JP6924450B2 - Search needs evaluation device, search needs evaluation system, and search needs evaluation method - Google Patents

Description

The present invention relates to a technique for evaluating a search intention (hereinafter, appropriately referred to as "search needs") of a word used as a search term of a search engine.

Google® technology utilizes search results and various behavioral data displayed in the search results (specifically, click rate, time spent on the site, etc.) to determine the search ranking. In a search engine, which is a service based on this technology, the more clicks and longer the site stays, the easier it is for the search ranking to rise. Details of this technique are disclosed in Patent Document 1 (particularly, paragraphs 0088-0090). SEO (Search Engine Optimization) is one of the methods for adjusting the configuration of a website so that a specific website is displayed at the top in the search results of a search engine. Patent Document 2 is a document that discloses a technique related to SEO. When a certain word is input as a target keyword, the Web page analysis device of Patent Document 2 sets each of a plurality of Web page data in the search result for the target keyword as the analysis target Web page, and uses the analysis target Web page data as the analysis target Web page data. Morphological analysis processing is performed, the content of each morphological element of the same type in the morphological element group obtained by the morphological analysis processing is aggregated, and evaluation by morphological element indicating the degree of contribution of each morphological element to the ranking of the analysis target Web page in the search results. The values are obtained, and a list in which the evaluation values for each morphological element are arranged for each analysis target Web page is presented as the analysis result. According to the technique of Patent Document 2, a morpheme having a high SEO effect can be efficiently found.

US 2012/0209838A1 Patent No. 6164436

However, in this technique (Patent Document 2), when one target search keyword is used in a plurality of different search needs, it is not possible to obtain a clear analysis result for each of the plurality of search needs. That is, since multiple Web page data in the search results are analyzed together without considering the existence of a plurality of different search needs, it is not possible to obtain an appropriate morpheme-specific evaluation value for each search need. There was a challenge.

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 assisting analysis of the nature of search needs.

According to one aspect of the present invention, a similarity acquisition means for acquiring the similarity of search needs between each search term based on the search results for each of the plurality of search terms, and a node associated with each search term. , And a display control means for displaying a screen including, the length of the edge is the degree of similarity between the search terms associated with the nodes connected via the edge. A corresponding search needs evaluation device is provided.

The display control means may move a specific node according to a user operation, and may move at least one node connected to the specific node via an edge according to the movement of the specific node. ..

A classification means for classifying each search term into clusters based on the search results for each of the plurality of search terms is provided, and the display control means displays the nodes in a display mode according to the cluster in which each search term is classified. You may let me.

The classification means can calculate how close each search term is to each of two or more clusters, and the display control means displays nodes in a display mode according to how close each search term is to which cluster. It may be displayed.

The classification means can classify each search term into a cluster with a plurality of levels of particle size, and even if each search term is classified into a cluster according to the set particle size each time the particle size is set according to the user operation. good.

The display control means may change the display mode of the node when the particle size is changed according to the user operation and the cluster into which each search term is classified changes.

The display control means may display the nodes in a display mode according to the number of searches for each search term in a certain period.

A quantification means for converting at least one of the content and structure of the document data which is a search result for each of a plurality of search terms into multidimensional feature vector data is provided, and the similarity acquisition means is the feature vector for each search term. The similarity between each search term may be acquired based on the similarity between data.

According to another aspect of the present invention, the step of acquiring the similarity of the search needs between the search terms by the similarity acquisition means based on the search results for each of the plurality of search terms, and the display control means. Each search term comprises a node associated with it, an edge connecting the nodes, and a step of displaying a screen including, and the length of the edge was associated with the node connected through the edge. A search needs evaluation method corresponding to the similarity between search terms is provided.

According to another aspect of the present invention, each search term is associated with a similarity acquisition means for acquiring the similarity of search needs between each search term based on the search results for each of a plurality of search terms. It functions as a display control means for displaying a screen including a connected node and an edge connecting the nodes, and the length of the edge is a search term associated with the node connected via the edge. A search needs evaluation program is provided that corresponds to the degree of similarity between the two.

According to another aspect of the present invention, at least one of the acquisition means for acquiring a plurality of document data in the search result based on a certain search term and the content and structure of the plurality of document data is converted into multidimensional feature vector data. Analysis of the nature of the search needs based on the quantification means to be converted, the classification means for classifying the plurality of document data into a plurality of subsets based on the feature vector data, and the relationship between the plurality of subsets. A search needs evaluation device is provided, which comprises an analysis result output means for outputting a result.

The classification means may process the feature vector data according to a clustering algorithm or a classification algorithm, and classify the plurality of document data into a plurality of subsets.

The acquisition means acquires the document data in the search result for each search term for each of the plurality of search terms, and the quantification means obtains the content and structure of the plurality of document data in the search result for each search term. It is provided with a synthesis means for converting at least one of the feature vector data into multidimensional feature vector data, performing predetermined statistical processing on the feature vector data for each document obtained by the quantification means, and synthesizing the feature vector data for each search term. You may.

The acquisition means acquires the document data in the search result for each search term for each of the plurality of search terms, and the quantification means obtains the content and structure of the plurality of document data in the search result for each search term. At least one of them is converted into multidimensional feature vector data, and the classification means classifies a plurality of document data into a plurality of subsets based on the feature vector data for each document, and predetermined statistics are obtained based on the processing result by the classification means. A synthesis means for performing processing and synthesizing the processing results for each search term may be provided.

The plurality of documents are provided with dimensional reduction means for dimensionally reducing the feature vector data to lower-dimensional feature vector data, and the classification means is based on the feature vector data that has undergone dimensional reduction of the dimensional reduction means. The data may be classified into multiple subsets.

According to another aspect of the present invention, at least one of the acquisition means for acquiring a plurality of document data in the search result based on a certain search term and the content and structure of the plurality of document data is converted into multidimensional feature vector data. A quantification means to be converted, a similarity identification means for specifying the similarity between the feature vector data of the plurality of document data, and a community detection for classifying the plurality of document data into a plurality of communities based on the similarity. A search needs evaluation device is provided, which comprises means and analysis result output means for outputting an analysis result of search needs based on the relationship between the plurality of communities.

The acquisition means acquires the document data in the search result for each search term for each of the plurality of search terms, and the quantification means obtains the content and structure of the plurality of document data in the search result for each search term. At least one of them is converted into multidimensional feature vector data, the similarity specifying means identifies the similarity between the feature vector data of a plurality of document data for each search term, and the community detecting means is for each search term. Features of a plurality of document data Based on the similarity between vector data, a plurality of document data for each search term is classified into a plurality of communities, and predetermined statistics are obtained in the processing result of community detection for each search term by the community detection means. A synthesis means may be provided which performs processing and synthesizes the processing result of community detection for each search term.

According to another aspect of the present invention, at least one of the acquisition step of acquiring a plurality of document data in the search result based on a certain search term and the content and structure of the plurality of document data is converted into multidimensional feature vector data. Analysis of the nature of the search needs based on the quantification step to be transformed, the classification step of classifying the plurality of document data into a plurality of subsets based on the feature vector data, and the relationship between the plurality of subsets. A search needs evaluation method is provided, which comprises an analysis result output step for outputting a result.

According to another aspect of the present invention, at least one of the acquisition step of acquiring a plurality of document data in the search result based on a certain search term and the content and structure of the plurality of document data is converted into multidimensional feature vector data. A quantification step to be converted, a similarity identification step for specifying the similarity between feature vector data of the plurality of document data, and a community detection for classifying the plurality of document data into a plurality of communities based on the similarity. A search needs evaluation method is provided, which comprises a step and an analysis result output step for outputting an analysis result of the search needs based on the relationship between the plurality of communities.

According to another aspect of the present invention, a computer has a multidimensional feature of at least one of an acquisition step of acquiring a plurality of document data in a search result based on a certain search term and at least one of the contents and structure of the plurality of document data. A quantification step for converting to vector data, a classification step for classifying the plurality of document data into a plurality of subsets based on the feature vector data, and a search need based on the relationship between the plurality of subsets. A search needs evaluation method is provided, which comprises executing an analysis result output step that outputs an analysis result of a property.

An acquisition step of acquiring a plurality of document data in a search result based on a certain search term on a computer, and a quantification step of converting at least one of the contents and structures of the plurality of document data into multidimensional feature vector data. A similarity identification step for specifying the similarity between feature vector data of the plurality of document data, a community detection step for classifying the plurality of document data into a plurality of communities based on the similarity, and the plurality of communities. A search needs evaluation method is provided, which comprises executing an analysis result output step that outputs an analysis result of search needs based on the relationship between the two.

According to the present invention, it is possible to quantitatively evaluate or display the variety of search needs for each search term. In addition, in the conventional technology, the evaluation of morphemes contained in the search result Web page, which could be evaluated only for each search term, can be evaluated for each search need. It will be easier to create pages, etc.

It is a figure which shows the whole structure of the evaluation system including the search needs evaluation apparatus which is 1st Embodiment of this invention. It is a flowchart which shows the flow of the evaluation method which the CPU of the search needs evaluation apparatus which is 1st Embodiment of this invention executes according to an evaluation program. It is a figure which shows the procedure of the clustering process of the search needs evaluation apparatus which is 1st Embodiment of this invention. It is a figure which shows the procedure of setting the evaluation axis of the search needs evaluation apparatus which is 1st Embodiment of this invention. It is a figure which shows the outline of the process of the search needs evaluation apparatus which is 1st Embodiment of this invention. It is a flowchart which shows the flow of the evaluation method which the CPU of the search needs evaluation apparatus which is 2nd Embodiment of this invention executes according to the evaluation program. It is a figure which shows the procedure of the classification process of the search needs evaluation apparatus which is 2nd Embodiment of this invention. It is a figure which shows the outline of the process of the search needs evaluation apparatus which is 2nd Embodiment of this invention. It is a flowchart which shows the flow of the evaluation method which the CPU of the search needs evaluation apparatus which is 3rd Embodiment of this invention executes according to the evaluation program. It is a figure which shows the outline of the process of the search needs evaluation apparatus which is 3rd Embodiment of this invention. It is a flowchart which shows the flow of the evaluation method which the CPU of the search needs evaluation apparatus which is 4th Embodiment of this invention executes according to the evaluation program. It is a figure which shows the outline of the process of the search needs evaluation apparatus which is 4th Embodiment of this invention. It is a flowchart which shows the flow of the evaluation method which the CPU of the search needs evaluation apparatus which is 5th Embodiment of this invention executes according to the evaluation program. It is a figure which shows the outline of the process of the search needs evaluation apparatus which is 5th Embodiment of this invention. It is a flowchart which shows the flow of the evaluation method which the CPU of the search needs evaluation apparatus which is 6th Embodiment of this invention executes according to the evaluation program. It is a figure which shows the outline of the process of the search needs evaluation apparatus which is 6th Embodiment of this invention. It is a flowchart which shows the flow of the evaluation method which the CPU of the search needs evaluation apparatus which is 7th Embodiment of this invention executes according to the evaluation program. It is a figure which shows the outline of the process of the search needs evaluation apparatus which is 7th Embodiment of this invention. It is a flowchart which shows the flow of the evaluation method which the CPU of the search needs evaluation apparatus which is 8th Embodiment of this invention executes according to the evaluation program. It is a figure which shows the outline of the process of the search needs evaluation apparatus which is 8th Embodiment of this invention. It is a flowchart which shows the flow of the evaluation method which the CPU of the search needs evaluation apparatus which is 9th Embodiment of this invention executes according to the evaluation program. It is a figure which shows the outline of the process of the search needs evaluation apparatus which is 9th Embodiment of this invention. It is a figure which shows the processing content of the search needs evaluation apparatus which is a modification of this invention. It is a figure which shows the processing content of the search needs evaluation apparatus which is a modification of this invention. It is a figure which shows the mapping image 7 of FIG. 11 more concretely. It is a figure which shows the state which moved the node n3 associated with "ABC business" in FIG. It is a figure which shows the mapping image 7 which the search term is classified into a cluster, and the node is displayed in the display mode corresponding to the classified cluster. It is a figure which shows the mapping image 7 in the case where the search term is not fixed in the classification into one cluster, but can be classified into a plurality of clusters. It is a figure which shows the mapping image 7 which a user can set the particle size. FIG. 29 is a diagram showing a state in which the particle size is finely set as compared with FIG. 29. It is a figure which shows the example of the interface of particle size adjustment. It is a figure which shows the example of the interface of particle size adjustment. It is a figure which shows the example of the interface of particle size adjustment. It is a figure which shows the example of the interface of particle size adjustment. It is a figure which shows the example of the interface of particle size adjustment. It is a figure which shows the mapping image 7 which displayed the node in the mode corresponding to the search number of each search term. It is a figure which shows the screen example at the time of displaying the analysis result in a table format. FIG. 37 is a diagram showing a state in which the particle size of FIG. 37 is coarsened. It is a figure which shows the screen example at the time of displaying the analysis result in the correlation matrix format. It is a figure which shows the state in which the search term of FIG. 39 is rearranged. It is a figure which shows the screen example at the time of displaying the analysis result in the dendrogram format. It is a figure which shows the state which moved the particle size setting bar 36 of FIG. 41. It is a figure which shows the screen example at the time of displaying the analysis result in the tree map format. It is a figure which shows the screen example at the time of displaying the analysis result in the sunburst format.

