each of the N document data index x _i based on the inflow speed (I = 1 to N) and y _i (i = 1 to N) may be used. Similarly, when the number of browsing pages is the access analysis data C _i and D _i , similarly, the number of browsing pages at the first point and the second point of each of the N document data is based on the number of browsing pages. The indices x _i (i = 1 to N) and y _i (i = 1 to N) may be used.

In addition, when the first point is set as the position of the searcher, the number obtained by converting the click through rate of each of the N pieces of document data into a logarithm (log) or a reciprocal (1 / click through rate) is the click through rate. an index x _{i (i} = 1~N) based on the logarithm of each of the click-through rate of the N document data in the case where the second point to the searcher's position (log) and the reciprocal (1 / click-through The number converted into (rate) may be used as an index y _i (i = 1 to N) based on the click-through rate. Similarly, when the stay time is the access analysis data C _i and D _i , the stay time at the first point and the second point of each of the N pieces of document data is similarly expressed as a logarithm (log) or a reciprocal (1 / stay time). converted number in), or as an index based on the stay time _x i (i = _1~N) and _y i (i = 1~N). Similarly, when the leaving rates are set as the access analysis data C _i and D _i , similarly, the leaving rates at the first point and the second point of each of the N document data are expressed by logarithm (log) or reciprocal (1 / leave rate). converted number in), or as an index based on the withdrawal rate _x i (i = _1~N) and _y i (i = 1~N). Similarly, when the number of inflows is set as the access analysis data C _i and D _i , the number of inflows at the first point and the second point of each of the N pieces of document data is expressed by logarithm (log) or reciprocal (1 / inflow number). the number of converted into), or as an index based on the inflow number _x i (i = _1~N) and _y i (i = 1~N). Similarly, when the number of browsing pages is set as the access analysis data C _i and D _i , the number of browsing pages at the first point and the second point of each of the N document data is similarly expressed as a logarithm (log) or an inverse number (1 / the number converted to number of pages viewed), or as an index based on the number of the pages viewed _x i (i = _1~N) and _y i (i = 1~N).

In addition, when the first point is set as the position of the searcher, a number obtained by multiplying each click-through rate of each of the N pieces of document data by N weighting factors is used as an index x _i (i based on the click-through rate). = 1 to N), and when the second point is the position of the searcher, an index based on the click-through rate is a number obtained by multiplying each click-through rate of the N document data by N weighting factors. It is good also as _yi (i = 1-N). In this case, the higher the click-through rate, the higher the weighting factor, or the higher the click-through rate, the lower the weighting factor. Similarly, when the stay time is set as the access analysis data C _i and D _i , the number obtained by multiplying the stay time at the first point and the second point of each of the N document data by N weighting factors or as an index based on the stay time _x i (i = _1~N) and _y i (i = 1~N). Similarly, when the leaving rate is set as the access analysis data C _i and D _i , the number obtained by multiplying the leaving rate at the first point and the second point of each of the N document data by N weighting factors The indices x _i (i = 1 to N) and y _i (i = 1 to N) based on the departure rate may be used. Similarly, when the number of inflows is set as the access analysis data C _i and D _i , similarly, a number obtained by multiplying the number of inflows at the first point and the second point of each of the N document data by N weighting factors is represented by: The indices x _i (i = 1 to N) and y _i (i = 1 to N) based on the number of inflows may be used. Similarly, when the number of browsing pages is the access analysis data C _i and D _i , similarly, the number obtained by multiplying the number of browsing pages at each of the first point and the second point of N document data by N weighting factors. May be used as indices x _i (i = 1 to N) and y _i (i = 1 to N) based on the number of browsed pages.

