JP2022133401A - Relevance score calculation system, method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a relevance score calculation system, relevance score calculation method, and relevance score calculation program for quantifying the strength of relation between a person and a word.

SOLUTION: In a relevance score calculation system, a keyword reception unit 6 receives a keyword from a searcher, and a search unit 7 acquires, from a storage unit 5, a relevance score based on the total operation time during which a user operated each file for each combination of a word included in each file name and each user who operated the file in an operation log, identifies a pair containing the word corresponding to the search keyword and the relevance score equal to or greater than a threshold from among pairs of a user name, word, and relevance score, and identifies the user name included in the pair as a search result.

SELECTED DRAWING: Figure 5

COPYRIGHT: (C)2022,JPO&INPIT

Description

The present invention relates to a relation score calculation system, a relation score calculation method, and a relation score calculation program for quantifying the strength of relation between a person and a word.

Japanese Patent Application Laid-Open No. 2002-200002 discloses a method for determining an application program being used by a user by means of an operation log containing information necessary for determining how much an application program installed in a personal computer or a mobile information terminal is being used. An information providing device is described. Patent Document 1 also describes that the information providing apparatus determines how much the user uses the application program and determines the user's knowledge level of the application program.

Patent Literature 2 describes a personnel search system that registers, deletes, updates, etc. information about personnel in a profile database and searches for personnel that match a search keyword by referring to profile information. Further, Japanese Patent Laid-Open No. 2002-200002 describes obtaining information about a human resource's specialized field and work by extracting keywords of a document authored by a person and obtaining high-level keywords.

JP 2013-37584 A Japanese Patent Application Laid-Open No. 2005-327028

It is preferable to be able to find a person who is familiar with a specific field or technology, or who has participated in a certain project, within an organization such as a company. It would also be desirable to be able to easily find out which fields and technologies a person is familiar with and which projects he or she has participated in. However, in large organizations such as large corporations, "who is familiar with which field and which technology?" and "who has participated in which project?" put away. This tendency is particularly strong for information such as "who was familiar with which field and which technology in the past" and "who participated in which project in the past". As a result, new employees and temporary employees, in particular, do not know who to ask what they want to hear. As a result, for example, the efficiency of product development may decrease.

Therefore, the inventors of the present invention thought that it would be preferable to be able to clarify the strength of association between people in an organization and words. For example, if the inventors of the present invention can clarify the strength of the relationship between a certain person and the word "artificial intelligence", they can determine whether the person is familiar with the field and technology of "artificial intelligence" and whether the person thought that it would be easy to estimate whether or not they have participated in a project related to "artificial intelligence".

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a relation score calculation system, a relation score calculation method, and a relation score calculation program capable of clarifying the strength of the relation between a person and a word in an organization.

A search system according to the present invention includes a keyword reception unit that receives a keyword from a searcher; a search unit for searching for a user who

A search method according to the present invention receives a keyword from a searcher, and searches for a user related to the keyword based on the total operation time of each file including the keyword in the file name among the operation times of each file of one user. do.

The search program according to the present invention provides a computer with a process of accepting a keyword from a searcher, and based on the total operation time of each file containing the keyword in the file name, out of the operation time of each file by one user. and a process of retrieving related users.

A search system according to the present invention includes a keyword reception unit that receives a keyword from a searcher; a search unit for searching for a user;

A search method according to the present invention receives a keyword from a searcher, and searches for a user related to the keyword based on the total participation time of each event including the keyword in the event name among the participation times of one user's event. .

A search program according to the present invention provides a computer with a process of accepting a keyword from a searcher, and based on the total participation time of each event that includes the keyword in the event name, among the participation times of one user's event, is related to the keyword. and a process of searching for a user who wants to do so.

According to the present invention, it is possible to clarify the strength of the relationship between people and words in an organization.

1 is a block diagram showing a configuration example of a related score calculation system according to the first embodiment of the present invention; FIG. FIG. 4 is a schematic diagram showing an example of an operation log; FIG. 4 is a schematic diagram showing an example of a related score calculated for each combination of a user and a word; 4 is a flow chart showing an example of the progress of processing in the related score calculation system of the first embodiment; It is a block diagram which shows the modification of 1st Embodiment. FIG. 4 is a schematic diagram showing an example of a set of a user name, a word, and a related score stored in a storage unit; It is a block diagram which shows the structural example of the related score calculation system of the 2nd Embodiment of this invention. FIG. 10 is a schematic diagram showing an example of a related score calculation result (a set of pairs of a user name, a word, and a related score); FIG. 4 is a schematic diagram showing an example of a first table; FIG. FIG. 11 is a schematic diagram showing an example of a second table; FIG. 9 is a flow chart showing an example of the progress of processing according to the second embodiment; 10 is a flowchart showing an example of processing for calculating search scores for selected users and keyword words. FIG. 10 is a flowchart showing an example of processing for calculating a score calculation target word and a search score for a keyword user name; FIG. It is a block diagram which shows the structural example of the related score calculation system of the 3rd Embodiment of this invention. FIG. 5 is a schematic diagram showing an example of a screen showing clustering results; FIG. 11 is a block diagram showing a configuration example of a related score calculation system according to a fourth embodiment; FIG. FIG. 4 is a schematic diagram showing an example of schedule information; 1 is a schematic block diagram showing a configuration example of a computer according to each embodiment of the present invention; FIG. 1 is a block diagram showing an overview of the present invention; FIG. FIG. 4 is a block diagram showing another example of the outline of the present invention;

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described with reference to the drawings.

Embodiment 1.
FIG. 1 is a block diagram showing a configuration example of a related score calculation system according to the first embodiment of the present invention. However, FIG. 1 also illustrates devices connected to the relevant score calculation system of the present invention via a communication network.

A personal computer (hereinafter referred to as PC) 91 is connected to the related score calculation system 1 of the present invention via a communication network 10 .

The PC 91 is used, for example, by a person belonging to an organization such as a company or enterprise. It is assumed that each PC 91 is provided within one organization. A case where the organization is a company will be described below as an example. However, the organization to which the person using the PC 91 belongs does not have to be a company or enterprise. Also, the organization may be an organization formed by a plurality of companies or the like, or may be a single company or a division of a company or the like.

Each PC 91 creates and stores one operation log regarding a file by a user who operates the file. As will be described later, the operation logs stored in each PC 91 are collected by the related score calculation system 1 (more specifically, the collection unit 2 of the related score calculation system 1).

FIG. 2 is a schematic diagram showing an example of an operation log created by the PC 91. As shown in FIG. The operation log associates, for example, the file name, the user name of the user who operated the file, the content of the operation, and the date and time when the operation was performed (see FIG. 2). In FIG. 2, the number of the operation log is also illustrated for convenience. Note that FIG. 2 is an example of an operation log, and the operation log is not limited to the example shown in FIG.

The PC 91 writes the file name including the path name as the file name in the operation log.

Also, the PC 91 may determine the user name described in the operation log from, for example, an ID (Identification) used when the user logs into the PC 91 . However, the method by which the PC 91 determines the user name is not limited to this method.

In addition, in each embodiment, a user name is used as a user ID (user identification information).

Examples of operation contents described in the operation log include "file open", "key touch", "update (save)", and "file close". However, the content of the operation described in the operation log is not limited to these, and may be "new creation" or the like.

For example, when the user "Yamada" opens the file "/.../artificial intelligence/.../Machine learning of company A.pptx" at 13:15 on October 10, 2017, the PC 91 Create the first operation log exemplified in 2. Further, for example, when the user "Yamada" performs a key touch (key input) on the file at 13:16 on October 10, 2017, the PC 91 outputs the second create an operation log for Also, for example, when the user "Yamada" updates (saves) the file at 15:17 on October 10, 2017, the PC 91 creates the m-1th operation log illustrated in FIG. do. Also, for example, when the user "Yamada" closes the file at 15:45 on October 10, 2017, the PC 91 creates the m-th operation log illustrated in FIG.

The operation logs from the (m+1)th to the nth shown in FIG. 2 are examples of operation logs when the user "Yamada" operates another file "/.../artificial intelligence/.../deep learning.docx". .

Each PC 91 similarly adds and stores an operation log each time the user operates a file.

A related score calculation system 1 includes a collection unit 2 , a word extraction unit 3 , a score calculation unit 4 and a storage unit 5 .

The collection unit 2 collects operation logs stored in each PC 91 from each PC 91 . As shown in FIG. 2, each operation log includes a file name including a path name and the user name of the user who used the file with that file name. Each operation log also includes information from which a relevance score, which is an index value representing the strength of relevance between a user and a word, can be derived. In the example shown in FIG. 2, the information described as "date and time" and "operation" corresponds to information from which a related score can be derived. However, the related score is derived not from one operation log, but from multiple operation logs.

The word extraction unit 3 performs morphological analysis on each file name (file name including path name) described in each operation log collected by the collection unit 2, thereby extracting the file name including the path name. Extract contained words. However, the word extractor 3 does not extract the same word redundantly. For example, when the word "artificial intelligence" has already been extracted, the word extraction unit 3 ignores the word "artificial intelligence" extracted for the second and subsequent times.