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 a search needs evaluation device 20 according to a first embodiment of the present invention. As shown in FIG. 1, the evaluation system 1 includes a user terminal 10 and a search needs evaluation device 20. The user terminal 10 and the search needs 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 patrolls the Internet 90 and indexes information obtained from web pages scattered on the Internet 90 as document data (data described in a markup language such as HTML (Hyper Text Markup Language)). The title and URL (Uniform Resource Locator) of the web page searched by receiving the HTTP (Hyper Text Transfer Protocol) request (search query) including the search term from the searcher's computer and using the search term in the search query. ), The search process of returning the search result in which the set of snippets is arranged in order from the highest (highest ranking) is performed. Although only one search engine server device 50 is shown in FIG. 1, the number of search engine server devices 50 may be plural.

The user terminal 10 is a personal computer. A unique ID and password are assigned to the user of the user terminal 10. The user accesses the search needs evaluation device 20 from his / her own user terminal 10 to perform an authentication procedure, and uses the service of the search needs evaluation device 20. Although only one user terminal 10 is shown in FIG. 1, the number of user terminals 10 in the evaluation system 1 may be plural.

The search needs evaluation device 20 is a device that plays a role of providing a search needs evaluation service. The search needs evaluation service receives the search term to be evaluated from the user, and classifies the top d (d is a natural number of 2 or more) web pages in the search result of the search term by a predetermined statistical classification processing algorithm. However, it is a service that presents a set of a plurality of web pages obtained by this classification as an analysis result.

As shown in FIG. 1, the search needs evaluation device 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 / receives data to / from a device 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. IPL (Initial Program Loader) and the like are stored in the ROM 24. The hard disk 25 stores an evaluation program 26 having a function peculiar to the present embodiment.

Next, the operation of this embodiment will be described. FIG. 2 is a flowchart showing a flow of an evaluation method executed by the CPU 22 of the search needs evaluation device 20 according to the evaluation program 26. By executing the evaluation program 26, the CPU 22 executes the acquisition process (S100), the quantification means for executing the quantification process (S200), the addition means for executing the addition process (S210), and the dimension reduction. Quantification means for executing processing (S300), classification means for executing clustering processing (S310), analysis result output means for executing analysis result output processing (S400), and evaluation for executing evaluation axis setting processing (S450). Functions as an axis setting means.

In the acquisition process of step S100, the CPU 22 receives the search term to be evaluated from the user terminal 10, and the document data D _k (k = 1 to 1) of the top d web pages in the search result based on the search term to be evaluated. d and k are indexes indicating the ranking). The document data D _k (k = 1 to d) describes the content and structure of the k-th web page in the search result by HTML. In the following, the document data D _{k (k = 1 to d} ) will be appropriately referred to as document data D 1, D ₂ ... D d.

The quantification process in step S200 includes a document content quantification process (S201) and a document structure quantification process (S202). The document content quantification process is _{a process of converting the contents of the document data D 1} , D ₂ ... D _d into n (n is a natural number of 2 or more) dimensional feature vector data. The document structure quantification process is _{a process of converting the structure of the document data D 1} , D ₂ ... D _d into m (m is a natural number of 2 or more) dimensional feature vector data. In the following, the n-dimensional feature vector data of each content of _{the document data D 1} , D ₂ ... D _{d is referred to as the feature vector data x 1} = {x ₁₁ , x ₁₂ ... x _1n }, x ₂ =. It is described as {x ₂₁ , x ₂₂ ... x _2n } ... x _d = {x _d1 , x _d2 ... x _dn }. Further, the m-dimensional feature vector data of each structure of the document data D ₁ , D ₂ ... D _{d can be used as the feature vector data y 1} = {y ₁₁ , y ₁₂ ... y _{1 m} }, y ₂ = {. y ₂₁ , y ₂₂ ... y _2m } ... y _d = {y _d1 , y _d2 ... y _dm }.

More specifically, in the document content quantification process, the CPU 22 _{converts the document data D 1} into a multidimensional vector according to algorithms such as Bag of Words (BoW), dmpv (Distributed Memory), and DBoW (Distributed BoW). The processing result is the feature vector data x ₁ = {x ₁₁ , x ₁₂ ... x _1n }, x ₂ = {x ₂₁ , x ₂₂ ... x _2n } ... x _d = {x _d1 , x _d2. ... x _dn }. The CPU 22 multidimensionally vectorizes the document data D ₂ ... D _d according to the same algorithm, and converts this processing result into the feature vector data x ₂ = {x ₂₁ , x _{22 of each of the document data D 2} ... D _d. ... x _2n } ... x _d = {x _d1 , x _d2 ... x _dn }. Here, dmpv and DBoW are a kind of Doc2Vec.

In the document structure quantification process, the CPU 22 _{converts the document data D 1} into a multidimensional vector according to algorithms such as Hidden Markov Model (HMM), Probabilistic Context Free Grammar (PCFGP), Recurrent Neural Network, and Recursive Neural Network, and this process. Let the result be the feature vector data y ₁ = {y ₁₁ , y ₁₂ ... y _{1 m} } of the document data D _1. The CPU 22 multidimensionally vectorizes the document data D ₂ ... D _d according to the same algorithm, and converts this processing result into the feature vector data y ₂ = {y ₂₁ and y _{22 of each of the document data D 2} ... D _d. ... y _2m } ... y _d = {y _d1 , y _d2 ... y _dm }.

The addition process of step S210 is a process of adding the processing result of step S201 and the processing result of step S202 and outputting the feature vector data of l (l = n + m) dimension. In the following, the l-dimensional feature vector data obtained by the addition processing for each of the _{document data D 1} , D ₂ ... D _{d is the feature vector data z 1} = {z ₁₁ , z ₁₂ ... z _1l }. , Z ₂ = {z ₂₁ , z ₂₂ ... z _2l } ... z _d = {z _d1 , z _d2 ... z _dl }.

In the dimension reduction process of step S300, the feature vector data z ₁ = {z ₁₁ , z ₁₂ ... z _1l }, z ₂ = {z ₂₁ , z ₂₂ ... z _2l } ... z _d = { This _{is a process of dimensionally reducing z d1} , z _d2 ... z _dl } to l'dimensional feature vector data having a smaller number of dimensions according to an algorithm such as an autoencoder or principal component analysis. In the following, the l'dimensional feature vector data obtained by dimensional reduction for each of the _{document data D 1} , D ₂ ... D _{d is referred to as the feature vector data z 1} = {z ₁₁ , z ₁₂ ... z. _{1 l'} }, z ₂ = {z ₂₁ , z ₂₂ ... z _2l' } ... z _d = {z _d1 , z _d2 ... z _dl' }.

The clustering process in step S310 is _{a statistical classification process for classifying document data D 1} , D ₂ ... D _d into a plurality of subsets (lumps) called clusters. In the clustering process, the CPU 22 uses the feature vector data z ₁ = {z ₁₁ , z ₁₂ ... z _1l' }, z ₂ = {z ₂₁ , z ₂₂ · _{of the document data D 1} , D ₂ ... D _d.・ ・ Z _2l' } ・ ・ ・ z _d = {z _d1 , z _d2・ ・ ・ z _dl' } is processed according to the algorithm of the shortest distance method of clustering, and the document data D ₁ , D ₂ ... Classify D _d into multiple clusters.

The details of the shortest distance method of clustering will be described. In FIGS. 3 (A), 3 (B), 3 (C), and 3 (D), _{the number d of the document data D k} is d = 9, and the number of dimensions l'is l'= 2. It is a figure which shows the classification example of the case. Clustering, for all combinations of two document data _{D k} in the document data _D k (k = 1 to d), determine the distance between the two document data _{D k.} The distance between the two document data _Dk may be the Euclidean distance, the Minkowski distance, or the Mahalanobis distance.

As shown in FIG. 3 (A), the two document data D _{k (D 1} and D _{2 in} the example of FIG. 3 (A)) that are closest to each other are grouped as the first cluster. After bundling the cluster, the representative point R (center of gravity) of the cluster is obtained, and the representative point R and the document data D _k outside the cluster (in the example of FIG. 3 (A), the document data D ₃ , D ₄ , D _{5 ,,} Find the distance to _{D 6} , D ₇ , D ₈ , D _9).

As shown in FIG. 3 (B), two document data D _k outside the cluster whose distance from each other is shorter than the distance from the representative point R (in the example of FIG. 3 (B), the document data D _3). , D ₄ ), the two document data D _k are bundled as a new cluster. Further, as shown in FIG. 3C, two clusters in which the distance between the representative points R of each other _{is shorter than the distance from the document data Dk} outside the cluster (in the example of FIG. 3C, If there is a cluster of document data D ₁ and D _{2 and} a cluster of document data D ₃ and D ₄ , the two clusters are grouped as a new cluster. As shown in FIG. 3D, the above processing is recursively repeated to generate a plurality of clusters having a hierarchical structure.

