JP5057512B2 - File search system - Google Patents

Description

  The present invention relates to a file search system for searching a file related to work from a file group having no link relationship by a keyword or the like.

  In recent file systems in servers, both the number and type of files have increased explosively. For this reason, the file system is configured to provide a hierarchical structure of directories and manage a logical arrangement in which files are stored. However, as the number of files handled increases, it has become difficult to easily find the files that you want to use. As a solution to such a problem, a file search system for finding a file including a keyword is provided.

  The file search system performs a search for a file name including a keyword (including a directory path name) or a file having a character string including the keyword in the file. For example, Patent Document 1 discloses an invention in which link importance is defined between files having a link relationship searched for by keyword so that a file having a high link importance can be easily extracted. .

Further, Non-Patent Document 1 describes a memory recall support tool that stores an image so that a scene in which a user is browsing a Web page can be reproduced, and that the user sees a playback scene and recalls a past memory. Are listed. As a search function in this memory recall support tool, it is assumed that a Web page that has been displayed during a period in which many Web pages including a keyword that the user remembered is actively displayed has a high degree of relevance with the keyword. It is disclosed to rank higher.
JP 2001-290843 A (paragraphs [0019] to [0033]) Tetsuyuki Morita, Tetsuo Hidaka, Akinori Tanaka, Yasuhisa Kato, “Memory-Retriever”, INTERACTION2007, March 15, 2007

  However, Patent Literature 1 and Non-Patent Literature 1 are based on a search using keywords. Therefore, it is difficult to detect a file name or an image file or a data file that does not include the keyword in the file even if the file includes content related to the keyword.

  In view of the above-described problems of the prior art, it is an object of the present invention to provide a file search system for searching for a file related to work from a file group having no link relationship by a keyword or the like.

In order to solve the above problems, a file search system is a file search system for searching for a file related to work from a file group stored in a server, and derives a degree of association between files from a history of accessing the file. Relevance degree deriving means, and output means for outputting a group of files having a relevance degree equal to or greater than a predetermined degree to the search file searched by the keyword, wherein the relevance degree deriving means refers to the history. And calculating the activity time interval during which the work was performed, calculating the file usage time using the file defined within the range of the activity time interval, and determining the relevance based on the duplication of the file usage time. calculated, the activity time interval, reads the file name and access time recorded in the history, the accession within a predetermined time width If there is a record of the file time, it is determined that the work was performed within the predetermined time width, and the close corresponding to the opening of the file recorded in the history is determined according to a predetermined condition, The file usage time is calculated as a time in which the activity time interval overlaps from the time when the file determined according to the predetermined condition is opened to the time when the file is closed .

According to such a configuration, even for a file that does not have a link relationship by a keyword or the like, the activity time interval (time zone in which the user is working) is estimated using the history of accessing the server, and overlaps with the activity time interval. Thus, it is possible to determine that the files that have been used are related to each other. In addition, it is not necessary to determine the working time (activity time interval) using a separate special application, and the activity time interval can be easily calculated from the access time record. Further, by examining the overlap with the activity time interval with respect to the time when the file is opened, it is possible to accurately estimate the time during which the file was used based on the actual work.

Further, in the file search system, the relevance degree deriving means is duplicated when the file usage time overlaps for one file selected from all the files in the file group and another file. If there are multiple co-occurrence times, the number is the number of co-occurrence times, and the co-occurrence interval from the end of the previous co-occurrence time to the start of the next co-occurrence time And when the difference between the start time of the file use time of the one file or the start time of the file use time of the other file and the start time of the co-occurrence time is used as the use start pattern, the co-occurrence time and the Performing a calculation based on at least one information of the number of co-occurrence, the co-occurrence interval, and the use start pattern, and calculating a degree of association between the selected one file and another file; And features.
And, in the calculation, when only the co-occurrence time is information, the accumulation of the co-occurrence time is the bottom, and when only the co-occurrence number is information, the co-occurrence number is the bottom, When only the co-occurrence interval is information, the reciprocal of the accumulation of the co-occurrence intervals is the bottom, and when only the use start pattern is information, the accumulation of the use start pattern is the bottom. The relevance calculation formula in which each predetermined value becomes an index is used, and further, a relevance calculation formula that combines two or more formulas by combining the relevance calculation formulas is used.