For example, the word extraction unit 3 performs morphological analysis on a file name including the path name "/.../artificial intelligence/.../machine learning of company A.pptx" to extract "artificial intelligence", Words such as "Company A" and "machine learning" are extracted. In the following explanation, it is assumed that the company (organization) to which each user belongs is "A company".

Further, for example, the word extraction unit 3 performs morphological analysis on a file name including a path name of "/.../artificial intelligence/.../deep learning.docx" to extract "artificial intelligence", " Extract words such as “deep learning”. However, since "artificial intelligence" has already been extracted as described above, the word extractor 3 ignores the word "artificial intelligence" extracted here.

The word extraction unit 3 obtains a set of words by performing similar processing on each file name described in each operation log. These words are different from each other.

The score calculation unit 4 calculates, for each combination of each user who operated the file and each word extracted by the word extraction unit 3, a relation score, which is an index value representing the strength of the relation between the user and the word. A user is represented by a user name described in the operation log.

In addition, in the second embodiment of the present invention, etc., an index value representing the strength of the relationship between words is also used. In the present invention, an index value representing the strength of the relationship between a user and a word is called a "relevance score", and an index value representing the strength of the relationship between words is called a "relevance level". Distinguish values.

The score calculation unit 4 calculates a related score for each combination of a user (user name) and a word, and stores the combination of the user name, word, and related score in the storage unit 5 .

The storage unit 5 is a storage device that stores a set of a user (user name), a word, and a related score.

There are multiple methods for calculating the related score. Three types of methods for calculating the related score will be described below. In any of the three methods described below, the score calculation unit 4 calculates a related score for each combination of the user and the word, and stores it in the storage unit 5 .

In the first calculation method, as a relation score between one user (hereinafter referred to as user U) and one word (hereinafter referred to as word W), user U This is a method of calculating the sum of the operation times of It can be said that the longer the operation time is, the stronger the relationship between the user U and the word W is, and the shorter the operation time is, the weaker the relationship between the user U and the word W is. Therefore, operation time can be used as a relevance score.

Here, first, a case where the actual reference time is treated as the operation time will be described as an example.

Note that the file name is a file name including a path name. Therefore, even if the word W is included in the path name, it is treated as being included in the file name. This point also applies to the second calculation method and the third calculation method, which will be described later.

The actual reference time is the time during which the window showing the contents of the file is active in the PC 91 (that is, the time during which the window showing the contents of the file is displayed in front of the user). be.

In the first calculation method of the relation score, the score calculation unit 4 calculates the sum of the actual reference times of the user U for each file containing the word W in the file name, and calculates the sum as the relation between the user U and the word W. score.

If a key is touched, the window representing the contents of the file is active. Therefore, for example, the score calculation unit 4 determines from the operation log the state in which the user U continues to make key touches with respect to one file including the word W in the file name, and calculates the time of the first key touch in that state. to the time of the last key touch is set as the substantial reference time of the user U in the file.

Furthermore, if there is another file that includes the word W in its file name and is operated by the user U, the score calculator 4 similarly calculates the effective reference time for that file.

Then, the score calculation unit 4 calculates the sum of the actual reference times calculated for each file that includes the word W in the file name and is operated by the user U, and sets the sum as the relation score between the user U and the word W.

In addition, if the operation log clearly indicates the start time and end time of the active state, the score calculation unit 4 calculates the actual reference time based on those times indicated in the operation log. You may

Further, the score calculation unit 4 may calculate the time from file open to file close as the operation time. In this case, the score calculation unit 4 calculates the time from the file opening to the file closing for each file that includes the word W in the file name and is operated by the user U. A related score may be used.

Note that it is preferable to treat the actual reference time as the operation time.

A second method of calculating the association score is the association score between one user (user U) and one word (word W). This is a method of calculating the total number of times. The greater the number of key touches, the greater the number of times the user U has operated files containing the word W in the file name. Therefore, it can be said that the larger the number of key touches, the stronger the relationship between the user U and the word W, and the smaller the number of key touches, the weaker the relationship between the user U and the word W. Therefore, the number of key touches can be used as a relevance score.

In the second calculation method, the score calculation unit 4 refers to the operation log and counts the number of key touches when the user U operates a file containing the word W in its file name. Find the number of key touches in .

Furthermore, if there is another file that includes the word W in its file name and is operated by the user U, the score calculator 4 similarly obtains the number of key touches for that file.

Then, the score calculation unit 4 calculates the sum of the number of key touches calculated for each file that includes the word W in the file name and is operated by the user U, and sets the sum as the association score between the user U and the word W. .

A third method of calculating the association score is a method of calculating the association score between one user (user U) and one word (word W) based on two ratios described below. Of these two ratios, one ratio is denoted R ₁ and the other ratio is denoted R ₂ .

_R1 is the ratio of the total operating time of user U for each file containing word W in its file name to the total operating time of all users in the organization for each file containing word W in its file name. That is, R ₁ is represented by formula (1) shown below.

R ₂ is the operation time of the user U for each file including the word W in the file name with respect to the sum of the operation time of the user U for each file including the word of interest in the file name when focusing on individual words. is the ratio of the sum of That is, R ₂ is represented by formula (2) shown below.

_R1 will be explained. The sum of the user U's operation time for each file including the word W in the file name (the numerator on the right side of equation (1)) corresponds to the relation score calculated by the first calculation method described above. That is, the score calculation unit 4 may calculate the sum of the operation time of the user U for each file including the word W in the file name by the method described in the first calculation method.

The sum of the operation times of all users in the organization (the denominator on the right side of equation (1)) for each file containing the word W in the file name will be described. The score calculation unit 4 also calculates the total operation time of the first user in the organization for each file including the word W in the file name, using the method described in the first calculation method above. Similarly, the score calculation unit 4 also calculates the total operation time of the second user in the organization for each file whose W file name includes a word, using the method described in the first calculation method above. Similarly, the score calculation unit 4 calculates the sum of the user's operation time for each file including the word W in the file name for each person belonging to the organization. Further, the score calculation unit 4 calculates the sum of the "sum of the user's operation time for each file containing the word W in the file name" calculated for each person belonging to the organization. This value corresponds to the total operation time of all users in the organization for each file containing the word W in the file name.

For example, assume that the word W is "artificial intelligence" and the user U is "Yamada". It is also assumed that there are 300 users in the organization "A company" to which Yamada belongs. In this case, the score calculation unit 4 obtains the sum of the operation times of the user "Yamada" for files including "artificial intelligence" in the file name as the numerator on the right side of Equation (1). In addition, the score calculation unit 4 calculates the sum of user operation times for files including "artificial intelligence" in the file name for each of the 300 individual users, and further calculates the total "operation time" for each individual user. The sum of "sum of time" is obtained as the denominator of the right side of equation (1). Then, the score calculation unit 4 calculates _R1 by the formula (1).

Next, _R2 will be explained. The numerator on the right side of equation (2) is the same as the numerator on the right side of equation (1). Therefore, the score calculation unit 4 may calculate the sum of the operation time of the user U for each file including the word W in the file name by the method described in the first calculation method.

A description will be given of the sum of the user U's operation time (the denominator on the right side of equation (2)) for each file containing the focused word in the file name when focusing on individual words. The score calculator 4 focuses on the individual words extracted by the word extractor 3 (in other words, selects the individual words one by one). Then, the score calculation unit 4 calculates the total operation time of the user U for each file whose file name includes the word of interest (selected word) by the method described in the first calculation method above. The score calculation unit 4 focuses on the next word (in other words, selects the next word), and calculates the sum of the user U's operation time for each file containing the focused word (selected word) in the file name. , is calculated by the method described in the first calculation method above. In this way, the score calculation unit 4 calculates, for each word, the sum of the operation times of the user U for each file including the word in the file name. Then, the score calculation unit 4 calculates the sum of "the sum of the operation times of the user U for each file containing the word in the file name" calculated for each word. This value corresponds to the denominator on the right side of Equation (2). "The total sum of the user U's operation time for each file containing the focused word in the file name when focusing on individual words (the denominator on the right side of Equation (2))" It can be said that it is the sum of the sums when the sum of the user U's operation time for the file is obtained for each word.

For example, as in the example above, assume that the word W is "artificial intelligence" and the user U is "Yamada". In this case, the score calculation unit 4 obtains the sum of the operation times of the user "Yamada" for files including "artificial intelligence" in the file name as the numerator on the right side of Equation (2). In addition, the score calculation unit 4 calculates the sum of the operation time of the user "Yamada" for each extracted word such as "artificial intelligence", "company A", "deep learning", etc. calculate. Furthermore, the sum of the "sum of operation times" calculated for each individual word is obtained as the denominator of the right side of Equation (2). Then, the score calculation unit 4 calculates _R2 by the formula (2).

After obtaining R ₁ and R ₂ , the score calculation unit 4 calculates the association score between the user U and the word W using the following equation (3).

Relevant score = R ₁ ×log(R ₂ ) (3)

When the relevance score is calculated by this third calculation method, words that are related to many people belonging to the organization have a low relevance score value, and words that are related to a specific person belonging to the organization have a low relevance score value. Relevance score increases. For example, each user belongs to Company A, so the word "Company A" may be associated with each user. However, it is self-evident that the word "A company" is strongly related to each user. Therefore, it is meaningless to increase the association score between the word "A company" and each user, and it is preferable to lower the association score. Also, if only a specific user belonging to an organization is associated with the word "artificial intelligence", it is preferable to increase the association score between that user and the word "artificial intelligence". In the third calculation method, the relevance score can be calculated as such.