In FIG. 2, the analysis result output process of step S400 is a process of outputting the analysis result of the nature of the search needs related to the search term to be evaluated based on the relationship between the clusters. As shown in FIG. 2, in the analysis result output processing, the CPU 22 transmits HTML data of the analysis result screen to the user terminal 10 and causes the user terminal 10 to display the analysis result screen on the display. The analysis result screen has a high-level page classification and a denogram 8. _{The upper page classification is a matrix of five frames F k} (k = 1 to d) in which a summary (title, snippet) of the top d web pages in the search results based on the search term to be evaluated is written. It is arranged in. In Figure 2, only the frame _F 1 _{to F 10} of web pages of the positions 1 to # 10 is displayed, the scroll bar operation, causing the appearance of the frame _{F k} of web pages of # 11 and subsequent You can also. _{The frame F k} (k = 1 to d) of the web page in the upper page classification is color-coded so that the ones distributed to the same cluster by clustering have the same color. For convenience, in FIG. 2, the first color frame F _k (in the example of FIG. 2, the first frame F ₁ , the third frame F ₃ , the fourth frame F ₄ , and the fifth frame F _5). , 7th frame F ₇ , 10th frame F ₁₀ ) with a thin line, 2nd color frame F _k (in the example of Fig. 2, 2nd frame F ₂ , 8th frame F ₈ , 9th place The frame F ₉ ) of is indicated by a thick line, and the frame F _k of the third color (in the example of FIG. 2, the frame F _{6 at the} 6th position) is indicated by a chain line. The dendrogram 8 shows the hierarchical structure of the clusters obtained in the process of clustering.

The evaluation axis setting process in step S450 is a process for setting the evaluation axis of the clustering process. As shown in FIG. 4A, there is an evaluation axis setting bar 9 on the denogram 8 of the analysis result screen. The evaluation axis setting bar 9 serves to specify the number of clusters in the clustering process. The evaluation axis setting bar 9 can be moved up and down by operating the pointing device of the user terminal 10. When the user wants to obtain an analysis result with coarser classification, the user moves the evaluation axis setting bar 9 to the upper (upper layer) side. Further, the user moves the evaluation axis setting bar 9 to the lower (lower layer) side when he / she wants to obtain the analysis result with the finer classification particle size. When the user performs an operation to move the evaluation axis setting bar 9, the CPU 22 sets the intersection position of the evaluation axis setting bar 9 and the vertical line of the dendrogram 8 after the movement as a new setting, and sets the new setting. Based on this, the clustering process is executed, and the analysis result including the process result of the clustering process is output.

The above is the details of this embodiment. According to this embodiment, the following effects can be obtained.
First, in the present embodiment, as shown in FIG. 5, the CPU 22 contains the contents _{of the top d document data D 1} , D ₂ ... D _{d in the search result of one search term to be evaluated.} And the structure features vector data z ₁ = {z ₁₁ , z ₁₂ ... z _1l' }, z ₂ = {z ₂₁ , z ₂₂ ... z _2l' } ... z _d = {z _d1 , z _{Converted to d2} ... z _dl' }, feature vector data z ₁ = {z ₁₁ , z ₁₂ ... z _1l' }, z ₂ = {z ₂₁ , z ₂₂ ... z _2l' } ... -Z _d = {z _d1 , z _d2 ... z _dl' } is subjected to clustering processing, and the document data D ₁ , D ₂ ... D _d are classified into a plurality of subsets (clusters). The CPU 22 outputs an analysis result of the nature of the search needs based on the relationship between a plurality of subsets which are the processing results of clustering of the document data D ₁ , D ₂ ... D _d. Therefore, according to the present embodiment, it is possible to efficiently analyze how much different needs are mixed in the search term and what the nature of the needs is.

Secondly, in the present embodiment, the upper page classification is output as the analysis result. The information on the web page in the upper page classification is color-coded so that the information distributed to the same subset (cluster) by clustering has the same color. In the present embodiment, the degree of variation in the nature of the needs of the search term to be evaluated can be visualized by this upper page classification. According to the present embodiment, in the case of verifying why the upper web page is higher due to the difference between the upper web page and the lower web page in the search result, the nature of the search needs is the same web. You can compare pages. Therefore, in the present embodiment, the upper web page can be verified more efficiently.

Third, in the present embodiment, the dendrogram 8 is output as an analysis result. When the operation to move the evaluation axis setting bar 9 in the denogram 8 is performed, the intersection position between the evaluation axis setting bar 9 and the vertical line of the denogram 8 is set as a new setting, and the clustering process is performed based on this new setting. Execute and output the analysis result including the processing result of the clustering process. Therefore, according to the present embodiment, the user can adjust the particle size of the classification in the upper page classification so as to match his / her intention.

<Second Embodiment>
A second embodiment of the present invention will be described. FIG. 6 is a flowchart showing a flow of an evaluation method executed by the CPU 22 of the search needs evaluation device 20 of the second embodiment according to the evaluation program 26. The CPU 22 executes the evaluation program 26 to execute the acquisition process (S100), the quantification means for executing the quantification process (S200), the addition means for executing the addition process (S210), and the dimension reduction. It functions as a dimension reduction means for executing the process (S300), a classification means for executing the class classification process (S311), and an analysis result output means for executing the analysis result output process (S400). The contents of the acquisition process, the quantification process, the addition process, and the dimension reduction process are the same as those in the first embodiment.

Comparing FIG. 6 with FIG. 2 of the first embodiment, in FIG. 6, the clustering process of step S310 is replaced with the classification process of step S311.

The classification process of step S311 is _{a statistical classification process for classifying document data D 1} , D ₂ ... D _d into a plurality of subsets (lumps) called classes. In the classification process, the CPU 22 uses the feature vector data z ₁ = {z ₁₁ , z ₁₂ ... z _1l' }, z ₂ = {z ₂₁ , z _{22 of the document data D 1} , D ₂ ... D _d. ... z _2l' } ... z _d = {z _d1 , z _d2 ... z _dl' } is processed according to the classification algorithm, and the document data D ₁ , D ₂ ... D _d Is classified into multiple classes.

The details of classification will be explained. In classification, set by the machine learning the weighting coefficients _{w 0, w 1, w 2} ··· w d using the feature vector data group known classes of linear classifier f (z) shown in the following equation (1) Then, in the linear classifier f (z), the feature vector data _{of the document data D 1} , D ₂ ... D _{d z 1} = {z ₁₁ , z ₁₂ ... z _1l' }, z ₂ = {z ₂₁ , z ₂₂ ... z _2l' } ... z _d = {z _d1 , z _d2 ... z _dl' }, and based on this result, the document data D ₁ , D ₂ ... D _d Determine the class of.

f (z) = w ₀ + w ₁ z ₁ + w ₂ z ₂ + ... + w _d z _d ... (1)

FIG. 7A is a diagram showing an example of class classification when the number of classes is two, class A and class B, and the number of dimensions l'is l'= 2. In machine learning, feature vector data group that becomes teacher data (in the example of FIG. 7A, feature vector data group associated with label information indicating that it is class A teacher data, and class B teacher data. Prepare a feature vector data group) associated with the label information indicating that.

Next, the weighting coefficients of the linear classifier f (z) (in the example of FIG. 7A, the two-dimensional linear classifier f (z) = w ₀ + w ₁ z ₁ + w ₂ z _{2) are initialized.} After that, the teacher data is assigned to the linear classifier f (z), and if the assignment result is different from the class indicated by the label information, the weighting coefficient is updated, and if the assignment result matches the class indicated by the label information, the weight coefficient is updated. The process of selecting another teacher data that has not been assigned to the linear classifier f (z) is repeated to optimize the weighting coefficient.

After optimization of the weighting coefficient by the machine learning, CPU 22 may belong document data _{D 1} by substituting the feature vector data _z 1 ₌ document data _{D 1 {z 11, z 12} } to linear classifier f (z) determine the class, the feature vector data _z of the document data _{D 2 2 = {z 21,} z 22} to determine the belonging class document data _{D 2} is substituted into the linear classifier f (z) · · · document data Features of D _{d Document data D 1} , D by substituting the vector data z _d = {z _d1 , z _d2 } into the linear classifier f (z) to determine the class _{to which the document data D n belongs. 2} ... _{Classify D d} into a plurality of classes.

The analysis result output process of step S400 in FIG. 6 is a process of outputting the analysis result of the search needs related to the search term to be evaluated based on the relationship between the classes. As shown in FIG. 6, in the analysis result output processing, the CPU 22 transmits HTML data of the analysis result screen to the user terminal 10 and causes the user terminal 10 to display the analysis result screen on the display. The analysis result screen has a higher page classification. _{The frame F k} (k = 1 to d) of the web page in the upper page classification of FIG. 6 is color-coded so _{that the frame F k of those belonging to the same class has the same color.}

The evaluation axis setting process in step S450 is a process of setting the evaluation axis of the classification process. As shown in FIGS. 7 (B) and 7 (C), the user sets the teacher data of the linear classifier f (z) to another (in the example of FIG. 7 (B), class A, class B1, and The teacher data of class B2, in the example of FIG. 7C, the teacher data of class C and class D) is replaced. When the user performs an operation to replace the teacher data, the CPU 22 optimizes the weight coefficient of the linear classifier f (z) by machine learning using the replaced teacher data, and the linear classifier f (z) uses the linear classifier f (z). The class to which the document data D ₁ , D ₂ ... D _{d belongs is determined.}

The above is the details of this embodiment. In the present embodiment, as shown in FIG. 8, the CPU 22 is characterized by the content and structure of _{the top d document data D 1} , D ₂ ... D _{d in the search result of one search term to be evaluated.} Vector data z ₁ = {z ₁₁ , z ₁₂ ... z _1l' }, z ₂ = {z ₂₁ , z ₂₂ ... z _2l' } ... z _d = {z _d1 , z _d2 ... Converted to z _dl' }, feature vector data z ₁ = {z ₁₁ , z ₁₂ ... z _1l' }, z ₂ = {z ₂₁ , z ₂₂ ... z _2l' } ... z _d = The {z _d1 , z _d2 ... z _dl' } is subjected to the classification process, and the document data D ₁ , D ₂ ... D _d are classified into a plurality of subsets (classes). The CPU 22 outputs an analysis result of the nature of the search needs based on the relationship between a plurality of subsets which are the processing results of the classification of the document data D ₁ , D ₂ ... D _d. The same effect as that of the first embodiment can be obtained by this embodiment as well.

<Third Embodiment>
A third embodiment of the present invention will be described. FIG. 9 is a flowchart showing a flow of an evaluation method executed by the CPU 22 of the search needs evaluation device 20 of the third embodiment according to the evaluation program 26. By executing the evaluation program 26, the CPU 22 executes the acquisition process (S100), the quantification means for executing the quantification process (S200), the addition means for executing the addition process (S210), and the similarity identification. The similarity identification means for executing the process (S320), the community detection means for executing the community detection process (S330), the analysis result output means for executing the analysis result output process (S400), and the evaluation axis setting process (S450) are executed. It functions as an evaluation axis setting means.

Comparing FIG. 9 with FIG. 2 of the first embodiment, in FIG. 9, there is no dimension reduction process in step S330 of FIG. In the present embodiment, document data _{D 1, D} 2 ··· _{D d} of the feature vector data _{z 1 = {z 11, z} 12 ··· z 1l '}, z 2 = {z 21, z 22 ··· With z _2l' } ... z _d = {z _d1 , z _d2 ... z _dl' } as the processing target, the similarity identification processing in step S320 and the community detection processing in step S330 are executed.

The similarity identification process in step S320 is a process for obtaining the similarity between _{document data Dk.} In the similarity identification process, _{the correlation coefficient between the document data D k is} obtained for all combinations of the _{two document data D k in the document data D k} (k = 1 to d), and this correlation coefficient is used as the document data. _{Let it be} the similarity between D k. The correlation coefficient may be a Pearson's correlation coefficient or a correlation coefficient in consideration of sparsity. Further, the variance-covariance matrix between the document data D _k , the Euclidean distance, the Minkowski distance, or the COS similarity may be used as the similarity between the _{document data D k.}

The community detection process in step S330 is _{a statistical classification process for classifying document data D 1} , D ₂ ... D _d into a plurality of subsets called communities. In the community detection process, the CPU 22 uses the feature vector data z ₁ = {z ₁₁ , z ₁₂ ... z _1l' }, z ₂ = {z ₂₁ , z _{22 of the document data D 1} , D ₂ ... D _d. ... z _2l' } ... z _d = {z _d1 , z _d2 ... z _dl' } is processed according to the community detection algorithm, and the document data D ₁ , D ₂ ... D _d Is categorized into multiple communities.