(11) In the first to sixth embodiments, the evaluation systems 1, 1A, 1B, 1C, 1D, and 1E including the evaluation devices 20, 20A, 20B, 20C, 20D, and 20E have been described. 20A, 20B, 20C, 20D, and 20E may be manufactured and sold alone. Moreover, in the said embodiment, by operating CPU22 according to the evaluation program 26, the said CPU22 was made to perform the evaluation method which shows the characteristic of this invention notably. However, each of the input means for executing the input process, the acquisition means for executing the acquisition process, and the analysis means for executing the analysis process is constituted by an electric circuit, and the evaluation devices 20, 20A of the present invention are combined by combining these means. 20B, 20C, 20D, and 20E may be configured.

(12) In the first to sixth embodiments, the evaluation program 26 that causes the CPU 22 to execute the evaluation method that clearly shows the features of the present invention is stored in advance in the hard disk 25 of the evaluation devices 20, 20A, 20B, 20C, 20D, and 20E. It was remembered. However, the evaluation program 26 may be distributed by being written on a computer-readable recording medium such as a DVD (Digital Versatile Disc) or a USB (Universal Serial Bus) memory, or the evaluation program can be downloaded via a telecommunication line such as the Internet. 26 may be distributed. By operating a general computer in accordance with the evaluation program 26 distributed in this manner, the computer can function as the evaluation devices 20, 20A, 20B, 20C, 20D, and 20E of the present invention.

(13) In the first to sixth embodiments, whether the change is caused by a search engine algorithm by comparing not only ρ in one search word but also ρ in many other search words, the search It is also possible to estimate the region-specific regional dependence. The search term here also conceptually includes a so-called second word, and it is possible to grasp the change in the appearance frequency in the above estimation process.

(14) In the first to sixth embodiments, the input procedure of the point input fields 52 and 53 on the input screen is performed by moving the mouse pointer to the point input fields 52 and 53 and performing a left click operation. The desired name is selected from the area names in the pull-down list that appears. However, the pull-down list may appear by moving the mouse pointer to the point input fields 52 and 53 and left-clicking.

(15) In the first to sixth embodiments, the user inputs a search word into the search word input field 51, and sets two points serving as a reference when evaluating the area dependence in the point input fields 52 and 53. When the search is performed based on the search word in the search word input field 51 at the first point indicated by the information in the point input field 52, the number N of search ranges is input to the analysis range input field 54. The document data of the top N-numbered web site, web page, or web content in the results was set as the analysis target. However, when a search is performed based on the search word in the search word input field 51 at the second point indicated by the information in the point input field 53, the top N-rank web site, web page, or web content in the search result Document data may be analyzed. In addition, the user may designate desired N web sites, web pages, or web content document data as analysis targets regardless of the first point and the second point. In short, it is only necessary to analyze N document data in the search result when a search based on a certain search word is performed.

(16) In the first to third embodiments, as shown in FIG. 17, an evaluation module for evaluating the relationship between search needs between any two points by a system having a neural network 85 and a learning evaluation unit 86. It may be generated. Specifically, for various search terms, a search result data group (specifically, a high-level web site or web page of the search result) for each search result obtained by searching various points as searcher positions. Or the morphological analysis processing is performed on the document data of the web content, and the processing result is expressed in a bag of words format) in the memory. In addition, data indicating the similarity between the search result at one point and the search result at the other point is stored in the memory for all combinations when two of the points selected as the searcher's position are selected. prepare. The data indicating the degree of similarity shows both the search result at one point and the search result at the other point to the subject. “How much do you feel the face of the search result is similar? 0.00 ～ 1. It is desirable that the answer data of the test subject be a questionnaire “Please answer between 00”.

  The learning evaluation unit 86, for various combinations of two locations such as “Tokyo” and “Hokkaido”, “Tokyo” and “Osaka”, “Hokkaido” and “Osaka”, When the search result data group of the other point is input to the neural network 85 as the input signals X1 and X2, the output signal Y output from the neural network 85 and the teacher signal T indicating the similarity of the combination are acquired and acquired. The processed signals Y and T are processed according to a deep learning algorithm, and based on the processing result, a weighting factor between the input layer and the intermediate layer and a weighting factor between the intermediate layer and the output layer in the neural network 85 And the learned model that has undergone the update of the weighting factor, the search needs of the first point and the second point for the search term And output as an evaluation module for evaluating the relationship between the search needs.