FIG. 3 is a schematic diagram showing an example of association scores calculated for each combination of a user and a word. The first related score shown in FIG. 3 is the related score calculated by the first calculation method. The second related score is the related score calculated by the second calculation method. The third related score is the related score calculated by the third calculation method. Although three types of related scores are shown in FIG. 3, the score calculation unit 4 may calculate any one type of related score. However, the score calculator 4 may calculate two or more types of related scores.

As already explained, the score calculation unit 4 calculates a relation score for each combination of a user and a word, and causes the storage unit 5 to store the set of the user name, the word, and the relation score.

The collection unit 2 is implemented by, for example, a CPU (Central Processing Unit) of a computer that operates according to a related score calculation program and a communication interface of the computer. For example, the CPU may read a related score calculation program from a program recording medium such as a program storage device of the computer, and operate as the collection unit 2 using a communication interface according to the related score calculation program. The word extractor 3 and the score calculator 4 are also implemented by, for example, the CPU of the above computer that operates according to the related score calculator program. That is, as described above, the CPU that has read the related score calculation program should operate as the word extractor 3 and the score calculator 4 according to the related score calculation program. The storage unit 5 is implemented by the storage device of the computer described above. Also, the collection unit 2, the word extraction unit 3, and the score calculation unit 4 may be realized by separate hardware.

Also, the related score calculation system 1 may have a configuration in which two or more physically separated devices are connected by wire or wirelessly. This point also applies to other embodiments described later.

Next, the process progress of the first embodiment will be described. FIG. 4 is a flow chart showing an example of the process progress of the related score calculation system of the first embodiment. In addition, detailed description is omitted about the matter already demonstrated.

First, the collection unit 2 collects operation logs from each PC 91 installed in the company (step S1).

Next, the word extraction unit 3 extracts words by performing morphological analysis on the file names (file names including path names) described in each operation log (step S2).

Next, based on each operation log, the score calculation unit 4 calculates the relationship between the user represented by the user name and the word for each combination of the user name described in the operation log and the word extracted in step S2. Calculates a relevance score that indicates the strength of the relevance. Then, the score calculation unit 4 causes the storage unit 5 to store the set of the user name, the word, and the related score (step S3). The score calculator 4 may calculate the related score by any one of the above-described first calculation method, second calculation method, and third calculation method.

As a result, the storage unit 5 stores a plurality of sets of user names, words, and related scores.

According to this embodiment, the score calculation unit 4 calculates a related score for each combination of a user and a word based on each operation log. The first calculation method, the second calculation method, and the third calculation method described above are all basically based on the amount of user U's operation (key touch number of times, operation time, etc.), a larger value is calculated as the related score. Therefore, the strength of the relationship between the user and the word is appropriately quantified for each combination of the user and the word. Therefore, it is possible to clarify the strength of the relationship between the person in the organization and the word.

Next, a modification of the first embodiment will be described. FIG. 5 is a block diagram showing a modification of the first embodiment. A related score calculation system 1 shown in FIG. Elements similar to those shown in FIG. 1 are denoted by the same reference numerals as in FIG. 1, and description thereof is omitted.

In the example shown below, it is assumed that a plurality of sets of user names, words, and related scores calculated by the third calculation method are stored in the storage unit 5 . FIG. 6 is a schematic diagram showing an example of a plurality of sets stored in the storage unit 5. As shown in FIG.

The related score calculation system 1 illustrated in FIG. 5 accepts a word as a search keyword, and searches for the username of the user according to the word. Alternatively, the related score calculation system 1 accepts a user name as a search keyword and searches for words corresponding to the user name. Also, the related score calculation system 1 may be capable of executing each of the above two types of searches.

The keyword reception unit 6 receives a search keyword from a searcher.

The search unit 7 executes a search according to the search keyword.

The output unit 8 outputs search results.

For example, the keyword reception unit 6 receives a search keyword from a terminal device (not shown) used by a searcher via a communication network, and the output unit 8 transmits the search result to the terminal device. do it. A case where the keyword reception unit 6 receives a search keyword in this manner and the output unit 8 outputs the search results in this manner will be described below as an example. However, the mode of accepting search keywords and the mode of outputting search results are not limited to this example. For example, the keyword accepting unit 6 may accept search keywords via an input device (not shown) included in the related score calculation system 1 . The output unit 8 may also output (display) the search results on a display device (not shown) provided in the related score calculation system 1 .

The keyword reception unit 6 and the output unit 8 are implemented, like the collection unit 2, by, for example, a CPU of a computer that operates according to a related score calculation program and a communication interface of the computer. Also, the search unit 7 is realized by the CPU of the computer that operates according to the related score calculation program. Moreover, the keyword reception unit 6, the search unit 7, the output unit 8, and other components may be realized by separate hardware.

Next, an example of processing in which the related score calculation system 1 accepts a word as a search keyword and searches for the username of the user according to the word will be described.

First, the keyword reception unit 6 receives a word as a search keyword from a searcher.

Next, the search unit 7 selects the word corresponding to the search keyword from among the sets of the user name, the word, and the association score stored in the storage unit 5, Identify the tuples that contain the scores, and identify the user names that are included in the tuples as search results.

For example, suppose that the keyword reception unit 6 receives the word "artificial intelligence" as a search keyword. It is also assumed that the above threshold is 0.5. In this case, the search unit 7 identifies a set containing the word "artificial intelligence" and a related score of 0.5 or higher from among the plurality of sets illustrated in FIG. In this example, the first set shown in FIG. 6 is identified. The search unit 7 obtains the user name "Yamada" included in the specified group as a search result. The output unit 8 outputs the search result. If there are a plurality of sets containing the word "artificial intelligence" and a related score of 0.5 or higher, the search unit 7 identifies all the sets, and retrieves the user name obtained from each set as a search result. and Therefore, the number of user names obtained as search results is not limited to one. The output unit 8 may output a plurality of user names obtained as search results, arranging them in descending order of related scores.

Since user names can be searched from words in this way, the searcher can easily know the user names of users who have been strongly involved in the field, technology, project, etc. indicated by the word specified as the search keyword.

Next, an example of processing in which the related score calculation system 1 accepts a user name as a keyword and searches for a word corresponding to the user name will be described.

First, the keyword reception unit 6 receives a user name as a search keyword from a searcher.

Next, the search unit 7 selects the user name corresponding to the search keyword and Identify the tuples containing the relevance scores and identify the words contained in the tuples as search results. At this time, if there are a plurality of sets containing a user name corresponding to the search keyword and a related score equal to or greater than the threshold, the search unit 7 identifies all the sets, and uses the words obtained from each set as the search result. do. Therefore, the number of words obtained as search results is not limited to one. The output unit 8 may output a plurality of words obtained as search results, arranging them in descending order of related scores.

For example, it is assumed that the keyword reception unit 6 has received the user name "Yamada" as a search keyword. It is also assumed that the above threshold is 0.5. In this case, the search unit 7 identifies a set including the user name "Yamada" and a related score of 0.5 or higher from among the plurality of sets illustrated in FIG. In this example, the first, third and fourth sets shown in FIG. 6 are identified. The search unit 7 obtains words included in each identified pair as a search result. That is, the search unit 7 obtains "artificial intelligence", "machine learning" and "deep learning" as search results. The output unit 8 outputs the search result.

In this way, since a user name can be searched for words, the searcher can easily guess the field, technology, etc., that the person with the user name specified as the search keyword is familiar with, or find out that the person has participated. A certain project etc. can be estimated easily.

Note that the above threshold "0.5" is an example, and the threshold does not have to be 0.5. Also, the threshold may be determined according to the calculation method of the related score.

Embodiment 2.
FIG. 7 is a block diagram showing a configuration example of a related score calculation system according to the second embodiment of the present invention. Components similar to those shown in the related score calculation system of the first embodiment (see FIG. 1) and the modified example of the first embodiment (see FIG. 5) are denoted by the reference numerals shown in FIGS. The same reference numerals are given, and the explanation is omitted.

In the second embodiment, the related score calculation system 1 includes a collection unit 2, a word extraction unit 3, a score calculation unit 4, a storage unit 5, a first table generation unit 11, and a second table generation unit. It includes a unit 12 , a keyword reception unit 6 , a search unit 17 and an output unit 8 .

The collection unit 2, the word extraction unit 3, and the score calculation unit 4 are the same as those shown in the first embodiment (see FIG. 1) and the modified example of the first embodiment (see FIG. 5). be.

Also, the keyword reception unit 6 and the output unit 8 are the same as those elements shown in the modified example of the first embodiment (see FIG. 5).

The storage unit 5 is the same as the storage unit 5 in the first embodiment (see FIG. 1) and the modification of the first embodiment (see FIG. 5). However, in this embodiment, the storage unit 5 not only stores a plurality of pairs of user names, words, and related scores, but also stores a first table 21 and a second table 22, which will be described later. Hereinafter, a set of pairs of a user name, a word, and a related score (for example, a plurality of pairs illustrated in FIG. 6) stored in the storage unit 5 by the score calculator 4 will be referred to as a related score calculation result 20 .