The details of community detection will be explained. Community detection is a type of clustering. In the community detection, each of the document data D ₁ , D ₂ ... D _d is used as a node, and a weighted undirected graph having an edge weighted by the similarity between the _{document data D k is generated.} Then, by repeating the calculation of the mediation centrality of each node in the weighted undirected graph and the removal of the edge having the maximum mediation centrality, the document data D ₁ , D ₂ ... D _d are hierarchically structured. Classify into multiple communities with.

The analysis result output process of step S400 is a process of outputting the analysis result of the search needs related to the search term to be evaluated based on the relationship between the communities. As shown in FIG. 9, in the analysis result output processing, the CPU 22 transmits HTML data of the analysis result screen to the user terminal 10 and causes the user terminal 10 to display the analysis result screen on the display. The analysis result screen has a high-level page classification and a denogram 8. _{The frame F k} (k = 1 to d) of the web page in the upper page classification of FIG. 9 is color-coded so _{that the frame F k of those belonging to the same community has the same color.} The denogram 8 shows the hierarchical structure of the community obtained in the processing process of the community detection process.

The content of the evaluation axis setting process in step S450 is the same as that of the first embodiment.

The above is the details of this embodiment. In the present embodiment, as shown in FIG. 10, the CPU 22 is characterized by the content and structure of _{the top d document data D 1} , D ₂ ... D _{d in the search result of one search term to be evaluated.} Vector data z ₁ = {z ₁₁ , z ₁₂ ... z _1l' }, z ₂ = {z ₂₁ , z ₂₂ ... z _2l' } ... z _d = {z _d1 , z _d2 ... Converted to z _dl' }, feature vector data z ₁ = {z ₁₁ , z ₁₂ ... z _1l' }, z ₂ = {z ₂₁ , z ₂₂ ... z _2l' } ... z _d = The {z _d1 , z _d2 ... z _dl' } is subjected to the processing of similarity identification and community detection, and the document data D ₁ , D ₂ ... D _d are classified into a plurality of subsets (communities). The CPU 22 outputs the analysis result of the nature of the search needs based on the relationship between the plurality of subsets which are the processing results of the community detection of the document data D ₁ , D ₂ ... D _d. The same effect as that of the first embodiment can be obtained by this embodiment as well.

<Fourth Embodiment>
A fourth embodiment of the present embodiment will be described. The search needs evaluation service of the first to third embodiments receives one search term from the user, and classifies the top d web pages in the search result of the search term by a predetermined statistical classification processing algorithm. However, a set of a plurality of web pages obtained by this classification was presented as an analysis result. On the other hand, in the present embodiment, a plurality of search terms A, B, C ... (For example, "AI intelligence", "AI artificial", "AI data") in which a nuclear word and various subwords are combined are used by the user. "... etc.) The upper d document data groups of each of the received and received multiple search terms A, B, C ... are classified by a predetermined statistical classification processing algorithm, and obtained by this classification. The set of a plurality of document data obtained is presented as an analysis result of the nature of the search needs of the search term itself, which is the core word.

FIG. 11 is a flowchart showing a flow of an evaluation method executed by the CPU 22 of the search needs evaluation device 20 of the fourth embodiment according to the evaluation program 26. The CPU 22 executes the evaluation program 26 to execute the acquisition process (S100), the quantification means for executing the quantification process (S200), the addition means for executing the addition process (S210), and the synthesis process (S210). Functions as a synthesis means for executing S250), a dimension reduction means for executing dimension reduction processing (S300), a classification means for executing clustering processing (S310), and an analysis result output means for executing analysis result output processing (S401). do.

Comparing FIG. 11 with FIG. 2 of the first embodiment, in FIG. 11, in the acquisition process of step S100, the CPU 22 receives a plurality of search terms A, B, C ... From the user terminal 10. For each of the plurality of search terms A, B, C ..., the document data of the top d web pages in the search results for each search term D _Ak (k = 1 to d), D _Bk (k = 1 to 1). _d), to obtain a D Ck (k = 1~d) ··· . Thereafter, CPU 22 is document data _D Ak _(k = _1~d) for each search _{term, D Bk (k = 1~d)} , the D Ck (k = 1~d) ··· , Determination of step S200 _{The feature vector data z A1} = {z _A11 , z _A12 ... z _A1l }, z _A2 = {z _A21, which is the processing result of the higher-level document of the search term A by executing the conversion process and the addition process of step S210. , Z _A22 ... z _A2l } ... z _Ad = {z _Ad1 , z _Ad2 ... z _{Adl }, Feature vector data z B1} = {z _B11 , which is the processing result of the higher-level document of the search term B. z _B12 ... z _B1l }, z _B2 = {z _B21 , z _B22 ... z _B2l } ... z _Bd = {z _Bd1 , z _Bd2 ... z _Bdl } Feature vector data z _C1 = {z _C11 , z _C12 ... z _C1l }, z _C2 = {z _C21 , z _C22 ... z _C2l } ... z _Cd = {z _Cd1 , z _Cd2 ... z _Cdl } ... Are generated individually.

In FIG. 11, there is a synthesis process of step S250 between the addition process of step S210 and the dimension reduction process of step S300. In the synthesis process, the CPU 22 uses the higher-level document feature vector data z _A1 = {z _A11 , z _A12 ... z _A1l }, z _A2 = {z _A21 , z _A22 ... z _A2l } ... z _Ad = {z _Ad1 , z _Ad2 ... z _Adl }, higher-level document feature vector data of search term B z _B1 = {z _B11 , z _B12 ... z _B1l }, z _B2 = {z _B21 , z _B22 ... z _B2l } ... z _Bd = {z _Bd1 , z _Bd2 ... z _Bdl }, higher-level document feature vector data of search term C z _C1 = {z _C11 , z _C12 ... z _C1l }, z _C2 = {z _C21 , z _C22 ... z _C2l } ... z _Cd = {z _Cd1 , z _Cd2 ... z _Cdl } ... Feature vector data z _A1 = {z _A11 , z _A12 ... z _A1l }, z _A2 = {z _A21 , z _A22 ... z _A2l } ... z _Ad = {z _Ad1 , z _Ad2 ... z _{Feature vector data z A} = {z _A1 , z _A2 ... z _Al }, which is a composite of _Adl }, higher-level document feature vector data z _B1 = {z _B11 , z _B12 ... z _B1l }, z _{B2 = {z B21, z B22} ··· z B2l} ··· z Bd = {z Bd1, z Bd2 ··· z Bdl} the combined feature vector data _{z B = {z B1, z} B2 ··· z _Bl }, higher-level document feature vector data of search term C z _C1 = {z _C11 , z _C12 ... z _C1l }, z _C2 = {z _C21 , z _C22 ... z _C2l } ... z _Cd = Feature vector data z _C = {z _C1 , z _C2 ... z _{Cl } ... Combining} {z _Cd1 , z _Cd2 ... z Cdl} ... is individually generated.

After that, the CPU 22 uses the feature vector data z _A = {z _A1 , z _A2 ... z _Al' } of the search term A, and the feature vector data z _B = {z _B1 , z _B2 ... z of the search term B. _Bl' }, feature vector data of search term C z _C = {z _C1 , z _C2 ... z _Cl' } ... The clustering process in step S310 and the analysis result output process in step S401 are performed. Run. That is, in the present embodiment, all documents are collectively clustered instead of clustering for each search term.

In the analysis result output process of step S401 of FIG. 11, the analysis result screen is displayed on the display of the user terminal 10. The analysis result screen has a mapping image 7. Mapping image 7, a two-dimensional plane, in which a plurality of search terms A, B, and C marks _MK 1 indicating the location of each of the _···, MK 2 ··· MK _L arranged. The mapping image 7 is generated based on the processing results of steps S250, S300, and S310.

The above is the details of this embodiment. In the present embodiment, as shown in FIG. 12, the CPU 22 has the top d document data D in the search result for each search term for each of the plurality of search terms A, B, C ... _{Ak (k = 1~d), D} Bk (k = 1~d), D Ck (k = 1~d) to get the ..., document data _D Ak in the search results for each search term (k = _{1~d), D Bk (k =} 1~d), D Ck (k = 1~d) the contents and structure of ... multidimensional feature vector data _{z A1, z A2 ··· z Ad} , z _{Converted to B1} , z _B2 ... z _Bd , z _C1 , z _C2 ... z _Cd ..., the feature vector data for each document is subjected to predetermined statistical processing, and the feature vector data for each search term is synthesized. do. Then, the synthesized feature vector data z _A , z _B , z _C ... are subjected to clustering processing, and the search term A, search term B, search term C ... are combined into a plurality of subsets (clusters). The mapping image 7 which is the analysis result of the nature of the search need is output based on the relationship between a plurality of subsets which are the result of classification and the clustering process. Therefore, according to the present embodiment, by referring to the mapping image 7, it is possible to intuitively grasp how close the nature of the search needs related to various search terms including common words is. Therefore, even in this embodiment, it is possible to efficiently analyze how much different needs are mixed in the words of the search term and what the nature of the needs is.

<Fifth Embodiment>
A fifth embodiment of the present invention will be described. FIG. 13 is a flowchart showing a flow of an evaluation method executed by the CPU 22 of the search needs evaluation device 20 of the fifth embodiment according to the evaluation program 26. By executing the evaluation program 26, the CPU 22 executes the acquisition process (S100), the quantification means for executing the quantification process (S200), the addition means for executing the addition process (S210), and the dimension reduction. Functions as a quantification means for executing the process (S300), a classification means for executing the clustering process (S310), a synthesis means for executing the synthesis process (S350), and an analysis result output means for executing the analysis result output process (S401). do.

Comparing FIG. 13 with FIG. 11 of the fourth embodiment, in FIG. 13, there is no synthesis process of step S250 of FIG. 11, and there is a synthesis process of step S350 between steps S310 and S401. In the present embodiment, the CPU 22 uses the higher-level document feature vector data z _A1 = {z _A11 , z _A12 ... z _A1l }, z _A2 = {z _A21 , z _A22 ... z _A2l } of the search term A ... · Z _Ad = {z _Ad1 , z _Ad2 ... z _Adl }, higher-level document feature vector data of search term B z _B1 = {z _B11 , z _B12 ... z _B1l }, z _B2 = {z _B21 , z _B22 ... z _B2l } ... z _Bd = {z _Bd1 , z _Bd2 ... z _Bdl }, higher-level document feature vector data of search term C z _C1 = {z _C11 , z _C12 ... z _C1l } , Z _C2 = {z _C21 , z _C22 ... z _C2l } ... z _Cd = {z _Cd1 , z _Cd2 ... z _Cdl } ... clustering process of step S310 is executed, the document data _{D Ak (k = 1~d),} D Bk (k = 1~d), the D Ck (k = 1~d) processing result of the clustering process.. get. In the synthesis process of step S350, the CPU 22 performs a predetermined statistical process on the clustering process result for each document, and synthesizes the clustering process result for each search term.