The procedure of updating the weighting coefficient by the learning evaluation unit 86 will be described in more detail. As shown in FIG. 18, the learning evaluation unit 86 calculates the error between the output signal Y of the neural network 85 and the teacher signal T by the following equation ( E in 2).

Learning evaluation unit 86 obtains the partial derivative ∂E / ∂W ₁ ² obtained by partially differentiating the error E in the weighting factor _W ^{1 2} 1 th intermediate layers and the output layers of neural networks 85, the weighting factor _W ^{1 2} Replace with W ₁ ²⁻ α (α is a positive integer) × (∂E / ∂W ₁ ² ). That, ∂E / ∂W ₁ ² is the _W ^{1 2} if it is positive in a predetermined amount by a small value, ∂E / ∂W ₁ ² is to a large value of _W ^{1 2} if it is negative predetermined amount. Similarly, the second and subsequent weight coefficients W ₂ ² , W ₃ ² ... Between the intermediate layer and the output layer are W ₂ ² −α × (∂E / ∂W ₂ ² ), W ₃ ² −α × (∂ E / ∂W ₃ ² ).

Learning evaluation unit 86 obtains the partial derivative ∂E / ∂W ₁ ¹ which is partially differentiated by the weight coefficient _W ^{1 1} of the first input layer and hidden layer of the neural network 85 the error E, the weighting coefficient _W ^{1 1} Replace with W ₁ ¹ −α (α is a positive integer) × (∂E / ∂W ₁ ¹ ). That is, if ∂E / ∂W ₁ ¹ is positive, W ₁ ¹ is reduced by a predetermined amount, and if ∂E / ∂W ₁ ¹ is negative, W ₁ ¹ is increased by a predetermined amount. Similarly, the second and subsequent weight coefficients W ₂ ¹ , W ₃ ¹ ... Between the intermediate layer and the output layer are W ₂ ¹ −α × (∂E / ∂W ₂ ¹ ), W ₃ ¹ −α × (∂ Replace with E / ∂W ₃ ¹ ).

  Each time the learning evaluation unit 86 acquires the output signal Y of the neural network 85 and the teacher signal T corresponding to the output signal Y, the learning evaluation unit 86 has repeatedly learned the set of weighting coefficients that have been updated a predetermined number of times or for a predetermined period. Model.

(17) In (16), the weighting factor in the neural network 85 may be updated not by deep learning with the teacher signal T but by deep learning without the teacher signal T. Further, the search result data group may be obtained by performing a morphological analysis process on document data of a web site, a web page, or a web content at a higher rank of the search result, and expressing the process result in a one-hot expression. In FIG. 17, for the sake of simplicity, the intermediate layer of the neural network 85 is one layer, but the intermediate layer may be a plurality of layers. When the intermediate layer is a plurality of layers, the learning evaluation unit 86 processes the signals Y and T in accordance with the deep learning algorithm, and based on the processing result, the learning evaluation unit 86 performs a phase change between the input layer and the intermediate layer. It is preferable to update the weighting factor between layers and between the intermediate layer and the output layer.

(18) In the fourth to sixth embodiments, the access analysis data of the access destination in the search result when a search for a predetermined search word is performed with a plurality of points as the positions of the searcher is input to the neural network 85. Based on the error between the output signal of the neural network 85 and the teacher signal, the weighting coefficient may be updated by the same procedure as in (17).

1, 1A, 1B, 1C, 1D, 1E ... evaluation system, 10 ... user terminal, 20, 20A, 20B, 20C, 20D, 20E ... evaluation device, 21 ... communication interface, 22 ... CPU, 23 ... RAM, 24 ... ROM, 25 ... hard disk, 26 ... evaluation program, 30 ... document management server device, 40 ... search result collection server device, 45 ... access analysis result collection server device, 50 ... search engine server device, 60 ... access analysis server device.

