JP5765691B1 - Patent literature analysis method - Google Patents

Abstract

Provided is a method for stratifying a relationship between a specific technology and an element technical item of the specific technology and an assumed item in a time-series patent map. The number of patent documents consisting of a reference patent classification using one axis as a reference for calculating similarity of time series transitions between patent classifications, and an analysis target patent classification for calculating a correlation coefficient in combination with the reference patent classification A time-series patent map showing the number of patent documents for each patent classification for each time-series section, with the other axis as a time axis having a time-series section having at least three or more analysis periods Convert to The correlation coefficient value in the time axis direction of all patent classifications in the table, which varies depending on the numerical value of the similarity between the base patent classification and the analysis patent classification and the number of samples used to calculate the correlation coefficient. By comparing with the correlation determination reference value, the relationship between the reference patent classification and the analysis target patent classification is stratified according to whether the similarity is positive / negative / none. [Selection] Figure 2

Description

The present invention relates to a method for analyzing a patent document for obtaining a direction for developing a new technology.

As a method of grasping elemental technical items such as elemental technology, development purpose, device method, device form, etc. of a specific technology currently in progress, a time-series patent map or matrix of a collection of patent documents related to the specific technology to be analyzed There is a method of creating a patent map and observing technical items assumed as elemental technical items of a specific technology from the patent map.

Specifically, a patent document set related to a specific technology is created, a time series patent map is created regarding the technical items that are assumed to be elemental technical items of the specific technology, and patents of patent classifications granted to the specific technology Observe the time series transition of the number of documents and the number of patent documents of the patent classification assigned to the assumed technical item as the elemental technical item of the specific technology, similar to the trend of the patent classification assigned to the specific technology Or the patent classification assigned to the specific technology and the technical classification assumed to be the element technical item of the specific technology. A matrix patent map in which the patent classification of one axis, the other axis is the applicant to be analyzed, and the intersection of both axes describes the patent classification of each axis and the number of patent documents to which the applicant is assigned at the same time The The ratio of the number of patent documents of the patent classification granted to a specific technology and the number of patent classifications assigned to an assumed technical item is almost constant regardless of the applicant. This is a method for estimating that a combination of patent classifications possessed is a combination of a specific technology and elemental technology items.

Patent No. 5527566

In the case of a time-series patent map, the time series of the number of patent documents of a patent classification assigned to a specific technology and the number of patent documents of a patent classification assigned to an assumed technical item as an elemental technical item of the specific technology The transition is similar in a certain period and may show completely different trends in a certain period, so it is difficult to accurately estimate the elemental technical items while observing them. A method for quantitatively evaluating the similarity of a series transition using a correlation coefficient is described in Patent Document 1, but the dummy used as a reference for the time series trend in Patent Document 1 is a monotonically increasing trend or a monotonically decreasing trend. In this method, it is difficult to classify patent classifications assumed as elemental technical items from the viewpoint of whether or not they are elemental technical items.

In the case of a matrix patent map, the ratio of the number of patent documents in the patent classification assigned to the assumed technical item and the patent classification assigned to the specific technology and the element technical item for each applicant of the matrix patent map. However, it is difficult to analyze a matrix patent map that does not have a scale for identifying an applicant to be analyzed.

Therefore, the object of the present invention is given to the technical items to be assumed as the number of patent documents of the patent classification assigned to the specific technology and the elemental technical items of the specific technology in the analysis target period of the time series patent map. Similarity between the number of patent documents of the specified technology and the number of patent documents of the specified technology and the number of patent documents to be assumed as the element technology item of the specified technology by the correlation coefficient of the scatter diagram of the time series The specific technique and the elemental technical items of the specific technique are assumed by comparing the quantified degree of similarity and the correlation discrimination reference value that changes depending on the number of samples used for calculating the correlation coefficient. It is to provide a method for stratifying the relationship of the selected items.

In addition, in the matrix patent map in which one axis is the patent classification axis and the other axis is the applicant axis, the object of the present invention is the number of patent documents of the patent classification assigned to a specific technique and the specific technique. The number of patent documents of the specific technology and the specific number of correlations in the scatter diagram for seeing the correlation of the number of patent documents of the patent classification assigned to the assumed technical items By quantifying the similarity of the number of patent documents of the assumed technical item technical item and the item, by comparing the quantified similarity and a correlation criterion value that changes depending on the number of samples used to calculate the correlation coefficient, It is to provide a method for stratifying the relationship between the specific technology and the elemental technical items of the specific technology and assumed items.

In order to solve the above-described problems and achieve the object, the present invention provides the following patent document analysis method.

The invention according to claim 1 is a patent document analysis method for processing a patent document set accumulated in a storage unit of a computer in which patent document analysis software is installed, by the patent document analysis software,
The number of patent documents of the analysis target patent classification that calculates the correlation coefficient in combination with the reference patent classification and the reference patent classification in which one axis is used as a reference for calculating the similarity of the number of patent documents of patent classification The number axis of patent documents that display
The other axis is a time series axis in which the analysis period is divided into time series sections having at least three or more data,
Creating a time series patent map indicating the number of patent documents for each patent classification for each time series section, and converting the time series patent map into a time series numerical table;
Create a correlation matrix in the time axis direction between all patent classifications of the time series numerical table, and the correlation coefficient values of the reference patent classification and the analysis target patent classification of the correlation matrix in the time series patent map The degree of similarity of the time series transition of the number of patent documents of the reference patent classification and the analysis target patent classification is quantified, and the correlation determination changes depending on the numerical value of the similarity and the number of samples used to create the correlation matrix. Stratifying the relationship between the reference patent classification and the analysis-target patent classification into a positive similarity, no similarity, and a negative similarity by comparison with a reference value;
The patent document analysis method characterized by comprising.

A patent document analysis method for processing a patent document set accumulated in a storage unit of a computer in which patent document analysis software is installed by the patent document analysis software,
Selection of the target patent category, which uses one axis as a standard for calculating the similarity of the number of patent documents between patent categories, the target patent category for calculating the correlation coefficient in combination with the reference patent category, and the target applicant for analysis A patent classification axis consisting of a specific patent classification and
The other axis is an applicant axis in which at least three specific applicants selected from among applicants having the number of patent documents to which the specific patent classification is assigned are arranged,
Create a matrix patent map in which the number of patent documents to which the patent classification and the applicant are simultaneously assigned is written at the intersection of the patent classification axis and the applicant axis, and the matrix patent map is Converting to a matrix numeric table;
Create a correlation matrix in the applicant axis direction of all patent classifications of the patent classification axis of the matrix numerical table, and use the correlation coefficient values of the reference patent classification and the analysis target patent classification of the correlation matrix to calculate the matrix Quantify the similarity between the number of patent documents of the reference patent classification and the analysis patent classification in the applicant axis direction in the patent map,
By comparing the digitized value of the similarity with a correlation criterion value that varies depending on the number of samples used to create the correlation matrix, the relationship between the reference patent classification and the analysis patent classification is positively similar. Stratifying relevant, non-similar, negative-similar,
The patent document analysis method characterized by comprising.