In the analysis result output process of step S401 of FIG. 13, the analysis result screen is displayed on the display of the user terminal 10. The mapping image 7 of the analysis result screen of FIG. 19 is generated based on the processing results of steps S300, S310, and S350.

The above is the details of the configuration of the present embodiment. In the present embodiment, as shown in FIG. 14, the CPU 22 has the top d document data D in the search result for each search term for each of the plurality of search terms A, B, C ... _{Ak (k = 1~d), D} Bk (k = 1~d), D Ck (k = 1~d) to get the ..., document data _D Ak in the search results for each search term (k = _{1~d), D Bk (k =} 1~d), D Ck (k = 1~d) the contents and structure of ... multidimensional feature vector data _{z A1, z A2 ··· z Ad} , z _{Converted to B1} , z _B2 ... z _Bd , z _C1 , z _C2 ... z _Cd ..., and processed the feature vector data for each document according to the clustering algorithm to generate multiple document data. Classify into a subset of. After that, the clustering processing result is subjected to predetermined statistical processing, the clustering processing result for each search term is synthesized, and the analysis result of the nature of the search needs is output based on the relationship between the synthesized subsets. .. The same effect as that of the fourth embodiment can be obtained by this embodiment as well.

<Sixth Embodiment>
A sixth embodiment of the present embodiment will be described. FIG. 15 is a flowchart showing a flow of an evaluation method executed by the CPU 22 of the search needs evaluation device 20 of the sixth embodiment according to the evaluation program 26. By executing the evaluation program 26, the CPU 22 executes the acquisition process (S100), the quantification means for executing the quantification process (S200), the addition means for executing the addition process (S210), and the synthesis process (S210). As a synthesis means for executing S250), a dimension reduction means for executing dimension reduction processing (S300), a classification means for executing classification processing (S311), and an analysis result output means for executing analysis result output processing (S401). Function.

Comparing FIG. 15 with FIG. 6 of the second embodiment, in FIG. 15, in the acquisition process of step S100, the CPU 22 receives a plurality of search terms A, B, C ... From the user terminal 10. For each of the plurality of search terms A, B, C ..., the document data of the top d web pages in the search results for each search term D _Ak (k = 1 to d), D _Bk (k = 1 to 1). _d), to obtain a D Ck (k = 1~d) ··· . Thereafter, CPU 22 is document data _D Ak _(k = _1~d) for each search _{term, D Bk (k = 1~d)} , the D Ck (k = 1~d) ··· , Determination of step S200 _{The feature vector data z A1} = {z _A11 , z _A12 ... z _A1l }, z _A2 = {z _A21, which is the processing result of the higher-level document of the search term A by executing the conversion process and the addition process of step S210. , Z _A22 ... z _A2l } ... z _Ad = {z _Ad1 , z _Ad2 ... z _{Adl }, Feature vector data z B1} = {z _B11 , which is the processing result of the higher-level document of the search term B. z _B12 ... z _B1l }, z _B2 = {z _B21 , z _B22 ... z _B2l } ... z _Bd = {z _Bd1 , z _Bd2 ... z _Bdl } Feature vector data z _C1 = {z _C11 , z _C12 ... z _C1l }, z _C2 = {z _C21 , z _C22 ... z _C2l } ... z _Cd = {z _Cd1 , z _Cd2 ... z _Cdl } ... Are generated individually.

In FIG. 15, there is a synthesis process of step S250 between the addition process of step S210 and the dimension reduction process of step S300. In the synthesis process, the CPU 22 uses the higher-level document feature vector data z _A1 = {z _A11 , z _A12 ... z _A1l }, z _A2 = {z _A21 , z _A22 ... z _A2l } ... z _Ad = {z _Ad1 , z _Ad2 ... z _Adl }, higher-level document feature vector data of search term B z _B1 = {z _B11 , z _B12 ... z _B1l }, z _B2 = {z _B21 , z _B22 ... z _B2l } ... z _Bd = {z _Bd1 , z _Bd2 ... z _Bdl }, higher-level document feature vector data of search term C z _C1 = {z _C11 , z _C12 ... z _C1l }, z _C2 = {z _C21 , z _C22 ... z _C2l } ... z _Cd = {z _Cd1 , z _Cd2 ... z _Cdl } ... Feature vector data z _A1 = {z _A11 , z _A12 ... z _A1l }, z _A2 = {z _A21 , z _A22 ... z _A2l } ... z _Ad = {z _Ad1 , z _Ad2 ... z Feature vector data of search term A synthesized from _{Adl } z A} = {z _A1 , z _A2 ... z _Al }, higher-level document feature vector data of search term B z _B1 = {z _B11 , z _B12 ... z _{B1l}, z B2 = {z} B21, z B22 ··· z B2l} ··· z Bd = {z Bd1, z Bd2 ··· z Bdl} feature vector a synthesized search word B data _z B = { z _B1 , z _B2 ... z _Bl }, higher-level document feature vector data of search term C z _C1 = {z _C11 , z _C12 ... z _C1l }, z _C2 = {z _C21 , z _C22 ... z _C2l } ・ ・ ・ z _Cd = {z _Cd1 , z _Cd2・ ・ ・ z _Cdl } ・ ・ ・ Characteristic vector data of search term C synthesized z _C = {z _C1 , z _C2・ ・ ・ z _Cl } ・ ・ ・To generate.

After that, the CPU 22 uses the feature vector data z _A = {z _A1 , z _A2 ... z _Al' } of the search term A, and the feature vector data z _B = {z _B1 , z _B2 ... z of the search term B. _Bl' }, feature vector data of search term C z _C = {z _C1 , z _C2 ... z _Cl' } ..., the classification process of step S311 and the analysis result output process of step S401. To execute. That is, in the present embodiment, all documents are collectively classified, instead of being classified for each search term.

In the analysis result output process of step S401 of FIG. 15, the analysis result screen is displayed on the display of the user terminal 10. The mapping image 7 of the analysis result screen of FIG. 15 is generated based on the processing results of steps S250, S300, and S311.

The above is the details of this embodiment. In the present embodiment, as shown in FIG. 16, the CPU 22 has the top d document data D in the search result for each search term for each of the plurality of search terms A, B, C ... _{Ak (k = 1~d), D} Bk (k = 1~d), D Ck (k = 1~d) to get the ..., document data _D Ak in the search results for each search term (k = _{1~d), D Bk (k =} 1~d), D Ck (k = 1~d) the contents and structure of ... multidimensional feature vector data _{z A1, z A2 ··· z Ad} , z _{Converted to B1} , z _B2 ... z _Bd , z _C1 , z _C2 ... z _Cd ..., the feature vector data for each document is subjected to predetermined statistical processing, and the feature vector data for each search term is synthesized. do. Then, the synthesized feature vector data z _A , z _B , z _C ... Is subjected to classification processing, and the search terms A, B, C ... Are classified into a plurality of subsets (classes). The analysis result of the nature of the search needs is output based on the relationship between a plurality of subsets which is the processing result of the classification. The same effect as that of the fourth embodiment can be obtained by this embodiment as well.

<7th Embodiment>
A seventh embodiment of the present invention will be described. FIG. 17 is a flowchart showing a flow of an evaluation method executed by the CPU 22 of the search needs evaluation device 20 of the seventh embodiment according to the evaluation program 26. The CPU 22 executes the evaluation program 26 to execute the acquisition process (S100), the quantification means for executing the quantification process (S200), the addition means for executing the addition process (S210), and the dimension reduction. As a dimension reduction means for executing the process (S300), a classification means for executing the class classification process (S311), a synthesis means for executing the synthesis process (S350), and an analysis result output means for executing the analysis result output process (S401). Function.

Comparing FIG. 17 with FIG. 15 of the sixth embodiment, in FIG. 17, there is no synthesis process of step S250 of FIG. 15, and there is a synthesis process of step S350 between steps S311 and S401. In the present embodiment, the CPU 22 uses the higher-level document feature vector data z _A1 = {z _A11 , z _A12 ... z _A1l }, z _A2 = {z _A21 , z _A22 ... z _A2l } of the search term A ... · Z _Ad = {z _Ad1 , z _Ad2 ... z _Adl }, higher-level document feature vector data of search term B z _B1 = {z _B11 , z _B12 ... z _B1l }, z _B2 = {z _B21 , z _B22 ... z _B2l } ... z _Bd = {z _Bd1 , z _Bd2 ... z _Bdl }, higher-level document feature vector data of search term C z _C1 = {z _C11 , z _C12 ... z _C1l } , Z _C2 = {z _C21 , z _C22 ... z _C2l } ... z _Cd = {z _Cd1 , z _Cd2 ... z _Cdl } ... run the classification processing in step S311, the document data _{D Ak (k = 1~d),} D Bk (k = 1~d), D Ck (k = 1~d) processing classification process ... Get the result. In the synthesis process of step S350, the CPU 22 performs a predetermined statistical process on the processing result of the class classification for each document, and synthesizes the processing result of the class classification for each search term.

In the analysis result output process of step S401 of FIG. 17, the analysis result screen is displayed on the display of the user terminal 10. The mapping image 7 of the analysis result screen of FIG. 17 is generated based on the processing results of steps S300, S311, and S350.

The above is the details of the configuration of the present embodiment. In the present embodiment, as shown in FIG. 18, the CPU 22 has the top d document data D in the search result for each search term for each of the plurality of search terms A, B, C ... _{Ak (k = 1~d), D} Bk (k = 1~d), D Ck (k = 1~d) to get the ..., document data _D Ak in the search results for each search term (k = _{1~d), D Bk (k =} 1~d), D Ck (k = 1~d) the contents and structure of ... multidimensional feature vector data _{z A1, z A2 ··· z Ad} , z _{Converted to B1} , z _B2 ... z _Bd , z _C1 , z _C2 ... z _Cd ..., processed the feature vector data for each document according to the classification algorithm, and searched for each search term. Classify multiple document data in the result into multiple subsets. After that, the processing result of the classification is subjected to predetermined statistical processing, the processing result of the classification for each search term is synthesized, and the analysis result of the nature of the search needs is analyzed based on the relationship between the synthesized subsets. Output. The same effect as that of the fourth embodiment can be obtained by this embodiment as well.

<8th Embodiment>
An eighth embodiment of the present embodiment will be described. FIG. 19 is a flowchart showing a flow of an evaluation method executed by the CPU 22 of the search needs evaluation device 20 of the eighth embodiment according to the evaluation program 26. By executing the evaluation program 26, the CPU 22 executes the acquisition process (S100), the quantification means for executing the quantification process (S200), the addition means for executing the addition process (S210), and the synthesis process (S210). Synthesis means for executing S250), similarity identification means for executing similarity identification processing (S320), community detection means for executing community detection processing (S330), analysis result output means for executing analysis result output processing (S401). Functions as.

Comparing FIG. 19 with FIG. 9 of the third embodiment, in FIG. 19, in FIG. 19, in the acquisition process of step S100, the CPU 22 receives a plurality of search terms A, B, C ... -Received, and for each of the plurality of search terms A, B, C ..., the document data of the top d web pages in the search results for each search term D _Ak (k = 1 to d), D _Bk ( _{k =} 1~d), to get the D Ck (k = 1~d) ··· . Thereafter, CPU 22 is document data _D Ak _(k = _1~d) for each search _{term, D Bk (k = 1~d)} , the D Ck (k = 1~d) ··· , Determination of step S200 _{The feature vector data z A1} = {z _A11 , z _A12 ... z _A1l }, z _A2 = {z _A21, which is the processing result of the higher-level document of the search term A by executing the conversion process and the addition process of step S210. , Z _A22 ... z _A2l } ... z _Ad = {z _Ad1 , z _Ad2 ... z _{Adl }, Feature vector data z B1} = {z _B11 , which is the processing result of the higher-level document of the search term B. z _B12 ... z _B1l }, z _B2 = {z _B21 , z _B22 ... z _B2l } ... z _Bd = {z _Bd1 , z _Bd2 ... z _Bdl } Feature vector data z _C1 = {z _C11 , z _C12 ... z _C1l }, z _C2 = {z _C21 , z _C22 ... z _C2l } ... z _Cd = {z _Cd1 , z _Cd2 ... z _Cdl } ... Are generated individually.