  According to this configuration, the relevance is calculated using a relevance calculation formula that uses at least one of the co-occurrence time, the number of times of co-occurrence, the co-occurrence interval, and the start time pattern as information. The files for which the degree of association is calculated may not have a link relationship with a keyword or the like.

In the file search system, the predetermined condition is:
If the closing history corresponding to the opening is missing, the first processing to be deleted from the history, the file that has been kept open, the next from the end of the previous activity time interval A second process for determining a time overlapping with the previous activity time interval as a closing interval corresponding to the opening of the file , when the time until the start of the activity time interval is a predetermined time or more away, For file types in which the interval between the time when the file was opened and the time when the file was closed is less than or equal to a predetermined value, the file from the time when it was first opened until the time when it was last closed in the activity time interval It is at least one of the 3rd processes determined to be close corresponding to the open of .

According to such a configuration, the first process is a process that allows the degree of relevance to be calculated when a history of access to a file (access log) lacks a closing history. Using such a history, the file search system recognizes that the file is still open and analyzes the relevance. Therefore, as a preprocessing, it is possible to reduce disturbances in the calculation of the relevance by deleting the open so that the correspondence between the open and the close can be taken.
The second process is a process that copes with, for example, a time when the file is kept open and is not used, although open and closed are taken in the history. In other words, if the file is open and there is a predetermined time or more between the end of the previous activity time period and the start of the next activity time period, the file is not used for work. Assuming you are not using the file. And, when the activity time interval before the activity time interval is more than the predetermined time and the time when the file is open is overlapped, it is judged as the file usage time, and the disturbance in the calculation of relevance is reduced It becomes possible to do.
The third process is a process for dealing with a case where open and close are recorded below a predetermined value, such as a file that closes (releases lock) as soon as it is opened. In this case, it is considered that the file has been used from the time when it was first opened until the time when it was last closed within the activity time interval. This makes it possible to estimate the usage time of a file that is closer to the actual situation.

  According to the present invention, it is possible to search for a file related to work from a file group having no link relationship by a keyword or the like.

  Next, the best mode for carrying out the present invention (hereinafter referred to as “embodiment”) will be described in detail with reference to the drawings as appropriate.

(Embodiment)
A configuration of a file search system 1 according to an embodiment of the present invention will be described with reference to FIG. FIG. 1 is a diagram showing the configuration of the file search system 1.
The file search system 1 is configured such that the servers 11 and 12, the terminals 21 and 22, and the file association degree management device 100 are connected via a network 30 and can communicate with each other.

The servers 11 and 12 accumulate a large amount of files and record a history (access log) regarding who accessed which file. Most of the current file systems provide a directory hierarchical structure, and a history is managed by attaching a directory path name to each file.
Terminals 21 and 22 are used by users who perform tasks such as document creation and file search. The terminal type may be a PC (Personal Computer) or a dedicated terminal connected to the server.
The file association degree management apparatus 100 acquires histories from the servers 11 and 12 and calculates the degree of association of files that are used simultaneously by the user. Then, when a keyword search is executed by the terminals 21 and 22, the file information (file name) including the detected keyword is acquired, the relevance of the acquired file name is referred to by referring to the relevance of the file And the file names having a high degree of relevance are displayed on the terminals 21 and 22.
The number of servers 11 and 12 and terminals 21 and 22 may be one or three or more.

Next, the hardware configuration of the file relevance management apparatus 100 will be described with reference to FIG. FIG. 2 is a diagram illustrating a hardware configuration of the file association degree management apparatus 100.
As shown in FIG. 2, the file relevance management apparatus 100 is configured such that the processing unit 110, the input / output unit 120, the storage unit 130, and the communication control unit 140 are connected via a bus and can communicate with each other. .

  The processing unit 110 includes a CPU (Central Processing Unit) 111 that executes arithmetic processing, and a main memory 112 that is a storage unit that the CPU 111 uses for arithmetic processing. The main memory 112 is realized by a RAM (Random Access Memory) or the like. Then, the application program stored in the storage unit 130 is expanded in the main memory 112, and the CPU 111 implements various processes by executing it.

  The input / output unit 120 includes an input device (not shown) such as a keyboard and a mouse connected to the file association degree management device 100, and a display device such as a display for displaying various data such as the results of arithmetic processing performed by the processing unit 110. (Not shown).