According to the analysis of the patent document, the present invention has the following effects.

In the invention described in claim 1, if the analysis is performed by applying the patent classification of the development target technology to the basic patent classification and applying the analysis to the analysis target patent classification by applying the patent classification assumed to be the element technical item of the development target technology, The target element technology item can be stratified into a stability element technology item, a decline element technology item, and a variability element technology item.

In the invention described in claim 1, the analysis period is divided into a plurality of analysis periods, and a quantitative value of the similarity between the reference patent classification and the analysis target patent classification is calculated for each analysis period. By connecting quantitative values in chronological order, it is possible to know the transition of the similarity between the standard patent classification and the analysis-target patent classification, in other words, the patent classification of the technology that is being developed to the standard patent classification. By applying a patent classification that is assumed to be an elemental technical item constituting the technology to be developed to the analysis-target patent classification, it is possible to know the time series transition of the elemental technical item of the technology to be developed. it can.

In the invention according to claim 2, the patent classification given to the technology targeted for development is applied to the standard patent classification, the patent classification assumed as the elemental technical item of the technology targeted for development is applied to the analysis target patent classification, If the patent classification of the company's specialty technology field is applied to the selector patent classification and analyzed, the element technology items constituting the technology as the development target adopted by the applicant who has the same specialty field as the company are adopted. Can be sieved.

Further, in the invention described in claim 2, if the base patent classification and the analysis target patent classification are the same and only the selector patent classification is analyzed, the similarity between the base patent classification and the analysis target patent classification in different technical fields Because it can know the difference, it can also lead to a change in idea.

Configuration block diagram of patent document analyzer Flowchart of analysis of time-series patent map showing the first embodiment of the invention Time-series patent map with analysis period of 1996-2000 Time series numerical table for analysis period 1996-2000 Correlation matrix obtained by correlation analysis of time series numerical tables with analysis period of 1996-2000 Similarity list of analysis target patent classifications of the analysis target patent classification 3C707AS01 of 1996-2000 Descending order similarity list of analysis target patent classification 3C707AS01 and analysis result patent classification stratified result list of analysis patent period of 1996-2000 Descending order similarity list with reference patent classification 3C707AS01 of analysis target patent classification of analysis period 2000-2004 and stratified result list of analysis target patent classification Descending order similarity list of the analysis target patent classification of the analysis target patent classification 3C707AS01 and analysis result patent classification stratified result list Descending order similarity list of analysis target patent classification 3C707AS01 of analysis target patent classification and analysis result patent classification stratification result list of analysis period 2008-2012 Table of changes in similarity between the analysis target patent classification and the reference patent classification due to changes in the analysis period Transition graph of similarity between the analysis target patent classification and the reference patent classification (Part 1) Transition graph of similarity between the analysis target patent classification and the standard patent classification (Part 2) Transition graph of similarity between the analysis target patent classification and the reference patent classification (Part 3) Transition graph of similarity between the analysis target patent classification and the reference patent classification (Part 4) Flowchart of matrix patent map analysis of the second embodiment of the present invention First matrix patent map First matrix numeric table Second matrix numerical table (selector patent classification 3C707BS10) Third matrix numerical table (selector patent classification 3C707BS10) Correlation matrix with 3C707AS01 as the reference patent classification and 3C707MT as the analysis target patent classification (selector patent classification 3C707BS10) 3C707MT F-term similarity list with 3C707AS01 (selector patent classification 3C707BS10) 3C707MT F-term similarity list with 3C707AS01 (selector patent classification 3C707BS26) Descending similarity list with reference patent classification 3C707AS01 of analysis target patent classification and stratified result list of analysis target patent classification (selector patent classification 3C707BS10) Descending similarity list with reference patent classification 3C707AS01 of analysis target patent classification and stratified result list of analysis target patent classification (selector patent classification 3C707BS26) Comparison list of similarity of the analysis target patent classification to the standard patent classification due to the difference of selector patent classification Comparison graph of similarity between the analysis target patent classification and the reference patent classification for different selector patent classifications

Hereinafter, an embodiment of the analysis method of the patent document of the present invention will be described. The embodiment of the present invention shows the most preferable mode of the present invention, and the present invention is not limited to this.

FIG. 1 shows a patent document analysis apparatus 1 having a computer 2 in which patent document analysis software used for the analysis method of all patent documents to which patent classification is assigned, such as published patent gazettes and patent gazettes, is installed. The configuration will be described. This patent document analysis apparatus 1 has a computer 2 to which a control device 3, an arithmetic device 4, a storage device 5 and an I / O 6 are connected via a bus line, and on the input side of the I / O 6 of the computer 2 A keyboard 7, a mouse 8, an input device 9, a CD-ROM drive 10 including an MO drive, etc. are connected, and on the output side of the I / O 6, a display device 11 comprising a CRT (including a liquid crystal display), a printer 12 etc. are connected. Further, it can be connected to an ASP patent database 13 such as PATOLIS through a communication line 14. The patent database 13 stores patent documents such as bibliographic information and specification contents.

The main functions of the patent document analyzer 1 are as follows.
A time-series patent map and matrix drawn on the display device 11 by operating the input device 9 to open the analysis-target patent document set file using the patent document analysis software preinstalled in the storage device 5 of the computer 2 A function to create a patent map by inputting conditions suitable for the purpose of analysis from the creation condition setting screen of various patent maps such as a patent map, a function to switch from a patent map to a numerical table, and from a numerical table to a patent map, a numerical table The function of deleting any row or column of the above, the rearrangement of the column of the numerical table is to arrange the number of patent documents in a specific row in ascending order or descending order, the rearrangement of the row is the number of patent documents in a specific column Functions that can be implemented by arranging them in ascending order or descending order, from the patent literature set installed in the storage device 5, the input device The search expression prepared by the operation, and has a function for creating a new patent document set.

In the correlation analysis, an analysis condition setting screen of correlation analysis software installed in the storage device 5 in advance is drawn on the display device 11 by operating the input device 9, and the range of analysis is displayed on the numerical table of the patent map to be analyzed. After specifying whether the analysis direction is the row direction or the column direction, specifying whether or not to use the first column or the first row as a label, a correlation matrix can be created by clicking the OK button on the analysis condition setting screen. .

A method for creating a collection of patent documents necessary for analysis is as follows. First, the input device 9 is operated to access the patent database 13 or the published patent publication CD set in the drive 10 such as a CD-ROM, A patent document set necessary for analysis is created using a search expression composed of patent classifications and keywords, a name is given to the patent document set required for analysis, and it is stored in the storage device 5 and stored. A set of patent documents necessary for analysis is converted into a format that can be processed by the patent document analysis software using a data format conversion program installed in the storage device 5 in advance, and the converted patent document set is screened and analyzed. It is necessary to make it a set of conforming patent documents.
The patent classification in the present application is any of IPC, FI subclass, main group, subgroup, F-term theme code, theme code + viewpoint, and F-term.