Moreover, in this embodiment, the case where the score calculation part 4 calculates a related score by the above-mentioned 3rd calculation method is made into an example, and is demonstrated. However, the score calculator 4 may calculate the related score by the first calculation method or the second calculation method described above.

In the following description, it is assumed that the score calculation unit 4 has already calculated the related score and the storage unit 5 has stored the related score calculation result 20 . FIG. 8 is a schematic diagram showing an example of a related score calculation result 20 (set of pairs of user names, words, and related scores). Here, a case where the related score calculation result 20 illustrated in FIG. 8 is stored in the storage unit 5 will be described as an example.

The first table generation unit 11 generates the first table 21 based on the related score calculation result 20 and causes the storage unit 5 to store the first table 21 . The first table 21 is a table describing the relationship between user IDs, words, and related scores. More specifically, the first table 21 arranges the user names included in the related score calculation result 20 on one of the vertical axis and the horizontal axis, and includes the user names included in the related score calculation result 20. It is a table in which words are arranged on the other axis, and the user having the user name and the related score of the word are described in the column where the user name and the word intersect.

In this example, the case where the first table generation unit 11 generates the first table by arranging user names on the vertical axis and arranging words on the horizontal axis will be described as an example. The generation unit 11 may arrange user names on the horizontal axis and words on the vertical axis.

When arranging the user names along the axis, the first table generation unit 11 arranges the user names of the same user without duplication. For example, in the related score calculation result 20 shown in FIG. 8, the user name "Yamada" is included in the first group, and the user name "Yamada" is included in the second group. This user name "Yamada" is the user name of the same user. Therefore, when arranging the user names along the axis, the first table generation unit 11 only needs to arrange the user name "Yamada" once.

Similarly, when arranging words along the axis, the first table generation unit 11 arranges the same words without duplication. For example, in the related score calculation result 20 shown in FIG. 8, the word "artificial intelligence" is included in a plurality of sets. However, when arranging the words along the axis, the first table generation unit 11 only needs to arrange the word "artificial intelligence" once.

An example of the first table generated by the first table generator 11 is shown in FIG. The first table generation unit 11 arranges the user names (“Yamada”, “Suzuki”, “Tanaka”, etc.) included in the related score calculation result 20 along the direction of the vertical axis (see FIG. 9). . In addition, the first table generation unit 11 arranges the words ("artificial intelligence", "company A", "machine learning", "deep learning", etc.) included in the related score calculation result 20 in the direction of the horizontal axis. (See FIG. 9).

Then, the first table generation unit 11 writes the user having the user name and the related score of the word in the column where the user name and the word intersect. For example, in the example shown in FIG. 8, the association score between the user name "Yamada" and the word "artificial intelligence" is "0.80". Therefore, the first table generator 11 writes "0.80" in the column where "Yamada" and "artificial intelligence" intersect in the first table 21 (see FIG. 9). Also, for example, in the example shown in FIG. 8, the association score between the user name "Yamada" and the word "A company" is "0.10". Therefore, the first table generation unit 11 writes “0.10” in the column where “Yamada” and “A company” intersect in the first table 21 . The first table generation unit 11 similarly writes a related score for each column where a user name and a word intersect.

The first table generation unit 11 causes the storage unit 5 to store the first table 21 generated as described above.

The second table generation unit 12 generates the second table 22 based on the first table 21 and stores it in the storage unit 5 . The second table is a table describing the degree of association, which indicates the strength of the association between words calculated based on the first table. More specifically, in the second table 22, words are arranged on both the vertical axis and the horizontal axis, and in the columns where the words intersect, the strength of the relationship between the words calculated based on the first table 21 is displayed. It is a table describing the degree of relevance that represents the degree of relevance. As already explained, in the present invention, the index value representing the strength of the relationship between the user and the word is called the "relevance score", and the index value representing the strength of the relationship between words is called the "degree of association". distinguish between two types of index values.

FIG. 10 is a schematic diagram showing an example of the second table. In this example, a case where the second table generation unit 12 arranges words in the same order as in the first table 21 on each of the vertical axis and the horizontal axis will be described. For example, in the first table illustrated in FIG. 9, the words are arranged in order of "artificial intelligence", "company A", "machine learning", "deep learning", and so on. The second table generation unit 12 arranges the words in the same order as the order on the vertical axis and the horizontal axis of the second table (see FIG. 10).

Then, the second table generation unit 12 calculates, for each combination of one word and another word, the degree of association representing the strength of the relationship between the words, and displays the Describe the degree of relevance. Note that the second table generation unit 12 also calculates the degree of association between the same words. For example, the second table generation unit 12 also calculates the degree of association between "artificial intelligence" and "artificial intelligence".

Here, the degree of association between words will be described. First, in the first table 21 illustrated in FIG. 9, a case will be described where attention is paid to the two words "artificial intelligence" and "machine learning".

When the association scores of "artificial intelligence" and individual users are arranged in the order of the user names in the first table 21, they are as follows.

0.80, 0.20, 0.60, 0.43,...

In addition, the relation scores of “machine learning” and individual users are arranged in the order of user names in the first table 21 as follows.

0.79, 0.22, 0.58, 0.45, ...

It can be said that the trends of changes in the related scores in the above related score sequences are similar. In this case, the degree of relevance between “artificial intelligence” and “machine learning” is high.

Next, a case of focusing on two words, "artificial intelligence" and "A company", will be described.

As described above, when the association scores of "artificial intelligence" and individual users are arranged in the order of the user name in the first table 21, they are as follows.

0.80, 0.20, 0.60, 0.43,...

Also, when the association scores of "company A" and individual users are arranged in the order of the user names in the first table 21, the results are as follows.

0.10, 0.40, 0.35, 0.05, ...

It can be said that the trend of changes in the related scores in the above sequence of related scores is not similar. In this case, the degree of relevance between "artificial intelligence" and "company A" is low.

The second table generation unit 12 may calculate the correlation coefficient between words as the degree of association between words. Here, a case will be described where the correlation coefficient between word _w1 and word _w2 is calculated as the degree of association between word _w1 and word _w2 . More specifically, the correlation coefficient between word w1 and word w2 is _a correlation function between word w1 and the _row of association scores of individual users and word _w2 and the _row of association scores of individual users. be.

The arrangement of the association scores (numerical values) when the association scores of one word and individual users are arranged in the order of the user name in the first table 21 is referred to as the word sequence. For example, in the example shown in FIG. 9, the series of "artificial intelligence" is as follows.

0.80, 0.20, 0.60, 0.43,...

Let the sequence of word w ₁ be (x ₁ , x ₂ , . . . , x _n ). Let this sequence be x. In the example shown in _FIG . 9, the sequence of words is _a sequence of _numerical values arranged in the vertical direction. Shown side by side. This point also applies to the word _w2 described below.

Also, assume that the sequence of word w ₂ is (y ₁ , y ₂ , . . . , y _n ). Let this sequence be y.

Let n be the number of related scores belonging to the series as described above.

The _second table generation unit 12 may calculate the correlation coefficient between x and y as the degree of association between word _w1 and word w2. Let r be the correlation coefficient between x and y. The second table generation unit 12 calculates the correlation coefficient r between x and y by calculating the following equation (4).

In equation (4), s _xy is the covariance of x and y. Also, s _x is the standard deviation of x, and s _y is the standard deviation of y. x _i is the i-th association score in x. y _i is the i-th association score in y.

は、ｘの平均値である。

is the average value of x.

は、ｙの平均値である。

is the mean value of y.

The second table generation unit 12 calculates the correlation coefficient by calculating the formula (4) for each combination of one word and one word, and uses the correlation coefficient as the degree of association to create a second table 22. When two words forming a combination are different words, the degrees of relevance of the two words are described in two places in the second table 22 . For example, in the example shown in FIG. 10, the degrees of association between "artificial intelligence" and "company A" are described in the first row, second column and the second row, first column, respectively.

The second table generation unit 12 causes the storage unit 5 to store the second table 22 generated as described above.

The related score calculation system 1 of the second embodiment accepts a word as a search keyword, and searches for the username of the user according to the word. Alternatively, the related score calculation system 1 of the second embodiment accepts a user name as a search keyword, and searches for words corresponding to the user name. Also, the related score calculation system 1 of the second embodiment may be capable of executing each of the above two types of searches.

When searching for a user name of a user corresponding to a word specified by a searcher, the keyword reception unit 6 receives the word from the searcher as a search keyword. Based on the first table 21 and the second table 22, the search unit 17 searches for user names corresponding to words corresponding to the search keyword.

When searching for a word corresponding to a user name specified by a searcher, the keyword reception unit 6 receives the user name from the searcher as a search keyword. Based on the first table 21 and the second table 22, the search unit 17 searches for a word corresponding to the user name corresponding to the search keyword.