  The storage unit 130 stores various data and calculation results used by the CPU 111 for calculation processing, or data transmitted and received by the input / output unit 120. The storage unit 130 is realized by a ROM (Read Only Memory) or a hard disk device (not shown).

  The communication control unit 140 includes a communication interface (not shown), transmits information processed by the processing unit 110 to other devices via the network 30 (see FIG. 1), and receives information from other devices. Control to receive.

  Next, functions of the file relevance management apparatus 100 will be described with reference to FIG. 3 (see FIG. 1 as appropriate). FIG. 3 is a diagram illustrating functions of the file association degree management apparatus 100.

  The processing unit 110 of the file relevance management apparatus 100 includes an access log collection unit 113, an access log analysis unit 114, an access log preprocessing unit 115, an activity time calculation unit 116, a file usage time calculation unit 117, and a relevance level calculation unit 118. Prepare.

The access log collection unit 113 acquires access logs (history) from the servers 11 and 12. The timing for acquiring the history may be regular, when necessary, or at any time.
The access log analysis unit 114 classifies the acquired access log (history) for each file name, or extracts an extension that immediately releases the lock, as will be described later.
The access log preprocessing unit 115 supplements or corrects the raw data of the access log (history) according to a predetermined condition before calculating the degree of association. The predetermined condition is, for example, deleting a history lacking close to open, or determining a file usage time for a file that has been open for a long time.
The activity time calculation unit 116 calculates a time interval (activity time interval) in which the user is working. In other words, it is not considered that the file is being used even if the file is open except during the activity time period.
The file usage time calculation unit 117 calculates a time (file usage time) in which the activity time interval calculated by the activity time calculation unit 116 overlaps with the time when the file is opened. In this process, the history corrected by the access log analysis unit 114 and the access log preprocessing unit 115 is used. Then, the file usage time is calculated for each file.
The relevance calculation unit 118 compares the file usage time calculated by the file usage time calculation unit 117 for each file name, and calculates the time (co-occurrence time) in which the file usage time overlaps. Then, the degree of association is calculated by a predetermined mathematical expression using the co-occurrence time and the number of times of co-occurrence as variables. This predetermined formula will be described later. The calculated degree of association is stored in the degree of association DB (Data Base) 131 for each file name.

  The relevance deriving means described in the claims includes the access log collection unit 113, the access log analysis unit 114, the access log preprocessing unit 115, the activity time calculation unit 116, the file usage time calculation unit 117, and the relevance level calculation unit 118. It is a general term for all functions.

The input / output unit 120 includes an operation input unit 121 for performing operations such as starting the file association degree management apparatus 100 and a display unit 122 for displaying various data.
The storage unit 130 stores the association degree calculated by the processing unit 110 in the association degree DB 131. When a file name is acquired from the communication control unit 140 or the operation input unit 121, the relevance level DB 131 is referred to, and a file group having a relevance level with respect to the acquired file name is extracted.
The communication control unit 140 receives the file name (file set F) as a result of the file search from the terminals 21 and 22, and transmits the extracted file name having the relevance to the terminals 21 and 22. Or

  Here, the history acquired by the access log collection unit 113 will be described with reference to FIG. 4 (see FIG. 3 as appropriate). FIG. 4 is a diagram illustrating an example of a history.

In FIG. 4, the history includes access time, access type, access ID, and file name as elements. Note that the history shown in FIG. 4 shows only elements necessary for calculating the degree of association.
The access time represents the time when the file was accessed. If number 1 is described, it indicates that access was made at 16: 51: 9 on May 2, 2007.
The access type represents open or close. That is, open indicates that the file has been opened (opened for use), and close indicates that the file has been closed (closed to stop using it).
The access ID is a display for identifying who accessed or from which terminal. For example, the access ID is a login name to the servers 11 and 12 (see FIG. 1) or an IP address or name assigned to the terminal.
The file name is a name given to the file. When the servers 11 and 12 have a file system that provides a hierarchical structure of directories, the file names include directory path names.
Note that FIG. 4 shows a history when the access ID is A11, but it is possible to extract a specific access ID even if a plurality of access IDs are mixed.