In FIG. 19, there is a synthesis process of step S250 between the addition process of step S210 and the dimension reduction process of step S300. In the synthesis process, the CPU 22 uses the higher-level document feature vector data z _A1 = {z _A11 , z _A12 ... z _A1l }, z _A2 = {z _A21 , z _A22 ... z _A2l } ... z _Ad = {z _Ad1 , z _Ad2 ... z _Adl }, higher-level document feature vector data of search term B z _B1 = {z _B11 , z _B12 ... z _B1l }, z _B2 = {z _B21 , z _B22 ... z _B2l } ... z _Bd = {z _Bd1 , z _Bd2 ... z _Bdl }, higher-level document feature vector data of search term C z _C1 = {z _C11 , z _C12 ... z _C1l }, z _C2 = {z _C21 , z _C22 ... z _C2l } ... z _Cd = {z _Cd1 , z _Cd2 ... z _Cdl } ... Feature vector data z _A1 = {z _A11 , z _A12 ... z _A1l }, z _A2 = {z _A21 , z _A22 ... z _A2l } ... z _Ad = {z _Ad1 , z _Ad2 ... z Feature vector data of search term A synthesized from _{Adl } z A} = {z _A1 , z _A2 ... z _Al }, higher-level document feature vector data of search term B z _B1 = {z _B11 , z _B12 ... z _{B1l}, z B2 = {z} B21, z B22 ··· z B2l} ··· z Bd = {z Bd1, z Bd2 ··· z Bdl} feature vector a synthesized search word B data _z B = { z _B1 , z _B2 ... z _Bl }, higher-level document feature vector data of search term C z _C1 = {z _C11 , z _C12 ... z _C1l }, z _C2 = {z _C21 , z _C22 ... z _C2l } ・ ・ ・ z _Cd = {z _Cd1 , z _Cd2・ ・ ・ z _Cdl } ・ ・ ・ Characteristic vector data of search term C synthesized z _C = {z _C1 , z _C2・ ・ ・ z _Cl } ・ ・ ・To generate.

After that, the CPU 22 uses the feature vector data z _A = {z _A1 , z _A2 ... z _Al } of the search term A, and the feature vector data z _B = {z _B1 , z _B2 ... z _{Bl of the search term B.} }, Feature vector data of search term C z _C = {z _C1 , z _C2 ... z _Cl } ..., Similarity identification processing in step S320, community detection processing in step S330, and step S401. Executes the analysis result output processing of. That is, in the present embodiment, the similarity is specified and the community is detected for each search term, but all the documents are collectively specified and the community is detected.

In the analysis result output process of step S401 of FIG. 19, the analysis result screen is displayed on the display of the user terminal 10. The mapping image 7 of the analysis result screen of FIG. 19 is generated based on the processing results of steps S250, S320, and S330.

The above is the details of this embodiment. In the present embodiment, as shown in FIG. 20, the CPU 22 has the top d document data D in the search result for each search term for each of the plurality of search terms A, B, C ... _{Ak (k = 1~d), D} Bk (k = 1~d), D Ck (k = 1~d) to get the ..., document data _D Ak in the search results for each search term (k = _{1~d), D Bk (k =} 1~d), D Ck (k = 1~d) the contents and structure of ... multidimensional feature vector data _{z A1, z A2 ··· z Ad} , z _{Converted to B1} , z _B2 ... z _Bd , z _C1 , z _C2 ... z _Cd ..., the feature vector data for each document is subjected to predetermined statistical processing, and the feature vector data for each search term is synthesized. do. Then, the synthesized feature vector data z _A , z _B , z _C ... are processed for similarity identification and community detection, and the search terms A, B, C ... Are classified into a plurality of communities. Outputs the analysis result of the nature of the search needs based on the relationship between multiple communities, which is the processing result of community detection. The same effect as that of the fourth embodiment can be obtained by this embodiment as well.

<9th embodiment>
A ninth embodiment of the present invention will be described. FIG. 21 is a flowchart showing a flow of an evaluation method executed by the CPU 22 of the search needs evaluation device 20 of the ninth embodiment according to the evaluation program 26. By executing the evaluation program 26, the CPU 22 executes the acquisition process (S100), the quantification means for executing the quantification process (S200), the addition means for executing the addition process (S210), and the similarity identification. Similarity specifying means for executing processing (S320), community detecting means for executing community detection processing (S330), synthesis means for executing synthesis processing (S350), analysis result output means for executing analysis result output processing (S401). Functions as.

Comparing FIG. 21 with FIG. 19 of the eighth embodiment, in FIG. 21, there is no synthesis process of step S250 of FIG. 19, and there is a synthesis process of step S350 between steps S330 and S401. In the present embodiment, the CPU 22 uses the feature vector data z _A1 = {z _A11 , z _A12 ... z _A1l }, z _A2 = {z _A21 , z _A22 ... z _A2l } of the higher-level document of the search term A. _.. z _Ad = {z Ad1, z _Ad2 ... z _Adl }, feature vector data of the higher-level document of the search term B z _B1 = {z _B11 , z _B12 ... z _B1l }, z _B2 = {z _B21 , Z _B22 ... z _B2l } ... z _Bd = {z _Bd1 , z _Bd2 ... z _Bdl }, Feature vector data of higher-level document of search term C z _C1 = {z _C11 , z _C12 ... z _C1l }, z _C2 = {z _C21 , z _C22 ... z _C2l } ... z _Cd = {z _Cd1 , z _Cd2 ... z _Cdl } ... run the community detection processing of a particular process and step S330, the document data _{D Ak (k = 1~d),} D Bk (k = 1~d), D Ck (k = 1~d) ··· community detection Get the processing result of the processing. In the synthesis process of step S350, the CPU 22 performs a predetermined statistical process on the community detection process result for each document, and synthesizes the community detection process result for each search term.

In the analysis result output process of step S401 of FIG. 21, the analysis result screen is displayed on the display of the user terminal 10. The mapping image 7 of the analysis result screen of FIG. 21 is generated based on the processing results of steps S320, S330, and S350.

The above is the details of the configuration of the present embodiment. In the present embodiment, as shown in FIG. 14, the CPU 22 has the top d document data D in the search result for each search term for each of the plurality of search terms A, B, C ... _{Ak (k = 1~d), D} Bk (k = 1~d), D Ck (k = 1~d) to get the ..., document data _D Ak in the search results for each search term (k = _{1~d), D Bk (k =} 1~d), D Ck (k = 1~d) the contents and structure of ... multidimensional feature vector data _{z A1, z A2 ··· z Ad} , z _{Converted to B1} , z _B2 ... z _Bd , z _C1 , z _C2 ... z _Cd ..., and subjected to similarity identification processing and community detection processing on the feature vector data for each document, and multiple document data. Is categorized into multiple communities. Then, the processing result is subjected to predetermined statistical processing, the processing result for each search term is synthesized, and the analysis result of the nature of the search needs is output based on the relationship between the synthesized communities. The same effect as that of the fourth embodiment can be obtained by this embodiment as well.

<10th Embodiment>
In the tenth embodiment, a display example of the analysis result using the weighted undirected graph will be specifically described.

FIG. 25 is a diagram showing the mapping image 7 of FIG. 11 more specifically. This mapping image 7 exemplifies the analysis result regarding the search term including the common word "ABC". It is assumed that there is a technical term "ABC", an electronic file extension "ABC", and a singer "ABC".

The mapping image 7 of FIG. 25 shows an analysis result as a graph (undirected graph) including nodes (for example, reference numerals n1 and n2) and edges connecting the nodes (for example, reference numeral e). Each search term is associated with the node. The edge length corresponds to the similarity of the search needs between the search term associated with the node at one end and the search term associated with the node at the other end. Specifically, the higher the similarity between one search term and another search term, the shorter the edge. Therefore, the nodes associated with the search terms having high similarities in the search needs are arranged close to each other. If the similarity between the two search terms is lower than the predetermined value, the edge between the nodes associated with both search terms may be omitted.

Here, the similarity may be the one described above in, for example, the eighth embodiment, or may be calculated by another method based on the search result for the search term.

By displaying in this way, highly relevant search terms become clear at a glance. According to FIG. 25, "ABC seminar", "ABC business", and "ABC venture" are highly related, "ABC live", "ABC album", and "ABC concert" are highly related, and "ABC". It can be seen that the "extension", "ABC data", and "ABC file" are highly related. This is because websites visited with the search term "ABC Seminar" are often visited with the search term "ABC Business" or "ABC Venture", but other "ABC Live" or "ABC Data" It means that it is rarely visited by the search term.

For example, when trying to create a web page related to the technology "ABC", create a web page with the search terms such as "ABC seminar", "ABC business", and "ABC venture" in mind. You just have to do it.

Further, in the undirected graph shown in FIG. 25, the user may be able to move the node. To move a node, for example, a method of clicking a desired node with a mouse or tapping a desired node with a touch panel to select a node and dragging the selected node to an arbitrary other location can be considered.

FIG. 26 is a diagram showing a state in which the node n3 associated with the “ABC business” in FIG. 25 is moved.

As the node n3 is moved by the user operation, other nodes (nodes n4 and n5 in FIG. 26) that are at least close to the node n3 (similarity is equal to or higher than a predetermined value) are automatically moved so as to be attracted to the node n3. Is good. At this time, the length of the edge is determined by a mechanical model such as a spring or a Coulomb force. Specifically, when the edge is pulled by the movement of the node, the edge is stretched, the pulling force is strengthened by the stretched force, and the force converges to a shortness in which the force can be balanced over time.

Although only a small number of nodes (search terms) are drawn in FIGS. 25 and 26, a large number of nodes (search terms) are actually displayed. Therefore, in some cases, nodes may be concentrated in one place. In this case, by moving the node associated with the search term of interest to an arbitrary location, the search term having a high degree of similarity can be displayed more easily.

FIG. 27 is a diagram showing a mapping image 7 in which search terms are classified into clusters and nodes are displayed in a display mode according to the classified clusters. For cluster classification, for example, the method described above may be applied in the fourth embodiment, or another method based on the search result for the search term may be applied. In addition, in FIG. 27 and the like, the search term itself is omitted.

The figure shows an example in which each search term is classified into any one of two clusters A, B, and C. Nodes associated with search terms classified in cluster A are displayed in black, nodes associated with search terms classified in cluster B are displayed in white, and nodes associated with search terms classified in cluster C are associated. It is displayed as a diagonal line. In addition, color coding may be performed according to the cluster.

FIG. 28 is a diagram showing a mapping image 7 in a case where the search term is not determined to be classified into one cluster but can be classified into a plurality of clusters. It is calculated how close each search term is to which cluster (how much the properties of which cluster are). In the example of FIG. 28, it is determined that a certain search term is 60% for cluster A, 30% for cluster B, and 10% for cluster C. In this case, the node n6 to which the search term is associated is displayed with 60% black, 30% white, and 10% diagonal lines, as in the pie chart.