If the collection conditions of the set of patent documents for analysis are too wide to be suitable for detailed analysis, use the search function attached to the patent document analysis software and use the conditions suitable for detailed analysis. If a set of patent document references is searched, a copy of a patent document set is created, and the copy of the copy of the patent document is stored as an object of the set of patent document sets for analysis, it can be used as needed.

Table 1 shows the technical items that are assumed to be the elemental technical items of the specific technology of the time-series patent map, whether they are the stationary element technical item, the decline element technical item, or the variability element technical item of the specific technology. In order to explain a method for stratifying, a patent classification given to the specific technology is referred to as a standard patent classification, and as a specific example of the standard patent classification, the standard patent classification A, the specific technology patent classification to be applied to the technical fields to assume that element technology items described analyzed patent classification, specific examples of the analyte patent classification, analyzed patent classification B1, analyzed patent classification B2, analyzed Patent classification It is a model figure of the time series patent map created using X1 and the analysis object patent classification | category X2.

In the time series patent map, when the horizontal axis is the time axis, the vertical axis is the patent document number axis, and when the horizontal axis is the patent document number axis, the vertical axis is the time axis, but in either case the result is In order to avoid confusion, the horizontal axis will be described as a time axis and the vertical axis will be described as a patent document number axis.

It is determined that the technical item to which the analysis target patent classification is assigned is the stability element technical item of the technology to which the reference patent classification is assigned because the reference patent classification and the analysis target patent classification are the same In the time-series patent map (see Table 1) of the set of patent documents for analysis, the patent documents of the reference patent classification and the analysis target patent classification in the period to be analyzed are included. This is a case where the time-series transitions that change with the number of cases are in a similar state (for example, the relationship between the standard patent classification A and the analysis target patent classification B1 in Table 1).

Quantitatively, the time series patent similarity of time-series transition of the patent document number of the reference patent classification and the analyzed patent classification of the analysis period in the map size beta _1, the beta ₁ is the analytical If it is defined that it is equal to the correlation coefficient magnitude R ₁ of the scatter diagram of the time series transition of the number of patent documents of the target patent classification and the analysis target patent classification in the target period, the β ₁ is 1 ≧ β ₁ ≧ α _{When 1} is a positive value, in other words, the analysis-target patent classification when the similarity of the time series transition of the number of patent documents of the reference patent classification and the analysis-target patent classification has a positive correlation is the standard It is defined as a stability element technical item of patent classification. However, α ₁ is a correlation determination reference value of the time-series patent map calculated by Equation ₁ .

但し、α_１は、相関係数をα_１とした場合、α_１ ^２＞４／（Ｎ＋２）（Ｎ：データ数）が成立すれば相関があると判定される値で、α_１を時系列特許マップの相関判別基準値と記載することとし、Ｎは基準特許分類と分析対象特許分類の相関係数を求める時のサンプル個数を意味している。
（なお、数１は「相関があるかを判定する式」として、「オペレーションズ・リサーチ」１９９７年７月号、上田太一郎 に掲載されているものである。）Here, when β ₁ = 1, the magnitude of the correlation coefficient of the scatter diagram of the time series transition of the number of patent documents of the reference patent classification and the analysis target patent classification is 1 (for example, in Table 1) Scatter chart of standard patent classification A and analysis target patent classification B1 (see Table 2), and β ₁ = α ₁ is a positive value, the number of patent documents of the reference patent classification and the analysis target patent classification When the correlation coefficient magnitude R ₁ of the scatter diagram of the time-series transition is equal to the positive value α ₁ calculated by Equation ₁ (the R ₁ is approaching the positive value of α ₁ ) Is a scatter diagram of the standard patent classification A and the analysis target patent classification X1 in Table 1, Table 3. The number of data in Table 3 is 10, the square value of the correlation coefficient of the scatter diagram is 0.334, The square value of the correlation determination reference value α _{1 in} the case of N = 10 in Equation ₁ is 0.333, and the data variation state in Table 3 Is close to a positive value of α ₁ ).

However, alpha _1, when the correlation coefficient was ^{_{α 1, α 1 2> 4}} / (N + 2): a value (N number of data) is determined to be correlated if satisfied, time the alpha ₁ sequence It is described as the correlation determination reference value of the patent map, and N means the number of samples when obtaining the correlation coefficient between the reference patent classification and the analysis target patent classification.
(Equation 1 is published in “Operations Research” July 1997, Taichiro Ueda, as an expression for determining whether there is a correlation.)

The number of patent documents to which the reference patent classification is assigned is determined that the technical item to which the analysis-target patent classification is assigned is a declining element technical item of the technique to which the reference patent classification is assigned. If the number of patent documents to which the analysis target patent classification is assigned decreases, the number of patent documents to which the reference patent classification is assigned decreases, and the number of patent documents to which the analysis target patent classification is assigned increases. In the time series patent map (see Table 1) of the relevant patent document set for analysis, the time series transition of the number of patent documents of the reference patent classification and the analysis target patent classification in the analysis target period is on the time axis. This is a case where there is a line-symmetric relationship (for example, the relationship between the standard patent classification A and the analysis target patent classification B2 in Table 1) with a specific parallel axis as the symmetry axis.

Quantitatively, when β ₁ is a negative value of −1 ≦ β ₁ ≦ α ₁ , in other words, the similarity of the time series transition of the number of patent documents of the reference patent classification and the analysis target patent classification is The analysis-target patent classification in the case of having a negative correlation is defined as a fading element technical item of the reference patent classification.

Here, when β ₁ = −1, the correlation coefficient of the scatter diagram of the time series transition of the number of patent documents of the reference patent classification and the analysis target patent classification is −1 (as an example 1 is a scatter diagram of the reference patent classification A and the analysis target patent classification B2, Table 4), and when β ₁ = a negative value of α ₁ , the patent documents of the reference patent classification and the analysis target patent classification size negative values of alpha ₁ wherein R ₁ is calculated by the number 1 of the correlation coefficient of the scatter plot of time-series transition of the number is equal to the case (the R ₁ approaches the negative value of the alpha ₁ Example 5 is a scatter diagram of the standard patent classification A and analysis target patent classification X2 in Table 1, Table 5. The number of data in Table 5 is 10, and the square value of the correlation coefficient of the scatter diagram is 0.334, On the other hand, the square value of the criterion value α _{1 in} the case of N = 10 in Equation ₁ is 0.333, and the data variation state in Table 5 Is close to the negative value of α ₁ ).

It is determined that the technical item to which the analysis target patent classification is assigned is a variability element technical item of the technology to which the reference patent classification is assigned because the reference patent classification and the analysis target patent classification are the same In the time-series patent map (see Table 1) of the set of relevant patent documents for analysis, the frequency of the reference patent classification and the analysis target patent classification in the analysis target period are randomly changed. When the time series transition of the number of patent documents is not similar, or the time series transition of the number of patent documents of the reference patent classification and the analysis target patent classification is a line with a specific axis parallel to the time axis as the axis of symmetry This is the case where there is no symmetric relationship.