In the second embodiment, the collection unit 2, the keyword reception unit 6, and the output unit 8 are implemented by a CPU of a computer that operates according to the related score calculation program and a communication interface of the computer. For example, the CPU reads a related score calculation program from a program recording medium such as a program storage device of a computer, and operates as the collection unit 2, the keyword reception unit 6, and the output unit 8 using a communication interface according to the related score calculation program. Just do it. The word extraction unit 3, the score calculation unit 4, the first table generation unit 11, the second table generation unit 12, and the search unit 17 are also realized by the CPU of the above computer that operates according to the related score calculation program, for example. be. That is, as described above, the CPU that has read the related score calculation program executes the word extraction unit 3, the score calculation unit 4, the first table generation unit 11, the second table generation unit 12, and the search unit according to the related score calculation program. It suffices to operate as the unit 17 . In addition, the collection unit 2, the keyword reception unit 6, the output unit 8, the word extraction unit 3, the score calculation unit 4, the first table generation unit 11, the second table generation unit 12, and the search unit 17 are each separate hardware. may be realized by

FIG. 11 is a flowchart showing an example of the progress of processing according to the second embodiment. Detailed descriptions of operations similar to those described in the first embodiment and operations already described in the second embodiment will be omitted.

Steps S1 to S3 are the same as steps S1 to S3 (see FIG. 4) in the first embodiment, and description thereof will be omitted.

After step S3, the first table generation unit 11 generates the first table 21 (see FIG. 9) based on the related score calculation result 20, and stores the first table 21 in the storage unit 5. (Step S4). Since the operation of generating the first table 21 has already been described, the description is omitted here.

Next, the second table generation unit 12 generates a second table 22 (see FIG. 10) based on the first table 21, and stores the second table 22 in the storage unit 5 (step S5). ). Since the operation of generating the second table 22 has already been explained, the explanation is omitted here.

After the first table 21 and the second table 22 are generated, the keyword reception unit 6 receives a search keyword from the searcher (step S6). The keyword accepting unit 6 may accept words as search keywords. Also, the keyword accepting unit 6 may accept a user name as a search keyword.

Next, the search unit 17 obtains search results corresponding to the search keyword based on the first table 21 and the second table 22 (step S7). When the search keyword is a word, the search unit 17 searches for user names corresponding to the word based on the first table 21 and the second table 22 . Moreover, when the search keyword is a user name, the search unit 17 searches for words corresponding to the user name based on the first table 21 and the second table 22 . Details of the operation of step S7 will be described later.

After step S7, the output unit 8 outputs the search result (step S8). The manner in which the output unit 8 outputs the search results is the same as in the modified example of the first embodiment.

Next, the operation of step S7 will be described. First, in step S6, the case where the keyword receiving unit 6 receives a word as a selected keyword and the searching unit 17 searches for a user name corresponding to the word will be described. A word corresponding to a search keyword is hereinafter referred to as a keyword word.

The search unit 17 sequentially selects individual user names and calculates a search score for the selected user names (hereinafter referred to as selected user names) and keyword words. The search score is an index value indicating the strength of the relationship between the user having the selected user name and the word, but it is calculated using the already-described relationship score, and the calculation method is different from that of the relationship score. Therefore, in the following description, the term "search score" is used to distinguish it from the related score.

Assume that the search unit 17 has selected one selected user name. The operation of calculating the search score of the selected user name and the keyword word by the search unit 17 will be described. FIG. 12 is a flow chart showing an example of processing for calculating the search score of a selected user and keyword words.

First, the search unit 17 initializes the value of the search score of the selected user name to 0 (step S11).

Next, the search unit 17 selects one word that has not yet been selected in step S12 from among the words arranged along the axis of the first table (or the axis of the second table) (step S12). The words selected in step S12 are hereinafter referred to as selected words. Note that the selected word may be the same as the keyword word.

Next, the search unit 17 calculates the product of the degree of association between the keyword word and the selected word and the association score between the selected user name and the selected word (step S13). The search unit 17 may read the degree of association used in step S13 from the second table, and read the association score used in step S13 from the first table.

Next, the search unit 17 adds the product calculated in step S13 to the search score (step S14).

Next, the search unit 17 determines whether or not there is a word that has not been selected in step S12 (step S15). If there is an unselected word (Yes in step S15), the search unit 17 repeats the processes after step S12.

If there is no unselected word (No in step S15), the search unit 17 determines the value of the search score at that time as the search score of the selected user name and keyword word, and terminates the process.

The search score calculation process described above can be represented by the following equation (5).

In equation (5), word i means the i-th selected word.

For example, it is assumed that the first table 21 shown in FIG. 9 and the second table 22 shown in FIG. It is also assumed that the keyword word (word corresponding to the search keyword) is "artificial intelligence" and the selected user name is "Yamada". In this case, the search unit 17 calculates the search score for the selected user name "Yamada" and the keyword word "artificial intelligence" using the following formula (see FIGS. 9 and 10).

Search score = 1.00 x 0.80 + 0.07 x 0.10 + 0.87 x 0.79 + 0.79 x 0.82 + ...

It is also assumed that the selected user name is "Suzuki". In this case, the search unit 17 calculates the search score for the selected user name "Suzuki" and the keyword word "artificial intelligence" using the following formula (see FIGS. 9 and 10).

Search score = 1.00 x 0.20 + 0.07 x 0.40 + 0.87 x 0.22 + 0.79 x 0.18 + ...

The search unit 17 obtains a search score for each user name described in the first table 21 through the above process. Then, the search unit 17 obtains user names whose search scores are equal to or higher than the threshold as search results. Therefore, a plurality of user names can be obtained as search results. The output unit 8 outputs the user name obtained as the search result. The output unit 8 may output a plurality of user names obtained as search results, arranging them in descending order of search score.

This search method also uses a second table that indicates the degree of association between words. Therefore, it is possible to obtain the user name of a user who is strongly related to another word that is highly related to the word specified as the search keyword by the searcher as a search result.

For example, in the company to which each user belongs, "Red Rockets" and "Green Rockets" are the product names of important products, and the two products are closely related. In this case, according to the above search method, even if "Red Rockets" is used as a search keyword, not only the user names of users strongly related to "Red Rockets" but also User names of users who are strongly related to a certain "Green Rockets" can also be obtained as a search result. Therefore, it is possible to search a wide range of user names based on the words specified by the searcher as search keywords.

In addition, in this embodiment, the second table is generated using the words included in the operation logs collected from the PC 91 provided in the company. Therefore, words that are often used uniquely to the company, such as "Red Rockets" and "Green Rockets" as exemplified above, can be included in the second table. If the first table and the second table were to be created manually, not only would it take an enormous amount of time and effort, but words and the like unique to the organization would easily escape from the first table and the second table. Therefore, it is possible to easily create the first table and the second table without omitting the words used uniquely by the organization. can be searched.

Next, in step S6, the case where the keyword receiving unit 6 receives a user name as a selected keyword and the searching unit 17 searches for a word corresponding to the user name will be described. A user name corresponding to a search keyword is hereinafter referred to as a keyword user name.

The search unit 17 sequentially selects individual words and calculates a search score for the selected words and the keyword user name. These selected words are referred to as score calculation target words. Further, even after selecting one score calculation target word, as will be described later, the words are separately selected sequentially (see step S22 in FIG. 13 described later). A word selected in step S22, which will be described later, is referred to as a selected word.

Assume that the search unit 17 has selected one score calculation target word. The operation of the search unit 17 for calculating the search score of the score calculation target word and the keyword user name will be described below. FIG. 13 is a flow chart showing an example of processing for calculating a search score for a score calculation target word and a keyword user name.

First, the search unit 17 initializes the value of the search score of the score calculation target word to 0 (step S21).

Next, the search unit 17 selects one word that has not been selected in step S22 from among the words arranged along the axis of the first table (or the axis of the second table) (step S22). As already mentioned, the words selected in step S22 are referred to as selected words. Note that the selected word may be the same as the score calculation target word.

Next, the search unit 17 calculates the product of the relevance score between the keyword user name and the selected word and the degree of relevance between the score calculation target word and the selected word (step S23). The search unit 17 may read the association score used in step S23 from the first table, and read the association degree used in step S23 from the second table.

Next, the search unit 17 adds the product calculated in step S23 to the search score (step S24).

Next, the search unit 17 determines whether or not there is a word that has not been selected in step S22 (step S25). If there is an unselected word (Yes in step S25), the search unit 17 repeats the processes after step S22.

If there is no unselected word (No in step S25), the search unit 17 determines the value of the search score at that time as the search score of the score calculation target word and the keyword user name, and ends the process.

The search score calculation process described above can be represented by the following equation (6).

In equation (6), word i means the i-th selected word.

For example, it is assumed that the first table 21 shown in FIG. 9 and the second table 22 shown in FIG. It is also assumed that the keyword user name (user name corresponding to the search keyword) is "Yamada" and the score calculation target word is "artificial intelligence". In this case, the search unit 17 calculates the search score of the score calculation target word “artificial intelligence” and the keyword user name “Yamada” by the following formula (see FIGS. 9 and 10).

Search score = 0.08 x 1.00 + 0.10 x 0.07 + 0.79 x 0.87 + 0.82 x 0.79 + ...

It is also assumed that the score calculation target word is "A company". In this case, the search unit 17 calculates the search score for the score calculation target word "Company A" and the keyword user name "Yamada" using the following formula (see FIGS. 9 and 10).

Search score = 0.08 x 0.07 + 0.10 x 1.00 + 0.79 x 0.09 + 0.82 x 0.11 + ...