Claims (28)

An input means for inputting a search term;
The search result data based on the search word when the first point is set as the position of the searcher is acquired, and the second point different from the first point is set as the position of the searcher. An acquisition means for acquiring search result data based on a search term;
Analysis means for obtaining an index indicating a correlation between the search result data of the first point and the search result data of the second point, and outputting the index as an analysis result unique to the search word. Search term evaluation device. The search result data is an order of N (N is a natural number of 2 or more) document data in a search result when a search based on the search word is performed from a document data group described in a markup language. ,
For each of the N pieces of document data, the analysis means includes the ranking of the document data when the first point is the position of the searcher and the order when the second point is the position of the searcher. The evaluation apparatus according to claim 1, wherein an index indicating the correlation is obtained by applying the order of document data to a predetermined function. The analysis means includes
For the i-th document data of the N pieces of document data, the ranking of the i-th document data or the index based on the ranking when the first point is the position of the searcher is x _i , the index Xm is an average value of the i-th document data when the second point is the position of the searcher, or an index based on the rank is y _i , and an average value of the index is ym, The evaluation apparatus according to claim 2, wherein the index ρ indicating the correlation is calculated according to a function represented by the following expression (1).
Information indicating a plurality of the second points is input to the input means,
The evaluation apparatus according to claim 3, further comprising grouping means for grouping each of the plurality of second points according to the value of the index ρ calculated by the analysis means. The search result data is predetermined in a snippet of N (N is a natural number of 2 or more) document data in a search result when a search based on the search word is performed from a document data group described in a markup language. Snippet analysis data obtained by performing the analysis process of
The analysis means, for each of the N document data, when the snippet analysis data of the document data and the second point are the position of the searcher when the first point is the position of the searcher The evaluation apparatus according to claim 1, wherein an index indicating the correlation is obtained by applying snippet analysis data of the document data in a predetermined function. The analysis means includes
For the i-th document data of the N document data, the snippet analysis data of the i-th document data or the index based on the snippet analysis data when the first point is the position of the searcher is x _i , the average value of the index is xm, and the snippet analysis data of the i-th document data or the index based on the snippet analysis data when the second point is the position of the searcher is y _i , 6. The evaluation apparatus according to claim 5, wherein an average value of the index is ym, and the index ρ indicating the correlation is calculated according to a function represented by the following formula (1).
A user terminal and an evaluation device that communicates with the user terminal;
The evaluation device is
An input means for inputting a search term;
The search result data based on the search word when the first point is set as the position of the searcher is acquired, and the second point different from the first point is set as the position of the searcher. An acquisition means for acquiring search result data based on a search term;
Analysis means for obtaining an index indicating a correlation between the search result data of the first point and the search result data of the second point, and outputting the index as an analysis result unique to the search word. A search word evaluation system. An input step in which the search word is input to an evaluation device that evaluates the search word to be evaluated ;
The evaluation device acquires search result data based on the search word when the first point is the searcher's position, and sets the searcher's position to a second point different from the first point. An acquisition step of acquiring search result data of a search based on the search term in the case of
An analysis step in which the evaluation device obtains an index indicating a correlation between the search result data of the first point and the search result data of the second point, and outputs the index as an analysis result unique to the search word; A method for evaluating a search term, comprising: An input means for inputting a search term;
The access analysis data of the access destination in the search result based on the search word when the first point is set as the searcher's position is acquired, and the second point different from the first point is determined as the searcher. An acquisition means for acquiring access analysis data of an access destination in a search result based on the search term when the position is
Analyzing means for obtaining an index indicating a correlation between the access analysis data at the first point and the access analysis data at the second point, and outputting the index as an analysis result specific to the search term. Search term evaluation device. The access analysis data is a click-through rate of N (N is a natural number of 2 or more) document data in a search result when a search based on the search word is performed from a document data group described in a markup language. And