Further, as described in the first embodiment, the particle size of the classification can be made finer or coarser. The finer the particle size, the more clusters are classified. Then, the user may be able to variably set this particle size.

FIG. 29 is a diagram showing a mapping image 7 in which the user can set the particle size. A slide bar 30 extending in the horizontal direction is displayed, and the user can set the grain size coarser by moving the bar 31 to the left and finer grain size by moving the bar 31 to the right. The particle size may be a plurality of stages, and the number of stages is not particularly limited.

FIG. 29 shows a state in which the particle size is coarsely set. In this example, each search term is classified into any one of two clusters A and B, and there are two types of node display modes (black and diagonal lines in the order of A and B).

FIG. 30 is a diagram showing a state in which the particle size is set more finely than in FIG. 29. In this example, each search term is classified into any one of four cluster rasters A1, A2, B1, and B2. Clusters A1 and A2 are further classified into cluster A, and clusters B1 and B2 are further classified into cluster B. In this case, there are four types of node display modes (black, white, diagonal lines, and wavy lines in the order of A1, A2, B1, B2).

In this way, each time the particle size is set (changed) according to the user operation, each search term is classified into a cluster according to the set particle size. Then, when the cluster into which each search term is classified changes, the display mode of the node is also automatically updated.

For example, when creating a Web page related to the technology "ABC" in general, it is possible to grasp a wide range of relatively highly relevant search terms by setting the particle size coarsely. On the other hand, when creating a Web page specialized in a more specific technology among the technologies called "ABC", it is possible to grasp a small number of highly relevant search terms with high accuracy by setting the particle size finely.

The interface for adjusting the particle size is not limited to the slide bar 30 shown in FIGS. 29 and 30. As shown in FIG. 31, a slide bar 30 extending in the vertical direction may be used. As shown in FIG. 32, a field 32 may be provided in which the user inputs a numerical value indicating the particle size. As shown in FIG. 33, the user may select the button (icon) 33 showing the particle size. The user may select from the pull-down 34 as shown in FIG. 34 or the radio button 35 as shown in FIG. 35. Other interfaces not illustrated may be used, but preferably an interface in which the user can selectively select one of a plurality of steps is preferable.

Further, the number of searches for each search term may be shown on the mapping screen 7.
FIG. 36 is a diagram showing a mapping image 7 in which nodes are displayed in an manner corresponding to the number of searches for each search term. The larger the number of searches for the search term associated with the node, the larger the node will be displayed. It is easy and intuitive to understand that the search term associated with the large node should be emphasized. The number of searches may be the number of searches in an arbitrary period (for example, the latest one month). Of course, the user may be able to set the period variably, for example, it may be possible to compare what kind of change has occurred between the last one month and two months ago.

By combining each of the above examples, the node corresponding to a certain search term may be displayed in a mode corresponding to the cluster in which the search term is classified and in a size corresponding to the number of searches of the search term. good. In addition, other additional information may be added to the undirected graph.

As described above, in the present embodiment, the analysis result for the search term is displayed as an undirected graph. Therefore, the user can intuitively understand the analysis results such as the similarity between the search terms and how they are clustered, and it becomes easy to select the search terms to be targeted.

<11th Embodiment>
The following is a modified example of the display mode of the analysis result.

FIG. 37 is a diagram showing a screen example when the analysis result is displayed in a table format. Each search term is classified into one of four clusters A to D, and the search term classified into each cluster is displayed in a table format associated with the cluster. In the figure, it can be seen that, for example, the search terms a to c are classified into cluster A.

Again, it is desirable for the user to be able to adjust the particle size. For example, in FIG. 37, the clusters were classified into four clusters, but when the user coarsens the particle size using the slide bar 30, the clusters E and F are classified and displayed as shown in FIG. 38. Similar to the case of the undirected graph, but each time the particle size is set (changed) according to the user operation, each search term is classified into clusters according to the set particle size. Then, when the cluster into which each search term is classified changes, the table is automatically updated.

Further, as shown in FIGS. 37 and 38, the number of searches may be associated with each search term and displayed. In this case, it is desirable that the search term having a large number of searches is placed above.

FIG. 39 is a diagram showing a screen example when the analysis result is displayed in the correlation matrix format. The search terms a to d are arranged side by side in the vertical direction and the horizontal direction. Then, the similarity between the search terms is shown in the cell at the intersection of the vertical direction and the horizontal direction. As the degree of similarity, a numerical value may be displayed in the cell, or the cell may be arranged according to the degree of similarity (the higher the degree of similarity, the darker the cell, etc. In FIG. 39, the density of the spots indicates the pseudo-concentration). It may be displayed with. Further, the number of searches may be associated with each search term and displayed.

Further, the user may change the order of the search terms. As an example, when the user selects a desired search term, the selected search term may be placed at the top, and other search terms may be placed from the top in descending order of similarity with the search term. It is assumed that the user selects the search term c in FIG. 39. In that case, as shown in FIG. 40, the search term c is arranged at the highest level, and the search terms b, d, and a are arranged below the search term c in descending order of similarity with the search term c.

FIG. 41 is a diagram showing a screen example when the analysis result is displayed in the dendrogram format. The search terms are arranged vertically, and the search terms with high similarity are placed close to each other. Then, it is shown that the search terms are gradually classified into clusters toward the right (direction away from the search terms).

In order to make the stepwise cluster classification easier to see, as in FIG. 4, a particle size setting bar (evaluation axis setting bar) 36 extending in a direction orthogonal to the dendrogram (vertical direction, direction in which search terms are lined up) is provided on the dendrogram. It is desirable to be displayed. The user can move the particle size setting bar 36 left and right, and the particle size becomes coarser as the particle size setting bar 36 is moved to the right (the farther away from the search term).

For example, when the particle size setting bar 36 is moved to the position shown in FIG. 41, the search term is classified into any of the three clusters A, B, and C, and when the particle size setting bar 36 is moved to the position shown in FIG. 42, the search term is classified into one of the three clusters A, B, and C. The search term is classified into one of two clusters D and E.

As shown in FIGS. 41 and 42, the number of searches may be associated with each search term and displayed. Further, the dendrogram may have search terms arranged in the horizontal direction. Further, although the particle size setting bar 36 is intuitive for setting the particle size, the particle size may be set by another interface as described in the tenth embodiment.

FIG. 43 is a diagram showing a screen example when the analysis result is displayed in the tree map format. Each search term a to n is classified into any of four clusters A to D. A cluster in which one rectangular cell corresponds to one search term, and the display mode of the cell (for example, the color of the cell. In the figure, the spot, the diagonal line, and the wavy line indicate the pseudo color). Indicates that the cell area indicates the number of searches in a predetermined period.

FIG. 44 is a diagram showing a screen example when the analysis result is displayed in the sunburst format. One Baumkuchen-type cell on the outermost side corresponds to each of the search terms a to h. The cells inside show the clusters in which each search term is classified, and the inside of the same layer is the clusters with the same particle size. For example, the innermost layer has three coarse-grained clusters A to C, search terms a to e are classified into cluster A, search terms f and g are classified into cluster B, and search term h is classified into cluster C. It is classified. The second layer from the inside has clusters A1 and A2, and cluster A is divided into two finer clusters A1 and A2, and each search term is classified into four clusters A1, A2, B, and C in total. It is shown. Indicates clusters (at a certain particle size) in which cell display modes (eg, cell color; spots, diagonal lines, and wavy lines are pseudo-colored in the figure) are classified, and the cell size is predetermined. It may indicate the number of searches in the period.

According to the treemap format and the sunburst format, it is possible to intuitively grasp the classification result and the number of searches. Even in these formats, it is desirable that the user can set the particle size variably.

<Modification example>
Although the first to eleventh embodiments of the present invention have been described above, the following description may be added to the embodiments.

(1) In the analysis result output processing of the first to third embodiments, the upper page classification is output as the analysis result. However, one or a combination of one or more of the following four types of information may be output as the analysis result.

_{First, after classifying the document data Dk} (k = 1 to d) into a plurality of subsets by classification processing such as clustering, classification, and community detection, the search for the evaluation target is performed based on the plurality of subsets. The needs purity may be obtained and the needs purity may be output as an analysis result. Here, the needs purity is an index indicating whether the variation in the properties of the needs purity in the search results is small or large. If the search result of a certain search term is occupied by web pages having similar properties, the needs purity of the search term is high. If the search term of a certain search term is occupied by web pages having different properties, the needs purity of the search term is low. The procedure for calculating the needs purity when the classification process is clustering class classification and when the classification process is community detection is as follows.

a1. When the classification process is clustering class classification In this case, _{the variance of the document data D k} (k = 1 to d) is calculated, and the needs purity is calculated based on this variance. More specifically, the feature vector data z ₁ = {z ₁₁ , z ₁₂ ... z _1l }, z ₂ = {z ₂₁ , z ₂₂ ... Of the document data D ₁ , D ₂ ... D _{d ...} z _2l } ... z _d = {z _d1 , z _d2 ... z _dl }, and find the average of all coordinates. Next, the feature vector data _{z 1 = {z 11, z} 12 ··· z 1l} of the document data _{D 1} distance from all coordinate average of the feature vector data _z of the document data _{D 2 2 = {z 21,} z ₂₂ ... _{Distance from all coordinate averages of z 2l } ... Features of document data D d} Find the distance from all coordinate averages of vector data z _d = {z _d1 , z _d2 ... z _dl }. Next, the variance of the distance from the average of all coordinates of the document data D ₁ , D ₂ ... D _d is obtained, and this variance is defined as the required purity. The required purity may be calculated based on the intra-cluster variance / intra-class variance instead of the variance of the distance from the average of all coordinates of the document data D ₁ , D ₂ ... D _d.

b1. When the classification process is community detection In this case, _{the average path length between the nodes of the document data Dk} in the undirected graph is calculated, and the needs purity is calculated based on this average path length. More specifically, _{a threshold of similarity between document data Dk} is set, and an unweighted undirected graph is generated by removing edges below the threshold. Next, the average path length between the nodes in this unweighted undirected graph is calculated, and the reciprocal of the average path length is used as the required purity. Similarly, the cluster coefficient, similarity selectivity, centrality distribution, and edge intensity distribution are obtained, and the values obtained by applying the cluster coefficient, similarity selectivity, centrality distribution, and edge intensity distribution to a predetermined function are obtained. It may be the required purity.

According to this variant, for example, as shown in FIG. 23, a first search term (storage in the example of FIG. 23) and a second search term including the first search term (in the example of FIG. 23, in the example of FIG. 23). cube storage) is a candidate for SEO, and if there is a difference in the number of searches per month for two search terms, the number of searches and needs purity of the first search term and the second search term By comparing the number of searches and the purity of needs, it becomes easy to determine which search term SEO is prioritized.

Second, as shown in FIG. 24, a first search term (storage in the example of FIG. 24) and a plurality of second search terms including the first search term (storage in the example of FIG. 24). Near me, storage sheds, cube storage, storage bins, storage boxes, mini storage, storage solutions, san storage, data storage) are evaluated, and the number of searches per month and document data D for each of multiple search terms. _{A list summarizing each product with the ratio of each subset to the entire k} (k = 1 to d) may be output as an analysis result.

According to this variant, the first search term and a plurality of second search terms including the first search term are candidates for SEO, and there is a difference in the number of searches of the plurality of search terms per month. If there is, it becomes easy to determine which search term SEO is prioritized. This variant is suitable for evaluating search terms with low needs purity.