  Next, the flow of the correction process for the history and the calculation of the relevance will be described with reference to FIG. 5 (see FIGS. 1 and 4 as appropriate). FIG. 5 is a diagram showing the flow of the correction process for the history and the relevance calculation.

In the following description, it is assumed that the acquired history is data as shown in FIG.
First, the file relevance management apparatus 100 acquires access logs (history) from the servers 11 and 12 (step S201).
Next, for the specific access ID (A11 in FIG. 4), the working time, that is, the activity time (activity time interval) is calculated (step S202).

  Here, the calculation of the activity time in step 202 will be described in more detail with reference to FIG. 6A is a diagram showing the concept of the activity time calculation process, FIG. 6B is a diagram showing the flow of the activity time calculation process, and FIG. 6C is the result of the activity time calculation process. It is a figure which shows an example.

  As for the activity time, if there is a history of access time within the range of the predetermined time width, it is determined that the section of the predetermined time width is regarded as work. For example, when the predetermined time width is set to 30 minutes, the activity time interval (activity time interval) is 16:00 as shown in FIG. From 30 minutes to 17:30 and from 18:30 to 19:00.

  FIG. 6B shows the flow of processing for calculating the activity time for one predetermined time width. First, an access log (history) of an access ID to be processed is acquired from the access logs acquired in step S201 (step S301). Then, it is determined whether or not an access log (history) exists within a predetermined time width set in advance (step S302). That is, it is determined whether or not the accessed time is within a predetermined time range. If an access log exists in a predetermined time width (Yes in step S302), the time zone having the predetermined time width is regarded as an activity time, and the activity flag is set to 1 (step S303). On the other hand, if there is no access log in the predetermined time width (No in step S302), the processing is terminated for the predetermined time width. Thereafter, for the next predetermined time width, the process returns to step S302 to execute the process.

FIG. 6C shows an example of the processing result. A time zone in which the activity flag is 1 represents an activity time section. The processing result is stored in the storage unit 130 (see FIG. 3). Then, the processing result of step S202 is delivered to steps S206 and S207.
The activity flag may not be 1 and may be another code as long as it is an identifiable code.

  Returning to FIG. 5, in step S203, the access logs (history) are classified for each file name from the access logs acquired in step S201. This process is necessary for calculating the file usage time for each file name. Then, the processing result of step S203 is delivered to steps S205, S206, and S207.

  Next, in step S204, statistical processing is performed on the access log acquired in step S201, and an extension that immediately releases the lock is extracted. “Immediately release the lock” refers to a case where the access logs (history) related to open and close are recorded at almost the same time. This is because opening a file reads it into memory and releases the lock. Then, even if the file is actually used, since it is opened for almost 0 seconds in the history, it is recognized as if it was not used. To prevent this, for each extension (for example, doc, txt, etc.) representing the type of file name, the average time is calculated by summing up the time interval (time interval from open to close) when the file is open, If the average value is less than or equal to a predetermined value, the extension is corrected. Then, the extension subject to correction is delivered to step S207.

  Next, preprocessing A (step S205) will be described. The input to the preprocessing A is the output in step S203, and is processing when the close corresponding to open is missing. In the history of FIG. 4, number 3 corresponds to this case. The lack of close can be found by checking whether there is a close corresponding to the opened file name. Then, such a history related to open is deleted, and correction is made so that the correspondence between open and close can be taken in the entire history.

The preprocess B (step S206) is a process for obtaining the file usage time corresponding to the case where the file is left open even though it is not actually used. The input to this preprocessing B is the output of steps S202 and S203. The difference from preprocessing A is that there is a close corresponding to the opened file name.
In the history of FIG. 4, numbers 4 and 13 correspond to this case. It opened at 16:54:15 on May 2, 2007 and closed at 8:37:56 on May 3, 2007. Moreover, the interval between the numbers 12 and 13 is 14 hours 6 minutes 34 seconds, and there is no access between them.
Thus, when there was no access for a long time, it was determined that the file was not used.

  This preprocessing B will be described in more detail with reference to FIG. 7A is a diagram showing a concept of processing in the preprocessing B, FIG. 7B is a diagram showing a flow of processing in the preprocessing B, and FIG. 7C is an example of processing results in the preprocessing B. FIG.