When each of the N pieces of document data has the first point as the searcher's position, the analysis means uses the document data click-through rate and the second point as the searcher's position. The evaluation apparatus according to claim 9, wherein an index indicating the correlation is obtained by applying a click-through rate of the document data in the above to a predetermined function. The analysis means includes
For the i-th document data of the N document data, the click-through rate of the i-th document data or the index based on the click-through rate when the first point is set as the position of the searcher is x _i , the average value of the index is xm, and the click-through rate of the i-th document data or the index based on the click-through rate when the second point is the position of the searcher is y _i , The evaluation apparatus according to claim 10, wherein an average value of the index is ym, and the index ρ indicating the correlation is calculated according to a function represented by the following expression (1).
Information indicating a plurality of the second points is input to the input means,
The evaluation apparatus according to claim 11, further comprising grouping means for grouping each of the plurality of second points according to the value of the index ρ calculated by the analysis means. The access analysis data is a stay of addresses of N (N is a natural number of 2 or more) document data in a search result when a search based on the search word is performed from a document data group described in a markup language. Time,
For each of the N pieces of document data, the analyzing means includes a stay time of the document data when the first point is the position of the searcher and a case where the second point is the position of the searcher. The evaluation apparatus according to claim 9, wherein an index indicating the correlation is obtained by applying a staying time of the document data to a predetermined function. The analysis means includes
For the i-th document data of the N document data, a stay time of the i-th document data or an index based on the stay time when the first point is the position of the searcher is x _i , The average value of the index is xm, and the stay time of the i-th document data or the index based on the stay time when the second point is the position of the searcher is y _i , and the average value of the index The evaluation apparatus according to claim 13, wherein ym is calculated, and an index ρ indicating the correlation is calculated according to a function represented by the following expression (1).
Information indicating a plurality of the second points is input to the input means,
The evaluation apparatus according to claim 14, further comprising grouping means for grouping each of the plurality of second points according to the value of the index ρ calculated by the analysis means. The access analysis data is a detachment of addresses of N (N is a natural number of 2 or more) document data in a search result when a search is performed based on the search word from a document data group described in a markup language. Rate,
For each of the N pieces of document data, the analysis unit is configured such that when the first point is the searcher's position, the document data withdrawal rate and the second point is the searcher's position. The evaluation apparatus according to claim 9, wherein an index indicating the correlation is obtained by applying a withdrawal rate of the document data to a predetermined function. The analysis means includes
For the i-th document data of the N document data, the i-th document data withdrawal rate or the index based on the withdrawal rate when the first point is the searcher's position is represented by x _i , The average value of the index is xm, and when the second point is set as the position of the searcher, the exit rate of the i-th document data or the index based on the exit rate is y _i , and the average value of the index The evaluation apparatus according to claim 16, wherein ym is calculated, and an index ρ indicating the correlation is calculated according to a function represented by the following expression (1).
Information indicating a plurality of the second points is input to the input means,
The evaluation apparatus according to claim 17, further comprising grouping means for grouping each of the plurality of second points according to the value of the index ρ calculated by the analysis means. The access analysis data is an inflow number of addresses of document data (N is a natural number of 2 or more) in a search result when a search is performed based on the search word from a document data group described in a markup language. And
For each of the N pieces of document data, the analyzing means includes the number of inflows of the document data when the first point is the position of the searcher and the second point is the position of the searcher. The evaluation apparatus according to claim 9, wherein the index indicating the correlation is obtained by causing the number of inflows of the document data to act on a predetermined function. The analysis means includes
For the i-th document data of the N document data, the number of inflows of the i-th document data or an index based on the number of inflows when the first point is set as the position of the searcher is x _i , When the average value of the index is xm, the number of inflows of the i-th document data or the index based on the number of inflows when the second point is the position of the searcher is y _i , and the average value of the index The evaluation apparatus according to claim 19, wherein ym is calculated, and an index ρ indicating the correlation is calculated according to a function represented by the following expression (1).