Further, this second modification may be applied to the search-linked advertisement. When the second modification is applied to the search-linked advertisement, it is possible to improve the accuracy of the advertisement related to the search term when a plurality of search needs exist in one search term. For example, when performing search-linked advertisements related to "storage" shown in the example of FIG. 24, what percentage of facility-type advertisements should be displayed, what percentage of furniture-type advertisements should be displayed, and what percentage of computer-type advertisements should be displayed. You will be able to judge whether it should be displayed in a split manner.

Thirdly, the B degree, which is an index showing how much the top web page of the search term to be evaluated meets the business needs, and the index showing how much the top web page of the search term to be evaluated meets the consumer needs. The C degree is obtained, and the B degree and the C degree may be output as the analysis result. The procedure for calculating the B degree and the C degree when the classification process is class classification is as follows.

First, the feature vector data group associated with the label information indicating that it is BtoB teacher data, the feature vector data group associated with the label information indicating that it is BtoC teacher data, and the CtoC teacher data. A group of feature vector data associated with label information indicating that there is is prepared, and the weighting coefficient of the linear classifier f (z) is made suitable for classification of BtoB, BtoC, and CtoC by machine learning using these. Set.

After optimizing the weighting coefficient by machine learning, the feature vector data z ₁ = {z ₁₁ , z ₁₂ ... z _{1l' } of the document data D 1} is substituted into the linear classifier f (z), and the document data D _{Determine which class 1} belongs to, and substitute the feature vector data z ₂ = {z ₂₁ , z ₂₂ ... z _2l' } of the document data D ₂ into the linear classifier f (z) to create the document data. Determine which class D ₂ belongs to ... Substitute the feature vector data z _d = {z _d1 , z _d2 ... z _{dl' } of the document data D n} into the linear classifier f (z). _{The document data D 1} , D ₂ ... D _d is classified into a BtoB class, a BtoC class, and a CtoC class by determining which class the document data D _{n belongs to.} .. Then, the B degree and the C degree are calculated based on the relationship of the ratio of each class of BtoB, BtoC, and CtoC to the entire document data D _{k (k = 1 to d).}

By the same procedure, the academic degree, which is an index showing how much the top web page of the search term to be evaluated meets the academic needs, and how much the top web page of the search term to be evaluated meets the conversational needs. The degree of conversation indicating the above may be obtained, and these indexes may be output as the analysis result.

(2) In the first to ninth embodiments, the web page in the search result is analyzed. However, a web site or web content may be included in the analysis target.

(3) In the quantification process of the first to ninth embodiments, _{only the content of the document data D k} (k = 1 to d) is quantified, and the feature vector data obtained by quantifying the content is classified. May be good. Further, in the quantification process, _{only the structure of the document data D k} (k = 1 to d) may be quantified, and the feature vector data obtained by quantifying the contents may be classified.

(4) In the document content quantification processing of the first to ninth embodiments, the document data _Dk (k = 1 to d) is summarized by an automatic sentence summarization algorithm, and the summarized document data is summarized as a multidimensional vector. The multidimensional vectorized feature vector data may be processed in whole or in part after step S210.

(5) In the document structure quantification processing of the first to ninth embodiments, the structure of the document data D _k (k = 1 to d) is divided into the part lyrics composition ratio, the HTML tag structure, the dependency structure, and the structural complexity. Quantification based on (Structure Complexity) may be performed.

(6) In the evaluation axis setting process of the first and third embodiments, the number of classifications (number of clusters and communities) is set by moving the evaluation axis setting bar 9 to the upper layer side or the lower layer side. On the other hand, as shown in FIG. 4 (B), a part (in the example of FIG. 4 (B), the portion indicated by the chain line) is excluded from the classification target among the plurality of subsets of the same hierarchy. The number of classifications may be set.

(7) In the clustering processing of the first, fourth, and fifth embodiments, the feature vector data z ₁ = {z ₁₁ , z ₁₂ ... z _{1l of the document data D 1} , D ₂ ... D _{d. '} }, Z ₂ = {z ₂₁ , z ₂₂ ... z _2l' } ... z _d = {z _d1 , z _d2 ... z _dl' } were subjected to the processing of the shortest distance method for clustering. However, processing other than the shortest distance method may be performed. For example, the feature vector data z ₁ = {z ₁₁ , z ₁₂ ... z _1l' }, z ₂ = {z ₂₁ , z ₂₂ ... z _2l'of the document data D ₁ , D ₂ ... D _d. } in _{··· z d = {z d1,} z d2 ··· z dl '}, Ward's method (Ward method), group average method, nearest neighbor method, the maximum distance method, or algorithm of Fuzzy C-meaps method The processing may be performed according to the above.

Further, the feature vector data z ₁ = {z ₁₁ , z ₁₂ ... z _1l' } of the document data D ₁ , D ₂ ... D _{d , z 2} = {z ₂₁ , z ₂₂ ... z _2l' } ... Z _d = {z _d1 , z _d2 ... z _dl' } may be subjected to a clustering process using deep learning.

Further, the feature vector data z ₁ = {z ₁₁ , z ₁₂ ... z _1l' } of the document data D ₁ , D ₂ ... D _{d , z 2} = {z ₂₁ , z ₂₂ ... z _2l' } ... Z _d = {z _d1 , z _d2 ... z _dl' } may be processed according to a non-hierarchical cluster classification algorithm such as k-means. Here, since k-means is a non-hierarchical cluster classification, the dendrogram 8 cannot be presented as an analysis result. In the case of clustering k-means, in the evaluation axis setting process, it is preferable to accept the input of the value k of the number of clusters from the user and perform the clustering process with the specified number of clusters as a new setting.

(8) In the classification processing of the second, sixth, and seventh embodiments, the CPU 22 uses the so-called perceptron linear classifier f (z) to _{perform each of the document data D k} (k = 1 to d). I decided which class to assign. However, the classes may be sorted by another method. For example, by perceptron, naive bays method, template matching, k-nearest neighbor identification method, decision tree, random forest, AdaBoost, Support Vector Machine (SVM), or deep learning, document data D _k (k = 1 to d). May be classified into multiple classes. Further, the classification may be performed by a non-linear classifier instead of the linear classifier.

(9) In the community detection processing of the third, eighth, and ninth embodiments, the document data D _k (k = 1 to d) is made into a weighted undirected graph, and mediation of each node in the weighted undirected graph. By repeating the calculation of the centrality and the removal of the edge having the maximum mediation centrality, the document data _Dk (k = 1 to d) was classified into a plurality of communities. _{However, the document data D k} (k = 1 to d) may be classified into a plurality of communities by a method other than that based on mediation centrality. For example, a document by community detection based on random walk, greedy method, community detection based on eigenvector, community detection based on multi-step optimization, community detection based on spin glass method, Infomap method, or community detection based on Overlapping Community Detection. Data D _k (k = 1 to d) may be classified into a plurality of communities.

(10) In the community detection process of the fifth to sixth embodiments _{, an unweighted undirected graph having each of the document data D k} (k = 1 to d) as a node is generated, and the unweighted undirected graph is used. Based on this, the document data D _k (k = 1 to d) may be classified into a plurality of communities.

(11) In the analysis result output processing of the fourth and fifth embodiments, the upper page classification based on the processing result of the clustering processing and the mapping image 7 may be output as the analysis result screen. Further, in the analysis result output processing of the sixth and seventh embodiments, the upper page classification based on the processing result of the classification processing and the mapping image 7 may be output as the analysis result screen. Further, in the analysis result output processing of the eighth and ninth embodiments, the upper page classification based on the processing result of the community detection processing and the mapping image 7 may be output as the analysis result screen.

(12) In the first, second, fourth, fifth, sixth, and seventh embodiments described above, classification processing such as clustering or classification is performed on the processing result of the addition processing without executing the dimension reduction processing. May be given. Further, in the third, eighth, and ninth embodiments, the dimension reduction process is executed, and the feature vector data that has undergone the dimension reduction by the dimension reduction process is subjected to the similarity identification process and the community detection process to perform the dimension reduction process. A plurality of document data may be classified into a plurality of subsets based on the feature vector data that has undergone about processing.

1 ... Evaluation system, 10 ... User terminal, 20 ... Search needs evaluation device, 21 ... Communication interface, 22 ... CPU, 23 ... RAM, 24 ... ROM, 25 ... Hard disk, 26 ... Evaluation program, 50 ... Search engine server device ..

Claims (10)

A means for generating a multidimensional feature amount of each of the plurality of search terms based on a high-order predetermined number of search results obtained from a search engine, which is a search result for each of the plurality of search terms.
A means for identifying the similarity of the multidimensional features between one search term and a plurality of other search terms, and
A means for obtaining the degree of how close each search term is to which subset by classifying each of the plurality of search terms into one or more of a plurality of subsets based on the identified similarity.
A second display control means for displaying the degree to which the search term is close to which subset in the figure corresponding to the search term is provided.
Figure is divided into a plurality of portions, the corresponding each of which in the subset, the evaluation device. The evaluation device according to claim 1, further comprising a first display control means for displaying each of the plurality of search terms and the similarity of the multidimensional feature amount with the other search terms in a tabular form. The first display control means
The plurality of search terms are arranged in a predetermined direction and in a direction orthogonal to the predetermined direction.
The evaluation device according to claim 2, wherein the similarity of the multidimensional features between the corresponding search terms is displayed at the intersection in each direction. The first display control means
The plurality of search terms are arranged in a predetermined direction and in a direction orthogonal to the predetermined direction.
The evaluation device according to claim 2, wherein the intersections in each direction are displayed in an manner corresponding to the degree of similarity between the corresponding search terms. The evaluation according to claim 3 or 4, wherein the first display control means arranges the plurality of search terms in the predetermined direction in an order according to the similarity of the multidimensional feature amount with other search terms. Device. When one of the plurality of search terms is selected by the user, the first display control means performs the plurality of search terms in descending order of similarity of the multidimensional feature amount with the selected search term. The evaluation device according to claim 5, wherein the device is arranged in the predetermined direction. The evaluation device according to any one of claims 1 to 6, wherein the plurality of search terms include common words. The evaluation device according to any one of claims 1 to 7, wherein the subset corresponds to a search need. Computer,
A means for generating a multidimensional feature amount of each of the plurality of search terms based on a high-order predetermined number of search results obtained from a search engine, which is a search result for each of the plurality of search terms.
A means for identifying the similarity of the multidimensional features between one search term and a plurality of other search terms, and
A means for obtaining the degree of how close each search term is to which subset by classifying each of the plurality of search terms into one or more of a plurality of subsets based on the identified similarity.
In the figure corresponding to the search term, it functions as a display control means for displaying the degree of how close the search term is to which subset.
An evaluation program in which a figure is divided into a plurality of parts, each of which corresponds to the subset. The multidimensional feature amount generating means is a search result for each of a plurality of search terms, and is a multidimensional feature amount of each of the plurality of search terms based on a high-order predetermined number of the search results acquired from the search engine. And the steps to generate
The similarity identification means includes a step of identifying the similarity of the multidimensional feature amount between one search term and a plurality of other search terms.
The classification means obtains the degree of how close each search term is to which subset by classifying each of the plurality of search terms into one or more of a plurality of subsets based on the identified similarity. the method comprising the steps of,
Display control means, in figure corresponding to the search word, look including the steps of: displaying a degree of how close to the search word which subsets,
An evaluation method in which a figure is divided into a plurality of parts, each of which corresponds to the subset.

Images