The search unit 17 obtains a search score by the above-described processing for each of the score calculation target words that are sequentially selected. Then, the search unit 17 obtains words whose search scores are equal to or higher than the threshold as search results. Therefore, a plurality of words can be obtained as search results. The output unit 8 outputs words obtained as search results. The output unit 8 may output a plurality of words obtained as search results, arranging them in descending order of search scores.

A second table indicating the degree of relevance between words is also used in the search method for searching for words from user names as described above. Therefore, not only words highly related to the user name represented by the keyword user name specified by the searcher, but also other words highly related to that word can be obtained as search results.

For example, as in the above example, in the company to which each user belongs, "Red Rockets" and "Green Rockets" are the product names of important products, and the two products are strongly related. It is also assumed that the user "Yamada" and the word "Red Rockets" have a strong relationship. In this case, when a searcher specifies "Yamada" as a keyword user name, not only the word "Red Rockets" which is strongly related to "Yamada" but also "Green Rockets" which is strongly related to the word "Red Rockets" will be retrieved. can be obtained as Since the relationship between "Red Rockets" and "Green Rockets" is strong, "Yamada", which has a strong relationship with "Red Rockets", is also considered to have a strong relationship with "Green Rockets". According to the above method, it is possible to widely search for words that are considered to be strongly related to the user indicated by the user name specified as the search keyword.

Moreover, as already explained, the first table and the second table can be easily created without omitting the words used uniquely to the organization.

As described above, in the present embodiment, when the keyword accepting unit 6 accepts a word as a selected keyword and the searching unit 17 searches for a user name according to the word, a wide range of user names can be searched. Further, when the keyword accepting unit 6 accepts a user name as a selected keyword and the searching unit 17 searches for a word corresponding to the user name, a wide range of words can be searched. In addition, the first table and the second table can be easily created without omitting the words used by the organization.

Embodiment 3.
FIG. 14 is a block diagram showing a configuration example of a related score calculation system according to the third embodiment of the present invention. The related score calculation system of the third embodiment includes, in addition to the components included in the related score calculation system of the second embodiment (see FIG. 7), a clustering unit 31, a cluster output unit 32, an exclusion target word and a reception unit 33 . 14, elements similar to those shown in FIG. 7 are denoted by the same reference numerals as in FIG. 7, and description thereof is omitted. Note that each of the first table generation unit 11 and the second table generation unit 12 performs operations described later in addition to the operations described in the second embodiment.

The clustering unit 31 reads the second table 22 (see, for example, FIG. 10) from the storage unit 5, and clusters the words described in the second table 22 based on the second table 22. conduct. The clustering unit 31 clusters words by a clustering method such as the k-means method or a hierarchical clustering algorithm. In a clustering method such as the k-means method or hierarchical clustering algorithm, the second table 22 is used as input data, and the words aligned along the axis of the second table 22 can be clustered. The clustering unit 31 assigns, for example, a cluster number as cluster identification information to each cluster, and stores the relationship between the cluster and the words belonging to the cluster in the storage unit 5 as the clustering result 23 .

The cluster output unit 32 receives, for example, a clustering result output request from a terminal device (not shown) of an administrator (hereinafter simply referred to as an administrator) of the related score calculation system 1 via a communication network, and outputs the clustering result. In response to the output request, the screen information of the screen showing the clustering result is transmitted to the terminal device (not shown). FIG. 15 is a schematic diagram showing an example of a screen showing clustering results. On the screen showing the clustering result, for example, the cluster number and each word belonging to the cluster are displayed for each cluster. A check box 41 is also displayed with each cluster number and each word. As shown in FIG. 15, the number of words belonging to each cluster can be different. A confirmation button 42 is included in the screen showing the clustering result.

The cluster output unit 32 reads the clustering result 23 from the storage unit 5 upon receiving a clustering result output request from the administrator's terminal device via the communication network. Then, based on the clustering result 23, the cluster output unit 32 displays the cluster number and the words belonging to the cluster indicated by the cluster number, displays the check box 41 with each cluster number and each word, and further, The screen information of the screen including the enter button 42 (for example, the screen illustrated in FIG. 15) is generated. Then, the cluster output unit 32 transmits the screen information to the administrator's terminal device via the communication network.

When receiving the screen information from the cluster output unit 32, the administrator's terminal device displays, for example, the screen shown in FIG. 15 based on the screen information.

The administrator confirms the screen illustrated in FIG. 15 (screen for displaying the clustering result of words), and puts a check in the check box 41 corresponding to the word determined by the administrator to be excluded. When the administrator determines that all words belonging to one cluster should be excluded, the administrator checks the check box 41 corresponding to that cluster. A check in the check box 41 corresponding to a word means that the word should be excluded. Also, the fact that the check box 41 corresponding to a cluster is checked means that it has been determined that all the words belonging to that cluster should be excluded. Here, "to be excluded" means to be excluded from the words arranged on the horizontal axis of the first table and the respective axes of the second table.

The confirm button 42 is a button for inputting that the determination of whether or not each word should be excluded has been completed. When the confirmation button 42 is clicked by the administrator, the terminal device of the administrator determines the word that the administrator has determined to be excluded according to which check box 41 is checked. The words determined to be excluded are sent to the related score calculation system 1.

The exclusion target word reception unit 33 of the related score calculation system 1 receives, via the communication network, the word transmitted from the administrator's terminal device (the word determined by the administrator to be excluded). The number of words received by the exclusion target word reception unit 33 is not limited to one. It can be said that the exclusion target word receiving unit 33 receives designation of words to be excluded from the words arranged on the horizontal axis of the first table and the respective axes of the second table.

The exclusion target word reception unit 33 notifies the first table generation unit 11 of the words (words to be excluded) received from the administrator's terminal device. When the first table generation unit 11 receives the notification of the word, the first table generation unit 11 removes the notified word from the words arranged on the horizontal axis of the already generated first table 21, and regenerates the first table 21. Generate. Then, the first table generation unit 11 replaces the first table 21 stored in the storage unit 5 with the regenerated first table 21 .

When the first table generator 11 regenerates the first table 21 , the second table generator 12 regenerates the second table based on the newly generated first table 21 . At this time, the second table generation unit 12 generates the second table by excluding the words received by the exclusion target word reception unit 33 from the words arranged along each axis. Then, the second table generation unit 12 replaces the second table 22 stored in the storage unit 5 with the regenerated second table 22 .

The operation of regenerating the first table 21 and the operation of regenerating the second table 22 are the same as those of the first table 21 described in the second embodiment, except that the number of words arranged along the axis is reduced. 21 and the second table 22 are the same.

The cluster output unit 32 and the exclusion target word reception unit 33 are realized by, for example, a CPU of a computer that operates according to the related score calculation program and a communication interface of the computer. Also, the clustering unit 31 is realized by the CPU of the computer that operates according to the related score calculation program.

According to this embodiment, words determined by the administrator to be excluded from the first and second tables can be excluded from the first and second tables. As a result, for example, when a user name is used as a search keyword to perform a word search, the words specified by the administrator as words to be excluded can be prevented from being included in the search results.

Examples of words that are determined by the administrator to be excluded will be described. For example, a user "Yamada" may include the user name "Yamada" in the name of a file that is newly created or browsed. In that case, when a word search is performed using the user name "Yamada" as a search keyword, the word "Yamada" may be included in the search results. However, when performing a word search using the user name "Yamada" as a search keyword, even if the word "Yamada" is included in the search results, it does not make much sense for the searcher.

Therefore, for example, when the screen shown in FIG. 15 is displayed, the administrator determines that the word "Yamada", which is clearly considered to be the same character string as the user name, should be excluded. , and click the confirm button 42. As a result, the first table generator 11 generates a new first table by excluding "Yamada" from the word arrangement on the horizontal axis. Subsequently, the second table generation unit 12 generates a new second table by excluding "Yamada" from the word sequences on the vertical and horizontal axes. As a result, for example, when performing a word search using the user name "Yamada" as a search keyword, it is possible to prevent the word "Yamada" from being included in the search results.

Note that the cluster output unit 32 may display a screen showing the clustering results on a display device (not shown) included in the related score calculation system 1 . In addition, the exclusion target word reception unit 33 operates the screen displayed on the display device (inputting a check in the check box 41 and confirming the confirmation button 42 ) may be accepted, and designation of words to be excluded may be accepted according to the operation.

Embodiment 4.
In the first to third embodiments, the collection unit 2 collects operation logs, the word extraction unit 3 extracts words from each file name described in each operation log, and the score calculation unit 4 shows a case in which a relevance score is calculated for each combination of each user who manipulated the file and each extracted word.

In the fourth embodiment, an example will be described in which the collection unit collects schedule information instead of file operation logs.

FIG. 16 is a block diagram showing a configuration example of a related score calculation system according to the fourth embodiment. The related score calculation system 1 shown in FIG. 16 replaces the collection unit 2, the word extraction unit 3 and the score calculation unit 4 in the first embodiment with a collection unit 52, a word extraction unit 53 and a score calculation unit 54. be. The storage unit 5 is the same as the storage unit 5 in the first embodiment.