Quantitatively, when β is in a range where α ₁ is a positive value>β> α ₁ is a negative value, in other words, when the number of patent documents of the reference patent classification and the analysis target patent classification is The analysis-target patent classification in the case where the degree of similarity of the series transition has a non-correlated relationship is defined as the variability element technical item of the reference patent classification.

Table 6 shows the technical items that are assumed to be the elemental technical items of the specific technology of the matrix patent map consisting of 10 applicants to be analyzed as the steady-state element technical items of the specific technology. In order to explain a method for stratifying whether the item is a variable element technical item, a patent classification given to the specific technology is described as a standard patent classification, and a standard patent classification is given as a specific example of the standard patent classification. C, a patent classification assigned to an element technical item of the specific technology and an assumed technical item is described as an analysis target patent classification, and as an example of the analysis target patent classification, an analysis target patent classification D1, an analysis target patent classification It is a model figure of the matrix patent map created using D2, analysis object patent classification Y1, and analysis object patent classification Y2.

Table 7 is a model diagram of a matrix patent map in which the number of patent documents of the basic patent classification C in Table 6 is arranged in descending order from the left to the right.

If the vertical axis is the patent classification, the horizontal axis is the applicant, and the vertical axis is the applicant, the horizontal axis is the patent classification, but in both cases the result is the same, To avoid confusion, the vertical axis is the patent classification and the horizontal axis is the applicant.

It is determined that the technical item to which the analysis target patent category is assigned is the stability element technical item of the technology to which the reference patent category is assigned. The number of patent documents to which the reference patent classification is assigned and the analysis-target patent classification are assigned in the matrix patent map (see Table 6) of the patent document collection of the applicant to be analyzed when the frequency is given to the documents. This is a case where the ratio of the number of patent documents being registered is constant regardless of the applicant (for example, the relationship between the reference patent classification C and the analysis target patent classification D1 in Table 6).

Quantitatively, the reference patent classification and the analysis of the similarity of the patent document number of applicants axis direction of the subject patent classification size beta ₂ in the matrix patent _map, the beta ₂ is the analyte and the reference Patent Classification If it is defined that it is equal to the magnitude R ₂ of the correlation coefficient of the scatter diagram of the number of patent documents in the applicant's axial direction in the patent classification, if β ₂ is 1 ≧ β ₂ ≧ α ₂ is a positive value, If the similarity of the number of patent documents in the applicant axis direction of the reference patent classification and the analysis target patent classification has a positive correlation, the analysis target patent classification is a stability element technical item of the reference patent classification. Define that there is. However alpha ₂ is the correlation determination reference value matrix patent map calculated by the number 2.

但し、α_２は、相関係数をα_２とした場合、α_２ ^２＞４／（Ｎ＋２）（Ｎ：データ数）が成立すれば相関があると判定される値で、α_２をマトリクス特許マップの相関判別基準値と記載することとし、Ｎは基準特許分類と分析対象特許分類の相関係数を求める時のサンプル個数を意味している。
（なお、数２は「相関があるかを判定する式」として、「オペレーションズ・リサーチ」１９９７年７月号、上田太一郎 に掲載されているものである。）Here, when β ₂ = 1, the correlation coefficient of the scatter diagram of the number of patent documents in the applicant's axial direction of the reference patent classification and the analysis target patent classification is 1 (for example, Table 6). Scatter chart of the reference patent classification C and the analysis target patent classification D1, see Table 8), and β ₂ = α ₂ is a positive value, the applicant axis of the reference patent classification and the analysis patent classification When the magnitude R ₂ of the correlation coefficient of the scatter diagram of the number of patent documents in the direction is equal to α ₂ calculated by Expression ₂ (the R ₂ approaches a positive value of α ₂ ) Examples are the scatter diagram of the reference patent classification C and the analysis target patent classification Y1 in Table 6, Table 9. The number of data in Table 9 is 10, the square value of the correlation coefficient of the scatter diagram is 0.3342, On the other hand, the square value of the correlation determination reference value α _{2 in} the case of N = 10 in Equation 2 is 0.333, and the variation of the data in Table 9 is as follows. The key state is close to a positive value of α ₂ ).

However, alpha _2, when the correlation coefficient and ^{_{α 2, α 2 2> 4}} / (N + 2): a value (N number of data) is determined to be correlated if satisfied, the alpha ₂ matrix patent It is described as a correlation discrimination reference value of the map, and N means the number of samples when obtaining the correlation coefficient between the reference patent classification and the analysis target patent classification.
(Note that Equation 2 is published in “Operations Research” July 1997, Taichiro Ueda, as an expression for determining whether there is a correlation.)

It is determined that the technical item to which the analysis target patent classification is assigned is a declining element technical item of the technology to which the reference patent classification is assigned because the number of patent documents to which the reference patent classification is assigned is determined. As the number of patent documents to which the analysis-target patent classification is assigned decreases, the number of patent documents to which the reference patent classification is assigned decreases, and the number of patent documents to which the analysis-target patent classification is assigned increases. In the case of a matrix patent map, the patent map 1 (for example, the matrix patent map of Table 6) in which the number of patent documents in the reference patent classification column is arranged in ascending order from left to right and the patent documents in the reference patent classification column. For example, the first column of the patent map 1 and the first column of the patent map 2 of the patent map 2 (for example, the matrix patent map of Table 7) in which the numbers are arranged in descending order from left to right. The number of patent documents to which the reference patent classification of the patent map 1 in the same column is assigned (for example, the number of patent documents of the basic patent classification C in Table 6) and the analysis target patent classification of the patent map 2 are This is a case where the ratio of the number of patent documents to be granted (for example, the number of patent documents of the analysis target patent classification D2 in Table 7) becomes constant. In other words, the reference patent classification is assigned in the patent map 1 This is a case where the number of patent documents and the number of patent documents to which the analysis object patent classification is assigned are in a plane-symmetric relationship.

Quantitatively, when β ₂ is negative, −1 ≦ β ₂ ≦ α ₂ , in other words, the similarity between the number of patent documents in the applicant's axial direction of the reference patent classification and the analysis target patent classification The analysis-target patent classification in the case where has a negative correlation is defined as a decaying element technical item of the reference patent classification.

Here, when β ₂ = −1, the correlation coefficient of the scatter diagram of the number of patent documents in the applicant's axis direction in the matrix patent map of the reference patent classification and the analysis-target patent classification is −1. In the case (for example, a scatter diagram of the standard patent classification C and the analysis target patent classification D2 in Table 6, Table 10), and when β ₂ = negative value of α ₂ , the reference patent classification and the analysis target When the correlation coefficient magnitude R ₂ of the scatter diagram of the number of patent documents in the applicant's axis direction in the matrix patent map of patent classification is equal to the negative value of α ₂ calculated by Equation ₂ (the R ₂ is An example of approaching the negative value of α ₂ is a scatter diagram of the standard patent classification C and the analysis target patent classification Y2 in Table 6, Table 11, where the number of data in Table 11 is 10, The square value of the number of relations is 0.3344, while the number 2 in the number 2 is 10 Since the square value of the correlation determination reference value α ₂ of the matrix patent map is 0.333, the variation state of the data in Table 11 is close to the negative value of α ₂ ).