First, in the preprocessing B, in order to calculate the file usage time, (1) the time when the file is open and the activity time (activity time interval) overlap, and (2) If the time that is not activity time (activity time interval) is equal to or longer than the predetermined time, use the history up to the activity time interval before the activity time interval,
Was set as a condition.
As described above, in the history of FIG. 4, the file names DDD.doc with numbers 4 and 13 are targeted.
In FIG. 7A, as shown by the horizontal bar 400 with dots, it is determined that the section that overlaps the activity time (activity time section) is the state in which the file is used, that is, the file use time. The Along with that, history of open and close is added.

  FIG. 7B shows a process flow of the pre-process B. First, an access log (history) of an access ID to be processed is acquired (step S401), and it is determined whether or not a time interval that is not an activity time is smaller than a predetermined time (step S402). If the time period that is not the activity time is smaller than the predetermined time (Yes in step S402), it is considered that the file is used for the activity time period, and the open or close of the access log (history) is corrected (step S403). . And it returns to step S402 and repeats a process. On the other hand, when the time interval that is not the activity time (from the end of the previous activity time interval to the start of the next activity time interval) is equal to or longer than the predetermined time (No in step S402), it corresponds to the access ID. The access log after that (after the time when the predetermined time or more has elapsed) is deleted (step S404), and the process ends.

  FIG. 7C shows an example of the processing result. Numbers 3-1, 3-2 and 3-3 are corrected access logs, which are corrected so as to overlap with the activity time. In addition, since there is a time period of no access for 14 hours or more between the numbers 12 and 13, the number 13 is deleted. However, in the embodiment of the present invention, the predetermined time is set to 5 hours, but is not limited thereto.

  Next, referring back to FIG. 5, the preprocessing C (step S207) will be described. The input to the preprocessing C is the output of step S202, step S203, and step S204.

  The preprocessing C will be described in more detail with reference to FIG. 8A is a diagram showing the concept of processing in the preprocessing C, FIG. 8B is a diagram showing the flow of processing in the preprocessing C, and FIG. 8C shows an example of processing results in the preprocessing C. FIG.

First, in the pre-processing C, the history related to the first open in the activity time section is targeted for the file with the file extension (html, tex, etc.) acquired from step S204 in order to calculate the file usage time. Then, the condition for deleting only the history related to the last close and deleting the other history is set.
For example, in the history of FIG. 4, the file name CCC.html of the numbers 5, 6, 8, 9, 11, and 12 is targeted.
In FIG. 8A, as indicated by the horizontal bar 500 with dots, in the activity time section, the section between the history related to the first open and the history related to the last close used the file. Determined to be in a state. Along with that, the history of open and close is deleted.

  FIG. 8B shows the flow of preprocessing C for one activity time interval. First, an access log (history) of an access ID to be processed is acquired (step S501), and it is determined whether the access is the last access in the activity time interval (step S502). If it is the last access (Yes in step S502), the first open and the last close of the activity time interval are left and the others are deleted (step S503). Similarly, steps S502 and S503 are repeatedly executed. On the other hand, if it is not the last access (No in step S502), the process returns to step S502 and continues.

  FIG. 8C shows an example of the processing result. That is, in the history of FIG. 4, numbers 6 and 8 that are not the first open and the last close are deleted.

Returning to FIG. 5, after the history is corrected by the preprocessing A, the preprocessing B, and the preprocessing C, the file usage time is totaled for each file name (step S208).
Here, the file usage time will be described with reference to FIG. 9A is an explanatory diagram showing an example of a file usage time table for a specific file name, FIG. 9B is an explanatory diagram showing an example of a file usage time table for another file name, and FIG. These are explanatory drawings showing an example of a co-occurrence time table in which both files are used in duplicate.
FIG. 9 shows the history acquired for the same access ID. Here, the file usage time table shown in FIGS. 9A and 9B is configured by associating the file usage start time (file start time) with the file usage time length. In addition, the co-occurrence time table shown in FIG. 9C is a file co-occurrence start time (start time of co-occurrence time) and a file co-occurrence time length that overlap two file use time tables and overlap the file use time. It is the result of having calculated.