Also, the related score calculation system 1 shown in FIG. 16 is connected via a communication network 10 to a schedule management server 61 that holds schedule information for each person belonging to an organization.

The collecting unit 52 collects from the schedule management server 61 the schedule information of each person belonging to the organization, which is stored in the schedule management server 61 . The schedule information may be stored in individual PCs (not shown in FIG. 16). In this case, the collection unit 52 may collect the schedule information of each person belonging to the organization from each individual PC.

Hereinafter, the identification information of a person belonging to an organization is referred to as person identification information. Also, a case where a person's name such as "Yamada" is used as the person identification information will be described as an example.

FIG. 17 is a schematic diagram showing an example of schedule information. For example, as shown in FIG. 17, the schedule information is described by associating the identification information of a person belonging to an organization, the event name, and the time zone in which the person was involved in the event having that event name. . A time period is represented by a start date and time and an end date and time.

The first schedule information shown in FIG. 17 is that a person named "Yamada" was involved in (in other words, attended) the "Artificial Intelligence Development Conference" from 10:00 on August 17, 2017 to 12:00 on the same day. That means. Hereinafter, the time during which a certain person participates in a certain event is referred to as event participation time. The event participation time for one event is the time obtained by subtracting the start date and time from the end date and time described in association with the event.

The word extraction unit 53 extracts words included in the event names by executing morphological analysis on each event name described in each schedule information collected by the collection unit 52 .

However, the word extraction unit 53 does not extract the same word redundantly. For example, when the word extraction unit 53 has already extracted the word "artificial intelligence", it ignores the word "artificial intelligence" extracted for the second and subsequent times. This point is the same as that of the word extraction unit 3 already described.

For example, the word extraction unit 53 extracts words such as "artificial intelligence", "development", and "meeting" by executing morphological analysis on the event name "artificial intelligence development conference".

The word extraction unit 53 obtains a set of words by performing this process on each event name described in each schedule information. The words belonging to this set are different from each other because the same word is not extracted redundantly.

The score calculation unit 54 calculates, for each combination of each person belonging to the organization and each word extracted by the word extraction unit 53, a relationship score indicating the strength of the relationship between the person and the word. This relevance score is similar to the relevance scores in the first to third embodiments. However, in the fourth embodiment, the number of key touches is not used as a related score. That is, the word extraction unit 53 does not calculate the related score by the second calculation method described in the first embodiment.

Two types of methods will be described below as the method of calculating the related score in the fourth embodiment. In any of the two types of methods described below, the score calculation unit 54 calculates a related score for each combination of a person and a word, and causes the storage unit 5 to store the score.

The first calculation method of the related score in the fourth embodiment is the same method as the first calculation method of the related score in the first embodiment. However, event participation time is used instead of file operation time.

A first method of calculating the association score in the fourth embodiment is to set word W as an association score between one person (hereinafter referred to as person H) and one word (hereinafter referred to as word W) as an event This is a method of calculating the sum of the event participation time of person H to each event included in the first name. It can be said that the longer the event participation time is, the stronger the relation between the person H and the word W is, and the shorter the event participation time is, the weaker the relation between the person H and the word W is. Therefore, event attendance time can be used as a relevance score.

In the first calculation method, the score calculation unit 54 calculates the sum of the event participation times of the person H for each event including the word W in the event name, and sets the sum as the relation score between the person H and the word W.

That is, the score calculation unit 54 calculates the event participation time for each event in which the person H participates in which the word W is included in the event name, and sets the sum as the association score between the person H and the word W. FIG.

The second calculation method of the relation score in the fourth embodiment is similar to the third calculation method of the relation score in the first embodiment. However, event participation time is used instead of file operation time.

A second method of calculating a relation score in this embodiment is a method of calculating a relation score between one person (person H) and one word (word W) based on two ratios described below. Of these _two proportions, _one is denoted as Q1 and the other as Q2.

_Q1 is the ratio of the total event participation time of person H for each event containing word W in the event name to the total event participation time of all persons in the organization for each event containing word W in the event name. . That is, _Q1 is represented by the following equation (7).

_Q2 is the event participation time of person H for each event including the word W in the event name with respect to the sum of the event participation time of the person H for each event including the focused word in the event name when focusing on individual words. is the ratio of the sum of That is, Q2 is represented by the following formula (8).

_Q1 will be explained. The sum of event participation times of person H (the numerator on the right side of equation (7)) for each event including the word W in the event name corresponds to the related score calculated by the first calculation method described above. That is, the score calculation unit 54 may calculate the sum of the event participation times of the person H for each event including the word W in the event name by the method described in the first calculation method.

The sum of event participation times of all persons in the organization (the denominator on the right side of Equation (7)) for each event containing the word W in the event name will be described. The score calculation unit 54 also calculates the sum of the event participation times of the first person in the organization for each event including the word W in the event name, using the method described in the first calculation method above. Similarly, the score calculation unit 54 calculates the sum of the event participation times of the second person in the organization for each event including the word W in the event name, using the method described in the first calculation method above. Similarly, the score calculation unit 54 calculates the sum of the event participation time of each person belonging to the organization for each event including the word W in the event name. Furthermore, the score calculation unit 54 calculates the sum of "the sum of the event participation time of people for each event including the word W in the event name" calculated for each person belonging to the organization. This value corresponds to the sum of event participation times of all persons in the organization for each event containing the word W in the event name.

The score calculator 54 calculates _Q1 by the calculation of Equation (7) using the numerator and denominator on the right side of Equation (7) calculated as described above.

Next, _Q2 will be explained. The numerator on the right side of equation (8) is the same as the numerator on the right side of equation (7). Therefore, the score calculation unit 54 may calculate the sum of the event participation time of the person H for each event including the word W in the event name by the method described in the first calculation method.

A description will be given of the total sum of the event participation times of the person H (the denominator on the right side of Equation (8)) for each event that includes the focused word in the event name when focusing on individual words. The score calculator 54 focuses on the individual words extracted by the word extractor 53 (in other words, selects the individual words one by one). Then, the score calculation unit 54 calculates the sum of the event participation time of the person H for each event including the focused word (selected word) in the event name by the method described in the first calculation method above. The score calculation unit 54 focuses on the next word (in other words, selects the next word), and calculates the sum of the event participation time of the person H for each event including the focused word (selected word) in the event name. , is calculated by the method described in the first calculation method above. In this manner, the score calculation unit 54 calculates, for each word, the sum of the event participation times of the person H for each event including the word in the event name. Then, the score calculation unit 54 calculates the sum of "the sum of the event participation time of the person H for each event including the word in the event name" calculated for each word. This value corresponds to the denominator of equation (8). "When focusing on individual words, the total sum of the event participation time of person H for each event including the focused word in the event name (the denominator on the right side of Equation (8))" It can be said that it is the "sum of the sums when the sum of the event participation times of the person H for the event is obtained for each word".

The score calculation unit 54 calculates _Q2 by the calculation of Equation (8) using the numerator and denominator on the right side of Equation (8) calculated as described above.

After obtaining Q ₁ and Q ₂ , the score calculation unit 54 calculates the association score between the person H and the word W using the following equation (9).

Association score = Q ₁ ×log(Q ₂ ) (9)

When the relation score is calculated by the second calculation method in the fourth embodiment, the same effect as when the relation score is calculated by the third calculation method described in the first embodiment can be obtained.

The collection unit 52 is implemented by, for example, a CPU of a computer that operates according to a related score calculation program and a communication interface of the computer. For example, the CPU may read the related score calculation program from a program recording medium such as a program storage device of the computer, and operate as the collection unit 52 using the communication interface according to the related score calculation program. The word extractor 53 and the score calculator 54 are also realized by the CPU of the above computer that operates according to the related score calculator program, for example. That is, as described above, the CPU that has read the related score calculation program should operate as the word extractor 53 and the score calculator 54 according to the related score calculation program. The storage unit 5 is implemented by the storage device of the computer described above. Also, the collecting unit 52, the word extracting unit 53, and the score calculating unit 54 may be realized by separate hardware.

The fourth embodiment may be applied to the modification of the first embodiment (see FIG. 5), the second embodiment (see FIG. 7), and the third embodiment (see FIG. 14). That is, in the modified example of the first embodiment (see FIG. 5), the collection unit 2, the word extraction unit 3 and the score calculation unit 4 are replaced with the collection unit 52, the word extraction unit 53 and the score described in the fourth embodiment. You may replace with the calculation part 54. FIG. Further, in the second embodiment (see FIG. 7), the collection unit 2, the word extraction unit 3 and the score calculation unit 4 are replaced with the collection unit 52, the word extraction unit 53 and the score calculation unit 54 described in the fourth embodiment. can be replaced with Further, in the third embodiment (see FIG. 14), the collection unit 2, the word extraction unit 3 and the score calculation unit 4 are replaced with the collection unit 52, the word extraction unit 53 and the score calculation unit 54 described in the fourth embodiment. can be replaced with

FIG. 18 is a schematic block diagram showing a configuration example of a computer according to each embodiment of the present invention. Computer 1000 includes CPU 1001 , main memory 1002 , auxiliary memory 1003 , interface 1004 , and communication interface 1005 .