It is determined that the technical item to which the analysis target patent classification is assigned is a variability element technical item of the technology to which the reference patent classification is assigned because the reference patent classification and the analysis target patent classification are the same The number of patent documents to which the reference patent classification is assigned and the analysis in the matrix patent map (see Table 6) of the set of patent documents of the applicant to be analyzed When the ratio of the number of patent documents to which the target patent classification is assigned varies depending on the applicant, or the number of patent documents to which the reference patent classification is assigned and the number of patent documents to which the analysis target patent classification is assigned Is not in a plane symmetry relationship.

Quantitatively, when β ₂ is in a range where α ₂ is a positive value> β ₂ > α ₂ is a negative value, in other words, the applicant axis of the reference patent classification and the analysis patent classification The analysis target patent classification when the similarity of the number of patent documents in the direction is uncorrelated is defined as the variability element technical item of the reference patent classification.

In the case of the time-series patent map, the relationship between the similarity β ₁ of the time-series transition of the number of patent documents in the time-series patent map and the analyzed patent classification in the time-series patent map and the correlation determination reference value α ₁ In this case, the analysis-target patent classification is determined based on the relation between the similarity β _{2 of the} number of patent documents in the applicant's axis direction of the matrix patent map and the patent classification of the analysis-target patent classification and the correlation determination reference value α ₂ . The method of stratifying whether it is a stability element technical item, a decline element technical item, or a variability element technical item is defined, but the correlation determination reference value α ₁ in the time-series patent map is Contact since calculation expression number 1 and the number 2 is a calculation formula of the correlation determination reference value alpha ₂ in the matrix patent map is are identical, in chronological Patent map and matrix patent map Or a stable element technology item the analyzed patent classification Te, or a decline element technology items, how to stratify on whether the variability component technology fields it is understood that handle the same.

As described above, the analysis in the time-series patent map and the matrix patent map is performed by rewriting the β ₁ in the time-series patent map as β, the α ₁ as α, the β ₂ in the matrix patent map as β, and the α ₂ as α. The categorization criteria for the target patent classification are summarized as follows.

In the time series patent map, β is the similarity of the time series transition of the number of patent documents in the analysis target period and the number of patent documents in the analysis target patent classification, and in the matrix patent map, the reference patent classification and analysis target of the applicant to be analyzed When the similarity of the number of patent documents in the applicant's axial direction of the patent classification is β, the patent classification to be analyzed is stratified into the following three types of elemental technical items by the value of β.
Here, α is a correlation determination reference value calculated by Equation 3.

但し、αは、相関係数をαとした場合、α^２＞４／（Ｎ＋２）（Ｎ：データ数）が成立すれば相関があると判定される値で、αを相関判別基準値と記載することとし、Ｎは基準特許分類と分析対象特許分類の相関係数を求める時のサンプル個数を意味している。
（なお、数２は「相関があるかを判定する式」として、「オペレーションズ・リサーチ」１９９７年７月号、上田太一郎 に掲載されているものである。）

However, when α is a correlation coefficient, α is a value determined to have a correlation if α ² > 4 / (N + 2) (N: number of data) is established, and α is described as a correlation determination reference value. N represents the number of samples when the correlation coefficient between the standard patent classification and the analysis target patent classification is obtained.
(Note that Equation 2 is published in “Operations Research” July 1997, Taichiro Ueda, as an expression for determining whether there is a correlation.)

When 1 ≧ β ≧ α is a positive value, the similarity between the reference patent classification and the analysis target patent classification is a positive similarity, and in the stratified determination, the analysis target patent classification is the stability of the reference patent classification. Judged as an elemental technical item.

When α is a negative value ≧ β ≧ −1, the similarity between the reference patent classification and the analysis target patent classification has a negative similarity relationship, and in the stratified determination, the analysis target patent classification is determined based on the reference patent classification. Judged to be a decline elemental technical item.

If α is a positive value> β> α is a negative value, there is no correlation between the similarity of the reference patent classification and the analysis-target patent classification. Judged as a variability element technical item.

(First embodiment)
The first embodiment relates to an analysis method based on a time-series patent map, and the analysis target patent document set uses the data of the patent document set A stored as an object of the file name and the manipulator patent document set. As will be described with reference to FIGS. 1 and 2 to 15, it is assumed that the patent document collection of the manipulator is already stored in the storage device 5.

The analysis purpose of the first embodiment is that the standard patent classification is F-term 3C707AS01 (manipulator use: transport, transfer, article supply, take-out), and the analysis target patent classification is F-term 3C707MT (manipulator control purpose). A specific period of the time-series patent map with F terms is divided into a plurality of analysis target periods, and the similarity between the time series transitions of the number of patent documents of 3C707MT and 3C707AS01 is calculated for each divided analysis target period. And stratifying the 3C707MT F-term into 3C707AS01 stability, decay, or variability component technical items using the similarity value; and The similarity value is connected to the time series of each F term of 3C707MT to clarify the time series transition of the similarity.

Specifically, the patent document set A consisting of the applicants with the top 20 documents in the patent code set of the theme code 3C707 (manipulator) whose application date is 1996-2012 is the analysis period of 1996-2000. , 2000-2004, 2004-2008, and 2008-2012, each period is divided into 1 year with 5 data, and 3C707AS01 and 3C707MT F-terms for each period The 3C707MT F-term is a stability element technical item of 3C707AS01 using the magnitude β of the similarity and the correlation determination reference value α calculated by the above equation 3 or It was categorized as an elemental technical item or a variable elemental technical item, and the similarity value was connected in time series by F-term of 3C707MT. A map is created to clarify the time series transition of the similarity.

In the first embodiment, how to count the number of patent documents of 1996 to 2012 to be analyzed and the number of patent documents of the analysis target patent classification is analyzed according to the number of patent documents belonging to 1996. There is a one-to-one correspondence between the patent application filing year and the subject year of analysis, such as the number of patent documents in 1996. To alleviate changes in the number of patent documents in the time-series patent map Furthermore, the number of patent documents analyzed in 1996 = (number of patent documents filed in 1995 and 1996) ÷ 2, the number of patent documents analyzed in 1996 = (patents whose application date was 1994-1996) A moving average such as a 2-year moving average or a 3-year moving average can also be taken as (number of documents) ÷ 3.

In the first embodiment, all the four analysis periods are equally divided so that each period has five data, but the time series of the number of patent documents of the reference patent classification and the analysis target patent classification. In order to improve the accuracy of the degree of similarity β of the transition, the time series transition of the number of patent documents of the standard patent classification is approximated by a plurality of straight lines, and the straight section is set as the analysis period. It is possible to analyze even if the analysis period includes three data, including 5 data in a certain analysis period, and different data numbers for each analysis period, but at least three or more data are included in each analysis period. It is necessary to include.