Returning again to FIG. 5, the calculation of the relevance level (step S209) will be described next.
First, the file usage start time and the file usage time are obtained from the file usage time tables of the two file names whose relations are to be examined, and collated. As shown in FIG. Calculate the file co-occurrence start time and the file co-occurrence time length for the current time zone. Then, the file co-occurrence time accumulation T, the file co-occurrence time count K, the similarity P of the file use start pattern, and the co-occurrence interval degree D are obtained, and the relevance calculation formula R = T ^α K ^The degree of association R is calculated from ^β P ^γ D ^δ (0 ≦ α, β, γ, δ ≦ 1). Α, β, γ, and δ are indices for weighting T, K, P, and D, respectively.
The mathematical formula for the degree of association is not limited to the above formula, and may be any one of variables T, K, P, and D described later, or a combination thereof.

  Here, the cumulative T of file co-occurrence times, the number K of file co-occurrence times, the similarity P of the file use start pattern, and the interval D of co-occurrence will be described in detail with reference to FIG. FIG. 10 is a diagram illustrating variables of the relevance calculation formula.

In FIG. 10, the horizontal axis indicates time. It is assumed that file usage time table data for file X and file Y is plotted. A co-occurrence time table of these files X and Y is plotted at the top.
The cumulative T of file co-occurrence times (co-occurrence times) is T = Σt _i (where i = 1 to n). In FIG. 10, n is 4. Note that t _i is the file co-occurrence time length.
The number of file co-occurrence times (number of co-occurrence) K is K = n.
The similarity P of the file use start pattern is P = 1 / Σp _i (where _i = 1 to n), and when p _i is 0, P = 1. P _i is the difference between the file use start time and the co-occurrence start time.
The co-occurrence interval degree D is D = Σd _{i (i + 1)} (where i = 1 to n−1), and when n = 1, D = 1. Note that d _{i (i + 1)} is the co-occurrence interval.

The variables T, K, P, and D are expressions that can express the following idea. That is,
(1) The longer the accumulation of co-occurrence times, the greater the degree of association.
(2) The degree of association increases as the number of co-occurrence increases.
(3) The degree of relevance increases as the file use start pattern is similar.
(4) The degree of relevance increases as the co-occurrence interval increases.
And (1)-(3) are matched with the user's intuition based on the actual use. In addition, (4) seems to be more closely related to the fact that co-occurrence occurs again even if there is a long interval between the previous co-occurrence and the next co-occurrence.

Then, as shown in FIG. 11, the relevance level R is stored in the relevance level DB 131 (see FIG. 3) of the storage unit 130 as a database indicating the relevance level between two files.
The relevance DB 131 may be updated by periodically adding a new history and performing the processing shown in FIG. 5 (steps S201 to S210). Also, when necessary, it may be updated at any time.

  Next, the flow of relevance calculation processing and search processing in the file relevance management apparatus 100 will be described with reference to FIG. FIG. 12 is a diagram illustrating the flow of the relevance calculation process and the search process in the file relevance management apparatus 100.

First, in the relevance calculation processing, the file relevance management device 100 transmits an access log request in order to obtain an access log (history) from the servers 11 and 12 (step S601). Upon receiving the access log request, the servers 11 and 12 transmit the access log (step S602). The file association degree management apparatus 100 acquires an access log (step 603), and calculates, stores, and updates the association degree (step S604).
Steps S601 to S604 may be performed periodically, or may be performed as necessary. When the functions of the file relevance management apparatus 100 are provided in the servers 11 and 12, the processing in steps S601 and S602 is unnecessary, and the process starts directly from acquiring the access log (step S603). Is done.

Next, in the search process, the terminals 21 and 22 instruct execution of keyword search (step S605), and a file including the keyword (file set F) is acquired (step S606). Then, the terminals 21 and 22 transmit the file set F including the keyword to the file association degree management apparatus 100 (step S607). On the other hand, the file relevance management apparatus 100 receives information on the file set F, refers to the relevance DB 131 (see FIG. 3) using the file name included in the file set F, and includes the files included in the file set F. The degree of association of the file name related to the name is extracted (step S608). Then, the file association degree management apparatus 100 transmits a file name having a high degree of association (large) to the terminals 21 and 22 (step S609). Then, the terminals 21 and 22 display the transmitted file names having a high degree of relevance (step S610).
Note that the servers 11 and 12 may perform steps S605 to S607.

  Note that the output means described in the claims is one in which the processing unit 110 (see FIG. 2) of the file association degree management apparatus 100 executes steps S608 to S609.