A related score calculation system 1 according to each embodiment of the present invention is implemented in a computer 1000 . The operation of the related score calculation system 1 is stored in the auxiliary storage device 1003 in the form of a related score calculation program. The CPU 1001 reads out the associated score calculation program from the auxiliary storage device 1003, develops it in the main storage device 1002, and executes the above processing according to the associated score calculation program.

Secondary storage 1003 is an example of non-transitory tangible media. Other examples of non-transitory tangible media include a magnetic disk, a magneto-optical disk, a CD-ROM (Compact Disk Read Only Memory), a DVD-ROM (Digital Versatile Disk Read Only Memory) connected via the interface 1004, A semiconductor memory etc. are mentioned. Further, when this program is distributed to the computer 1000 via a communication line, the computer 1000 receiving the distribution may develop the program in the main storage device 1002 and execute the above processing.

Also, the program may be for realizing part of the above-described processing. Furthermore, the program may be a differential program that achieves the above-described processing in combination with another program already stored in the auxiliary storage device 1003. FIG.

Also, part or all of each component may be realized by general-purpose or dedicated circuitry, processors, etc., or combinations thereof. These may be composed of a single chip, or may be composed of multiple chips connected via a bus. A part or all of each component may be implemented by a combination of the above-described circuit or the like and a program.

When a part or all of each component is realized by a plurality of information processing devices, circuits, etc., the plurality of information processing devices, circuits, etc. may be arranged centrally or distributedly. For example, the information processing device, circuits, and the like may be implemented as a client-and-server system, a cloud computing system, or the like, each of which is connected via a communication network.

Next, an outline of the present invention will be described. FIG. 19 is a block diagram showing the outline of the present invention. The related score calculation system of the present invention includes a collection unit 82 , a word extraction unit 83 and a related score calculation unit 84 .

The collecting unit 82 (for example, the collecting unit 2) collects operation logs, which are records of user's file operations, from the terminal device.

A word extraction unit 83 (for example, word extraction unit 3) extracts words from the file names described in each operation log.

The association score calculator 84 (for example, the score calculator 4) calculates a association score representing the strength of the association between the user and the word based on each operation log.

Such a configuration can clarify the strength of association between people and words in an organization.

FIG. 20 is a block diagram showing another example of the outline of the present invention. The related score calculation system of the present invention includes a collection unit 86 , a word extraction unit 87 and a related score calculation unit 88 .

The collection unit 86 (for example, the collection unit 52) collects schedule information describing a person, an event name, and a time period during which the person was involved in an event having that event name.

The word extractor 87 (for example, the word extractor 53) extracts words from the event names described in each schedule information.

The association score calculator 88 (for example, the score calculator 54) calculates a association score representing the strength of the association between a person and a word based on each piece of schedule information.

Such a configuration also makes it possible to clarify the strength of the relationship between people and words in the organization.

Each of the embodiments of the present invention described above can also be described in the following supplementary notes, but is not limited to the following.

(Appendix 1)
a collection unit that collects operation logs, which are records of file operations performed by users, from terminal devices;
a word extraction unit that extracts words from file names described in each operation log;
A relation score calculation system, comprising: a relation score calculation unit that calculates a relation score representing a strength of relation between a user and a word based on each operation log.

(Appendix 2)
The relation score calculation unit calculates, as the relation score between the one user and the one word, the sum of the operation times of the one user for each file containing the one word in the file name. calculation system.

(Appendix 3)
The relation score calculation unit calculates, as a relation score between one user and one word, the sum of the number of key touches when the one user operates each file including the one word in the file name. The associated score calculation system described in .

(Appendix 4)
The related score calculation unit
The relevance score for one user and one word,
a ratio of the total operating time of the one user for each file containing the one word in the file name to the total operating time of all users in the organization for each file containing the one word in the file name; ,
Said one user for each file including said one word in its file name with respect to the sum of said sums when the sum of operation times of said one user for each file including said word in its file name is obtained for each word The related score calculation system according to Appendix 1.

(Appendix 5)
a keyword reception unit that receives a search keyword;
a search unit for executing a search according to a search keyword,
The keyword reception unit is
accepts words as search keywords,
The search unit is
The related score calculation system according to any one of supplementary notes 1 to 4, wherein the user ID corresponding to the word is retrieved based on the related score.

(Appendix 6)
a keyword reception unit that receives a search keyword;
a search unit for executing a search according to a search keyword,
The keyword reception unit is
Receiving a user ID as a search keyword,
The search unit is
The related score calculation system according to any one of supplementary notes 1 to 5, wherein a word corresponding to the user ID is searched based on the related score.

(Appendix 7)
a first table generation unit that generates a first table that describes the relationship between user IDs, words, and association scores;
a second table generation unit for generating a second table describing the degree of association representing the strength of association between words calculated based on the first table;
a keyword reception unit that receives a search keyword;
a search unit for executing a search according to a search keyword,
The keyword reception unit is
accepts words as search keywords,
The search unit is
The related score calculation system according to any one of appendices 1 to 4, wherein a user ID corresponding to the word is searched based on the first table and the second table.

(Appendix 8)
a first table generation unit that generates a first table that describes the relationship between user IDs, words, and association scores;
a second table generation unit that generates a second table describing the degree of association representing the strength of association between words calculated based on the first table;
a keyword reception unit that receives a search keyword;
a search unit for executing a search according to a search keyword,
The keyword reception unit is
Receiving a user ID as a search keyword,
The search unit is
The related score calculation system according to any one of appendices 1 to 4 and 7, wherein a word corresponding to the user ID is searched based on the first table and the second table.

(Appendix 9)
a clustering unit that clusters words based on the second table;
a word presentation unit that presents words for each cluster;
a deletion target reception unit that receives designation of words to be excluded from the first table and the second table among the presented words,
The first table generation unit
Regenerate the first table by excluding words specified as words to be excluded;
The second table generation unit
The related score calculation system according to appendix 7 or appendix 8, wherein a second table is generated again based on the first table.

(Appendix 10)
a collection unit for collecting schedule information describing a person, an event name, and a time period during which the person was involved in an event having the event name;
a word extraction unit for extracting words from event names described in each schedule information;
A relation score calculation system, comprising: a relation score calculation unit that calculates a relation score representing a strength of relation between a person and a word based on each schedule information.

(Appendix 11)
Collect operation logs, which are records of user's file operations, from the terminal device,
Extract words from the file name described in each operation log,
A relation score calculation method, comprising: calculating a relation score representing a strength of relation between a user and a word based on each operation log.

(Appendix 12)
collecting schedule information that describes a person, an event name, and the time period during which the person was involved in an event having the event name;
Extract words from the event name described in each schedule information,
A relation score calculation method characterized by calculating a relation score representing the strength of relation between a person and a word based on each piece of schedule information.

(Appendix 13)
to the computer,
Collection processing for collecting operation logs, which are records of user's file operations, from the terminal device;
A word extraction process that extracts words from the file name described in each operation log, and
A relation score calculation program for executing a relation score calculation process for calculating a relation score representing the strength of relation between a user and a word based on each operation log.

(Appendix 14)
to the computer,
a collection process of collecting schedule information describing a person, an event name, and a time period during which the person was involved in an event having the event name;
Word extraction processing for extracting words from the event name described in each schedule information, and
A relation score calculation program for executing a relation score calculation process for calculating a relation score representing the strength of relation between a person and a word based on each schedule information.

INDUSTRIAL APPLICABILITY The present invention is preferably applied to a relation score calculation system that quantifies the strength of relation between a person and a word.

1 related score calculation system 2, 52 collection unit 3, 53 word extraction unit 4, 54 score calculation unit 5 storage unit 6 keyword reception unit 7, 17 search unit 8 output unit 31 clustering unit 32 cluster output unit 33 exclusion target word reception unit

Claims (9)

a keyword reception unit that receives keywords from a searcher;
a search unit that searches for a user associated with the keyword based on a sum of operation times of files containing the keyword in the file name among operation times of files by one user. The search unit is
Based on the ratio of the total operation time of the one user for each file containing the keyword in the file name to the total operation time of all users in the organization for each file containing the keyword in the file name, 2. The search system of claim 1, wherein searching for users associated with a keyword. 3. The search system according to claim 1, wherein said operation time is a time during which a window showing contents of a file is active. Receiving keywords from searchers,
A search method for searching for a user associated with a keyword based on a sum of operation times of files including the keyword in the file name, among operation times of files by one user. to the computer,
a process of accepting keywords from searchers;
A search program for executing a process of searching for a user associated with the keyword based on the total operation time of each file including the keyword in the file name among the operation times of each file by one user. a keyword reception unit that receives keywords from a searcher;
A search system for searching for a user associated with the keyword, based on a sum of participation times of events including the keyword in the event name, among participation times of one user in the event. The search unit is
Based on the ratio of the total participation time of the one user for each event containing the keyword in the event name to the total participation time of all users in the organization for each event containing the keyword in the event name, 2. The search system of claim 1, wherein searching for users associated with a keyword. Receiving keywords from searchers,
A search method for searching for a user associated with a keyword based on a sum of participation times of events including the keyword in the event name, among participation times of one user's event. to the computer,
a process of accepting keywords from searchers;
A search program for executing a process of searching for a user associated with the keyword based on the total participation time of each event including the keyword in the event name, among the participation times of one user's event.

Images