However, since the magnitude α of the correlation determination reference value varies depending on the value N according to the above formula 3, if it is desired to accurately compare the magnitude β of the similarity every analysis period, correction is necessary using the formula 4. However, if you want to roughly grasp the transition of β for each period, the correction is negligible.

ここで、
α_１は数３のＮがＮ_１の場合の相関判別基準値
α_２は数３のＮがＮ_２の場合の相関判別基準値
Ｎ_１、Ｎ_２は通常３〜５の値をとるため、Ｎ_１＝３、Ｎ_２＝５の場合でα_２＝０．８５α_１となり、α_１とα_２の違いは約１５％である。This is because the influence of N on α in Equation 3 can be expressed by Equation 4.

here,
Since alpha ₁ can take values of the correlation determination reference value _N 1, _{N 2} is usually 3 to 5 when the number 3 of N is correlated discrimination reference value alpha ₂ is the number 3 in the case of _{N 1} N is _{N 2,} In the case of N ₁ = 3 and N ₂ = 5, α ₂ = 0.85α ₁ , and the difference between α ₁ and α ₂ is about 15%.

Since the analysis methods of the 3C707AS01 and 3C707MT F terms in the four analysis periods are all the same, the description of the analysis method is limited to the analysis period of 1996 to 2000, and only the analysis results are described for the other periods.

Hereinafter, an analysis method using a time-series patent map will be described with reference to the flowchart of FIG.

The patent document analysis software is activated, and the analysis target patent document set file is opened (A-S1).

The patent document analysis software installed in advance in the storage device 5 of the patent document analysis apparatus 1 is activated by operating the input device 9 to open a file name stored in the storage device 5 in advance and a patent document collection of manipulators.

A condition for creating a time-series patent map as basic data for analysis is set (A-S2).

The time series patent map creation condition screen is displayed on the display device 11 by operating the input device 9, and the time series patent map creation condition to be analyzed is input from the display screen by the input device 9.

The main input items necessary for creating the time series patent map to be analyzed are the patent document set A stored as an object of the patent document set of the manipulator, and the display of the number of patent documents on the time series patent map. Target 3C707AS01, 3C707MT F term, analysis period displayed on the horizontal axis of the time-series patent map, 1996-2000, yearly division unit of the analysis target period, display item on the vertical axis It is the number of patent documents which is a unit.

Here, the number of patent classifications that can be input by the patent document analysis software will be described. The maximum number of patent classifications that can be input is 100 in the case of a time-series patent map, and normally, there is one standard patent classification. Therefore, the maximum number that can be input for the analysis target patent classification is 99.

A time-series patent map is created based on the time-series patent map creation conditions (A-S3).

When input of time series patent map creation conditions is completed on the time series patent map creation condition screen and a tab on the screen is operated by the input device 9, a time series patent map (see FIG. 3) is drawn on the display device 11. Is done.

The period of the horizontal axis of the time-series patent map drawn on the display device 11, the division unit of the period, the transition of the number of patent documents by patent classification, etc. are overlooked to confirm whether they are suitable for the purpose of analysis.

The time series patent map is switched to a numerical table to create a time series numerical table of the time series patent map (A-S4).

The time series patent map can be switched to the time series numerical table (see FIG. 4) by operating the tab on the screen on which the time series patent map is drawn with the input device 9.

A correlation matrix is created by executing correlation analysis in the time axis direction of all patent classification combinations in the time-series numerical table (A-S5).

The correlation matrix is created by operating the input device 9 to open the correlation analysis software installed in the storage device 5 of the patent document analysis device 1 in advance and drawing the correlation analysis software drawn on the display device 11. By inputting the data of all patent classification rows of the time-series numerical table (see FIG. 4) in the analysis range of the analysis condition setting screen and specifying the data direction of analysis as rows, the correlation analysis is executed. A correlation matrix (see FIG. 5) in which correlation coefficients in the time axis direction of all combinations of patent classifications in the series numerical value table are described can be created.

The correlation coefficient in the correlation coefficient list composed of a column describing the reference patent classification and the analysis target patent classification of the correlation matrix created in A-S5 and a column describing the reference patent classification is read as similarity. Then, a similarity list of the analysis target patent classification and the reference patent classification is created (A-S6).

If 3C707AS01 is arranged in the first row of the data string at the time of creating the time-series numerical table (see FIG. 4), the similarity list with 3C707AS01 of the 3C707MT F-term is created (see FIG. 6). If the column containing the patent classification name in the first column of the correlation matrix (see FIG. 5) and the 3C707AS01 column in the first data column adjacent to the column are extracted by the operation of the input device 9, A similarity list (see FIG. 6) of the 3C707MT F term 3C707AS01 can be created.

Rearranging the rows of the similarity list of the analysis target patent classification with the reference patent classification, rearranging in descending order using the similarity degree of the similarity list as a sorting key, the analysis target patent A descending similarity list with the reference patent classification of the classification is created (A-S7).

Since the 3C707MT F-term is stratified by the range of the degree of similarity with 3C707AS01, in order to create a set of 3C707MT F-terms belonging to the same stratification range, the degree of similarity with 3C707AS01 is descending. The descending-order similarity list (see FIG. 7) with the C707MT F-term 3C707AS01 is created.

FIG. 7 is a list of descending similarity of 3C707MT of F-term of 3C707MT whose analysis period is 1996 to 2000 with 3C707AS01 (however, except for the stratified result column of the list, FIG. 8, FIG. 9 and FIG. The part excluding the separate result column is the similarity degree descending list), but the analysis results of 2000 to 2004 analyzed by the same method as FIG. 7 are the analysis of FIG. 8 and 2004 to 2008 The result is FIG. 9, and the analysis result from 2008 to 2012 is FIG. 10.

The analysis target patent classification is determined by comparing the similarity degree β of the descending similarity list between the analysis target patent classification and the reference patent classification with the correlation determination reference value magnitude α calculated by the equation (3). Layered into the following three types (A-S8).

The number of samples in the time-series numerical table (see FIG. 4) defining the correlation analysis range of the correlation matrix (see FIG. 5) for calculating the correlation coefficient that becomes the replacement source of the similarity of the 3C707MT F term with 3C707AS01 is 1996. Since there are five years from 2000 to 2000, the magnitude α of the correlation discrimination reference value calculated by the above equation 3 is ± 0.76.

Each patent classification in the descending order similarity list with the 3C707AS01 of the 3C707MT F term of the analysis period from 1996 to 2000 is stratified using the correlation discriminating reference value, α = ± 0.76. FIG.

The patent classification in which the magnitude β of similarity of C707MT F term with 3C707AS01 is in the range of 1 ≧ β ≧ 0.76 is the stability element technical item of 3C707AS01 (FIG. 7 shows the patent classification corresponding to the stability element technical item. Is pear).

The analysis target patent classification in which the degree of similarity β between the reference patent classification and the analysis target patent classification is in the range of 0.76> β> −0.76 is the variability element technical item of the standard patent classification 3C707AS01 (see FIG. 7). NO2 to NO17 correspond to variability element technical items).