In step S608, it is also effective to shorten the search time by executing a search only for file names that are equal to or higher than a threshold set in advance in the degree of relevance.
For this purpose, the threshold value _Rth is calculated as follows.
First, let H be the set of files for the learning set. For each file h (where hεH), information is collected from subjects who participated in the experiment to collect learning data as to whether or not it was determined to be related to the co-occurrence time.
Here, a set of files co-occurring with the file h is defined as Q (h). Next, each element q _{i of} the file set Q (h) is associated as a correct answer when the subject determines that it is related and as an incorrect answer when it is determined that the subject is not related.
Next, among the elements of Q (h), an element q _k that is determined to be correct and has the smallest relevance to h is obtained, and a threshold value R (h, q _k ) for h is obtained. Then, a threshold value R (h, q _k ) is calculated for all h included in the file set H, and an average value thereof is set as R _th .
That is,
R _th = _Σh∈HR (h, q _k ) / | H |
To calculate the threshold value R _th . However, | H | is the number of files included in the file set H.

(Example of evaluation experiment)
Next, an evaluation experiment performed in advance with respect to the embodiment of the present invention will be described.

(Experiment environment)
The access log was collected using Samba, a Windows (registered trademark) compatible file server that is widely used at present. In the evaluation experiment, Samba was used because the usage time of the file can be extracted without installing a special application in the user's environment. However, even with a file server other than Samba, opening / closing of the file is possible. If information can be acquired, the proposed method can be applied in the same way.
In the evaluation experiment, the file server Samba2.2.3a was started at log level 2, and two subjects used the file server from the terminal for about 4 months. During this period, subjects prepared for academic conference presentations, and access logs for text files, image files, experimental data files, etc. were collected. The extensions of the analysis target are bib, doc, gif, htm, html, jpg, mpg, mpeg, pdf, ppt, tex, txt, and xls so that access to system files etc. is ignored. In addition, the activity time width is 30 minutes, the predetermined value relating to the non-activity time in the preprocessing B is 5 hours, and the predetermined value relating to the average value of the open times in step S204 is 10 seconds.

(Evaluation index)
As the evaluation index of the evaluation result, the recall rate (Recall) and the precision rate (Precision), which are generally used in the field of information retrieval systems, were used.
The recall is an index of completeness indicating how many matching files can be searched for a matching file (correct answer file) in the file group to be searched. The relevance ratio is an accuracy index indicating how many suitable files are included in the file group obtained as a search result.
Reproducibility = number of matched files searched / number of correct files in all target documents precision = number of matched files searched / number of files searched

(Evaluation results)
First, a file usage time table was created from each subject's access log. Next, the co-occurrence time was calculated for a file randomly selected from all files. The subject evaluated the co-occurrence time as a correct answer when judged to be related to the work, and an incorrect answer when judged to be unrelated to the work.
Table 1 shows the number of files evaluated and the total number of files.

Next, half of the files were randomly selected from the set of evaluated files shown in Table 1 to obtain a learning set. The relevance R was calculated by R = T ^α K ^β P ^γ D ^δ . At this time, the index (α, β, γ, δ) of the relevance R is a value when the average value of the recall ratios of all the files included in the learning set is maximized. The relevance threshold G _th was also calculated. The results are shown in Table 2.

  Using the indices shown in Table 2, the recall and precision were determined for each file in the evaluation set. The results are shown in Table 3.

  The results shown in Table 3 are considered to be practically acceptable. Then, it becomes possible to search for a file related to the work from a file group having no link relationship by a keyword or the like.