Information indicating a plurality of the second points is input to the input means,
The evaluation apparatus according to claim 20, further comprising grouping means for grouping each of the plurality of second points according to the value of the index ρ calculated by the analysis means. The access analysis data is a browsing of addresses of N (N is a natural number of 2 or more) document data in a search result when a search based on the search word is performed from a document data group described in a markup language. The number of pages,
In the case where each of the N pieces of document data has the first point as the searcher's position, the analysis unit has the number of browsing pages of the document data and the second point as the searcher's position. The evaluation apparatus according to claim 9, wherein the index indicating the correlation is obtained by causing the number of pages to be browsed of the document data to act on a predetermined function. The analysis means includes
For the i-th document data of the N pieces of document data, when the first point is the searcher's position, the number of pages to be browsed of the i-th document data or an index based on the number of browsed pages is x _i , the average value of the index is xm, and the index of the i-th document data when the second point is the searcher's position or the index based on the number of browsed pages is y _i , The evaluation apparatus according to claim 22, wherein an average value of the index is ym, and the index ρ indicating the correlation is calculated according to a function represented by the following expression (1).
Information indicating a plurality of the second points is input to the input means,
The evaluation apparatus according to claim 23, further comprising grouping means for grouping each of the plurality of second points according to the value of the index ρ calculated by the analysis means. A user terminal and an evaluation device that communicates with the user terminal;
The evaluation device is
An input means for inputting a search term;
The access analysis data of the access destination in the search result based on the search word when the first point is set as the searcher's position is acquired, and the second point different from the first point is determined as the searcher. An acquisition means for acquiring access analysis data of an access destination in a search result based on the search term when the position is
Analyzing means for obtaining an index indicating a correlation between the access analysis data at the first point and the access analysis data at the second point, and outputting the index as an analysis result unique to the search term. A search word evaluation system. An input step in which the search word is input to an evaluation device that evaluates the search word to be evaluated ;
The evaluation device acquires access analysis data of an access destination in a search result based on the search word when the first point is a searcher's position, and a second different from the first point. An acquisition step of acquiring access analysis data of an access destination in a search result based on the search term when the point is a searcher's position;
An analysis step in which the evaluation device obtains an index indicating a correlation between the access analysis data at the first point and the access analysis data at the second point, and outputs the index as an analysis result unique to the search term; A method for evaluating a search term, comprising: A neural network including an input layer, an intermediate layer, and an output layer;
An output signal output from the neural network when a search result data group when a search for a predetermined search word is performed with a plurality of points as the positions of searchers is input to the neural network, or the output signal and A teacher signal is acquired, and the acquired signal is processed according to a deep learning algorithm. Based on the processing result, weight coefficients between the input layer and the intermediate layer and between the intermediate layer and the output layer are calculated. Has a learning evaluation unit to update,
The learned model that has undergone the updating of the weighting coefficient by the learning evaluation unit,
An evaluation module production method comprising: outputting as an evaluation module for evaluating a relationship between a search need for a first point and a search need for a second point for the search word. A neural network including an input layer, an intermediate layer, and an output layer;
An output signal output from the neural network when an access analysis data group of access destinations in a search result when a search is performed for a predetermined search word with a plurality of points as the positions of searchers is input to the neural network Alternatively, the output signal and the teacher signal are acquired, the acquired signal is processed according to a deep learning algorithm, and based on the processing result, between the input layer and the intermediate layer and between the intermediate layer and the output layer A learning evaluation unit that updates the weighting coefficient between
The learned model that has undergone the updating of the weighting coefficient by the learning evaluation unit,
An evaluation module production method comprising: outputting as an evaluation module for evaluating a relationship between a search need for a first point and a search need for a second point for the search word.