The analysis target patent classification in which the degree of similarity β between the reference patent classification and the analysis target patent classification is in the range of −0.76 ≧ β ≧ −1 is a destructive element technical item of the standard patent classification 3C707AS01 (in FIG. The patent classification corresponding to the declining elemental technical item is “No”.

The analysis result from 2000 to 2004 analyzed by the same method as FIG. 7 is FIG. 8, the analysis result from 2004 to 2008 is FIG. 9, and the analysis result from 2008 to 2012 is FIG.

Further, FIG. 11 shows a list of changes in similarity between the 3C707MT F-term and the 3C707AS01 as the analysis period changes, and FIGS. 12 to 15 show the 3C707MT F-term 2008-2012 in the list of FIG. The values of the degree of similarity β of the year are arranged in descending order and divided into an appropriate number and graphed.

With the analysis method described in the first embodiment, the technical items assumed as the elemental technical items of the specific technology can be stratified into the stability element technical item, the variability element technical item, and the decline element technical item. It becomes possible to know the situation that changes with the era of the elemental technical items constituting the specific technology.

(Second Embodiment)
The second embodiment relates to an analysis method based on a matrix patent map, and uses FIG. 1 and FIGS. 16 to 27 by using data of a patent document set B stored as a file name and an object of a patent document set of a manipulator. As will be described with reference, it is assumed that the patent document set of the manipulator is already stored in the storage device 5.
In this embodiment, the same components as those in the first embodiment will be described with the same reference numerals.

The purpose of the analysis of the second embodiment is from the patent document set B consisting of the applicants up to the top 250 in the patent document set of the theme code 3C707 (manipulator) whose application date is 2008-2012, Patent document set BS10 consisting of the top ten applicants of patent documents to which F-term 3C707BS10 (vertical articulated type), which is a patent classification for creating applicants (hereinafter referred to as selector patent classification), is assigned A patent document set BS26 composed of the top ten applicants of patent documents to which selector patent classification F-term 3C707BS26 (double-armed manipulator) is assigned is created, and the patent document set BS10 and the patent document set BS26 The standard patent classification is the same F-term 3C707AS01 (use of manipulator: transport, transfer, article supply) ), And the analysis target patent classification is also the F term of the same patent classification 3C707MT (manipulator control purpose), and the reference patent classification and the analysis target patent of each of the patent document set BS10 and the patent document set BS26 The degree of similarity β of the classification is obtained, and the analysis target patent classification is stratified into the three types of elemental technical items by using the correlation determination reference value α calculated using the β and Equation 3, and different. This is to clarify the difference in the degree of similarity β between the patent document set BS10 extracted by the selector patent classifications 3C707BS10 and 3C707BS26 and the 3C707AS01 of the 3C707MT F-term in the patent document set BS26.

The patent document set BS10 is a group of applicants whose field of expertise is a vertical articulated manipulator (3C707BS10), and the patent document set BS26 is a group of applicants whose field of expertise is a multi-arm manipulator (3C707BS26). If it is read as a group of applicants with different fields of expertise, the elements of different technical fields are determined by clarifying the difference in similarity between the patent document set consisting of applicants of different fields of expertise and the reference patent classification of the analyzed patent classification. It can be used as a hint for introducing new technology because it shows the utilization status of technical items.

  Since the analysis methods of the analysis target patent classifications of the patent document set BS10 and the patent document set BS26 are exactly the same, the description in each analysis step will be described only for the patent document set BS10, and the patent document set BS26. Only the analysis result is described.

Hereinafter, B. of FIG. A description will be given along an analysis flowchart of the matrix patent map.

The patent document analysis software is activated to open the patent document set B (B-S1).

The patent document analysis software preinstalled in the storage device 5 of the patent document analysis device 1 is activated by operating the input device 9 to open a collection of patent documents of file names and manipulators stored in the storage device 5 in advance.

A condition for creating a first matrix patent map serving as base data for obtaining the similarity between the reference patent classification and the analysis target patent classification is set (B-S2).

The input items necessary for creating the first matrix patent map are the patent document set B stored as an object of the patent document set of the manipulator, the patent classification name displayed on the vertical axis of the first matrix patent map, and 3C707AS01. 3C707BS10, 3C707MT F-term, applicant name displayed on the horizontal axis of the first matrix patent map, applicants 1, 2,..., 250, and the number of patent documents as the type of aggregation target.

The creation condition of the first matrix patent map is input, and the first matrix patent map (see FIG. 17) is created (B-S3).

Specifically, by operating the input device 9, a matrix patent map creation condition screen is displayed on the display device 11, and the first matrix patent map creation condition is input from the display screen by the input device 9, When the input is completed, the first matrix patent map (see FIG. 17) is created by operating the tab with the input device 9, and the bird's eye view confirms whether the first matrix map is suitable for the purpose of analysis.

The first matrix patent map is switched to a numerical table to create a first matrix numerical table (B-S4).

The first matrix patent map (see FIG. 17) is drawn on the display device 11, and the screen switching tab is operated with the input device 9 to display the first matrix patent map (see FIG. 17) in the first matrix numerical table (see FIG. 17). 18).

The rearrangement of the columns of the first matrix numerical table is performed by rearranging the number of patent documents in the selector patent classification row in descending order from the left side to the right side to create a second matrix numerical table (B-S5).

The patent document analysis software has a function of rearranging the number of designated rows or columns in the matrix numerical table in ascending or descending order by operation of the input device 9, so in this step, the first matrix numerical table ( The rearrangement of the column of (see FIG. 18) is performed by rearranging the number of rows of the selector patent classification and 3C707BS10 from the left side to the right side in descending order to create the second matrix numerical value table (see FIG. 19).

The applicant's range is determined using the number of patent documents of the selector patent classification of the second matrix numerical table as an evaluation measure, and a third matrix numerical table is created (B-S6).

The applicant's range is specified by using the number of patent documents to which the selector patent classification is assigned as an evaluation scale, and the third matrix numerical table for correlation analysis (see FIG. 20) is created by narrowing down to applicants having a specific technology. For this reason, in the present embodiment, the applicants up to the top 10 in the number of patent documents to which the selector patent classification is assigned are the applicants to be analyzed.

A correlation matrix is created by executing correlation analysis in the applicant's axial direction for all combinations of patent classifications in the third matrix numerical table (B-S7).

The correlation matrix is created by operating the input device 9 to open the correlation analysis software installed in the storage device 5 of the patent document analysis device 1 in advance and drawing the correlation analysis software on the display device 11. By executing the correlation analysis by specifying the data of all patent classification rows of the third matrix numerical table (see FIG. 20) as the analysis range of the analysis condition setting screen and specifying the analysis data direction as rows. A correlation matrix (see FIG. 21) is created in which the correlation coefficients in the applicant's axial direction of all combinations of patent classifications in the third matrix numerical table (see FIG. 20) are described.