It is a figure which shows the structure of a file search system. It is a figure which shows the hardware constitutions of a file related degree management apparatus. It is a figure which shows the function of a file related degree management apparatus. It is a figure which shows an example of a log | history. It is a figure which shows the flow of the correction process with respect to a log | history, and relevance calculation. (A) is a diagram showing the concept of processing for calculating activity time, (b) is a diagram showing the flow of processing for calculating activity time, and (c) is an example of processing results for calculating activity time. FIG. (A) is a figure which shows the concept of the process in the pre-process B, (b) is a figure which shows the flow of the process in the pre-process B, (c) is a figure which shows an example of the process result in the pre-process B. . (A) is a figure which shows the concept of the process in the pre-process C, (b) is a figure which shows the flow of the process in the pre-process C, (c) is a figure which shows an example of the process result in the pre-process C. . (A) is explanatory drawing which shows an example of the file usage time table about a specific file name, (b) is explanatory drawing which shows an example of the file usage time table about another file name, (c) is both It is explanatory drawing which shows an example of the co-occurrence time table in which these files are used repeatedly. It is a figure shown about the variable of a relevance calculation formula. It is a figure which shows the association degree between files. It is a figure which shows the flow of the calculation process and search process of a relevance degree in a file relevance degree management apparatus.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 File search system 11,12 Server 21,22 Terminal 30 Network 100 File relevance management apparatus 110 Processing part 113 Access log collection part 114 Access log analysis part 115 Access log pre-processing part 116 Activity time calculation part 117 File use time calculation part 118 relevance calculation unit 120 input / output unit 130 storage unit 131 relevance DB
140 Communication control unit

Claims (8)

A file search system for searching for files related to work from a group of files stored in a server,
Relevance deriving means for deriving the relevance between files from the history of accessing the file;
An output means for outputting a file group having a degree of relevance greater than or equal to a predetermined degree with respect to the search file searched by the keyword,
The relevance deriving means refers to the history, calculates an activity time section in which work has been performed, calculates a file use time using a file defined within the range of the activity time section, Calculate the relevance based on the overlap of file usage time ,
Read the access time and file name recorded in the history for the activity time interval, and if the access time is recorded within a predetermined time width, it is determined that the work was performed within the predetermined time width. Calculated,
Confirming the closing corresponding to the opening of the file recorded in the history according to a predetermined condition;
The file search system characterized in that the file usage time is calculated as a time when the activity time interval overlaps from a time when a file determined according to the predetermined condition is opened to a time when the file is closed . The relevance deriving means includes:
For one file selected from all files in the file group stored in the server and other files,
If the file usage time overlaps, the overlapping time is the co-occurrence time,
If there are multiple co-occurrence times, the number is the number of co-occurrence,
The co-occurrence interval is from the end of the previous co-occurrence time to the start of the next co-occurrence time,
When the difference between the start time of the file use time of the one file or the start time of the file use time of the other file and the start time of the co-occurrence time is used as the use start pattern,
Performing a calculation based on at least one piece of information of the co-occurrence time, the number of times of co-occurrence, the co-occurrence interval, and the use start pattern, and calculating a degree of association between the selected one file and another file. ,
The file search system according to claim 1 . The operation is a file search system according to claim 2 in which the accumulation of pre-Symbol co-occurrence time becomes the bottom, characterized by using a relevance calculation formula predetermined value is exponential. The operation is a file search system of claim 2, before Symbol co-occurrence count becomes the bottom, characterized by using a relevance calculation formula predetermined value is exponential. The operation is a file search system according to claim 2, reciprocal those accumulated before Symbol co-occurrence interval becomes the bottom, characterized by using a relevance calculation formula predetermined value is exponential. The operation is a file search system according to claim 2 in which the cumulative prior Symbol used starting pattern becomes the bottom, characterized by using a relevance calculation formula predetermined value is exponential. The calculation includes a relevance calculation formula in which the accumulation of the co-occurrence time is a base and a predetermined value is an exponent, a relevance calculation formula in which the number of co-occurrence is a base and a predetermined value is an exponent, and the co-occurrence interval In the combination of the relevance calculation formula in which the reciprocal number of the cumulative value of the cumulative value of the above and the predetermined value becomes the exponent, and the relevance calculation formula in which the accumulation of the use pattern becomes the base and the predetermined value becomes the exponent, 3. The file search system according to claim 2 , wherein a relevance calculation formula for multiplying the above formula is used. The predetermined condition is:
A first process of deleting from the history when a history of closing corresponding to the opening is missing;
For the file that remains open, if there is a predetermined time or more between the end of the previous activity time interval and the start of the next activity time interval, the previous activity time interval A second process for determining a time overlapping with a closing interval corresponding to the opening of the file;
For file types in which the interval between the time when the file was opened and the time when the file was closed is less than or equal to a predetermined value, the file from the time when it was first opened until the time when it was last closed in the activity time interval A third process for determining a close corresponding to the open of
Being at least one of
The file search system according to claim 1 .

Images