The correlation coefficient in the correlation coefficient table composed of a column describing the reference patent classification and the analysis target patent classification of the correlation matrix created in B-S7 and a column describing the reference patent classification is read as similarity. Then, a similarity list of the analysis target patent classification with the reference patent classification is created (B-S8).

If 3C707AS01 is arranged in the first row of the data string at the stage of creating the matrix numerical value table (see FIG. 20), the similarity list with 3C707AS01 of the 3C707MT F-term (see FIG. 22) If the 3C707AS01 column, which is the first column of the data column adjacent to the column in which the patent classification name in the first column of the correlation matrix (see FIG. 21) is written, is extracted by the operation of the input device 9, 3C707MT A similarity list (see FIG. 22) with F-term 3C707AS01 can be created.

FIG. 23 is a similarity list of 3C707MT F-term 3C707AS01 when the selector patent classification 3C707BS26 is used, and comparing FIG. 22 and FIG. 23, the degree of similarity β of the same analysis target patent classification It can be seen that there are various things such as almost the same thing, greatly different things.

Rearranging the rows of the similarity list of the analysis target patent classification with the reference patent classification, rearranging in descending order using the similarity degree of the similarity list as a sorting key, the analysis target patent A descending similarity list with the standard patent classification of the classification is created (B-S9).

Since the 3C707MT F-terms are stratified according to the range of the degree of similarity with 3C707AS01, in order to create a set of 3C707MT F-terms belonging to the same stratification range, the degree of similarity with 3C707AS01 is descended in descending order. A descending-order similarity list (see FIG. 24) with the C707MT F-term 3C707AS01 is created.

24 shows the descending similarity degree list of the selector patent classification 3C707BS10 (however, the portion excluding the stratified result column of the list, and the portion excluding the stratified result column of FIG. 25 is the similar descending order list. However, FIG. 25 shows the analysis result of the selector patent classification 3C707BS26 analyzed by the same method as in FIG.

The analysis target patent classification is determined by comparing the similarity degree β of the descending similarity list between the analysis target patent classification and the reference patent classification with the correlation determination reference value magnitude α calculated by the equation (3). Layered into the following three types (B-S10).

The matrix numerical value table (see FIG. 20) that defines the correlation analysis range of the correlation matrix (see FIG. 21) that calculates the correlation coefficient that is the replacement source of the similarity of the 3C707MT F term with the 3C707AS01 is shown in FIG. Since there are people, the number of samples is 10, and the magnitude of the correlation determination reference value calculated by Equation 3 is α = ± 0.58.

When the selector patent classification is 3C707BS10, each patent classification in the descending similarity list of the 3C707MT F-term of the 3C707MT and the 3C707AS01 is stratified using the correlation discrimination reference value α = ± 0.58. It is.

The patent classification in which the degree of similarity β of 3C707MT F term with 3C707AS01 is in the range of 1 ≧ β ≧ 0.58 (NO2 to NO10 in FIG. 24) is a stability element technical item of 3C707AS01.

The patent classification in which the magnitude β of similarity of 3C707MT F term with 3C707AS01 is in the range of 0.58> β> −0.58 (NO11 to NO18 in FIG. 24) is a variable element technical item of 3C707AS01.

3C707MT F-term 3C707AS01 similarity β is in the range of −0.76 ≧ β ≧ −1 (the corresponding patent classification is not applicable in FIG. 24), and the patent classification is a degrading element technical item of 3C707AS01. .

FIG. 25 shows the analysis result in the case of selector patent classification 3C707BS26 analyzed by the same method as in FIG.

FIG. 26 is a similarity comparison list between the analysis target patent classification and the reference patent classification when the selector patent classification is different, and FIG. 27 is a graph of the similarity comparison list.

From FIG. 27, it can be seen that if the selector patent classification is different, the similarity of the analysis target patent classification with the reference patent classification may be the same or different.

By performing the analysis as described above, it is possible to know the element technical items of the specific technology in the specific applicant set, and further, if the specific applicant set is an applicant set having different fields of expertise, This will lead to new ideas with hints of elemental technology items in different fields.

This invention is important for new technology development because it can easily grasp the transition of element technology items of a specific technology and easily compare the degree of adoption of element technology items among a group of applicants with different strengths. It is thought that it can obtain a simple hint.

DESCRIPTION OF SYMBOLS 1 Patent document analyzer 2 Computer 3 Controller 4 Arithmetic device 5 Storage device 6 I / O
7 Keyboard 8 Mouse 9 Input device 10 Drive such as CD-ROM 11 Display device 12 Printer 13 Patent database 14 Communication circuit

Claims (2)

A patent document analysis method for processing a patent document set accumulated in a storage unit of a computer in which patent document analysis software is installed by the patent document analysis software,
The number of patent documents of the analysis target patent classification that calculates the correlation coefficient in combination with the reference patent classification and the reference patent classification in which one axis is used as a reference for calculating the similarity of the number of patent documents of patent classification The number axis of patent documents that display
The other axis is a time series axis in which the analysis period is divided into time series sections having at least three or more data,
Creating a time series patent map indicating the number of patent documents for each patent classification for each time series section, and converting the time series patent map into a time series numerical table;
Create a correlation matrix in the time axis direction between all patent classifications of the time series numerical table, and the correlation coefficient values of the reference patent classification and the analysis target patent classification of the correlation matrix in the time series patent map The degree of similarity of the time series transition of the number of patent documents of the reference patent classification and the analysis target patent classification is quantified, and the correlation determination changes depending on the numerical value of the similarity and the number of samples used to create the correlation matrix. Stratifying the relationship between the reference patent classification and the analysis-target patent classification into a positive similarity, no similarity, and a negative similarity by comparison with a reference value;
The patent document analysis method characterized by comprising. A patent document analysis how the patent document set stored in the storage unit of the computer patent document analysis software is installed for processing by the patent document analysis software,
Selection of the target patent category, which uses one axis as a standard for calculating the similarity of the number of patent documents between patent categories, the target patent category for calculating the correlation coefficient in combination with the reference patent category, and the target applicant for analysis A patent classification axis consisting of a specific patent classification and
The other axis is an applicant axis in which at least three specific applicants selected from among applicants having the number of patent documents to which the specific patent classification is assigned are arranged,
Create a matrix patent map in which the number of patent documents to which the patent classification and the applicant are simultaneously assigned is written at the intersection of the patent classification axis and the applicant axis, and the matrix patent map is Converting to a matrix numeric table;
Create a correlation matrix in the applicant axis direction of all patent classifications of the patent classification axis of the matrix numerical table, and use the correlation coefficient values of the reference patent classification and the analysis target patent classification of the correlation matrix to calculate the matrix Quantify the similarity between the number of patent documents of the reference patent classification and the analysis patent classification in the applicant axis direction in the patent map,
By comparing the digitized value of the similarity with a correlation criterion value that varies depending on the number of samples used to create the correlation matrix, the relationship between the reference patent classification and the analysis patent classification is positively similar. Stratifying relevant, non-similar, negative-similar,
The patent document analysis method characterized by comprising.

Images