JPWO2007088576A1 - File search program, method and apparatus - Google Patents

Abstract

The target file is searched from the file storage unit registered in association with the attribute information extracted for each file. Attribute information is set as a search condition in the search condition setting section file. For each file in the file storage unit, the search unit searches for a file by calculating similarity using weights from the attribute information of the search condition and the registered attribute information. The search result display unit displays the search result of the file and generates a display state corresponding to the browsing operation for the search result. The browsing operation extraction unit extracts a browsing operation performed by the search result display unit. The weight calculation unit calculates a value indicating the search viewpoint, which is a feature of the file the user is looking for, from the browsing operation history as a weight used for similarity calculation, and sets the weight calculated when performing the search again. Calculate similarity.

Description

The present invention relates to a file search program, a method, and an apparatus for searching for a target file from files stored in a personal computer, a file server or the like, and in particular, a file search program for searching for a target file using attribute information of the file. , Methods and apparatus.

  In recent years, with the spread of computer systems and networks, a lot of electronic data is exchanged through networks and storage media and stored on computers. In such a situation, there is a problem that it is difficult to search for target information while a large amount of information can be acquired.

  As a method for searching for a target file from files stored in a personal computer, a file server, or the like, a full-text search for words or phrases stored in the file is generally used.

  When there are a large number of search targets or the target file is ambiguous, an AND or OR search is performed using a plurality of keywords, or keywords are added to the search results to narrow down the search results.

Further, the method of feeding back the determination result to the next search by determining whether the user himself / herself is compatible with the search result file, that is, determining whether the file is similar to or not similar to the target file. There is also.
International Publication No. W04 / 0319566 JP 2001-209660 A

  However, in such a conventional file search method, the general full-text search method has a problem that it is difficult to set a keyword when there are a large number of files to be searched or the target file is ambiguous. There is. If simple keywords are set in order to eliminate search omissions, the search results become large, and the burden of determination work by the user himself increases.

  Furthermore, even if the user determines the suitability for the search results, if there are many search results, the determination work for searching for similar or dissimilar files becomes a heavy burden.

  For this reason, under the present circumstances, trial and error are repeated, keywords are changed, and target files are selected over time from a large amount of search results, which is a heavy burden on the user.

It is an object of the present invention to provide a file search program, method, and apparatus that reduce a user's burden and make it easy to search for a target file.

(program)
The present invention provides a file search program. The file search program of the present invention is provided in a computer including a file storage unit registered in association with attribute information extracted for each file.
A search condition setting step for setting attribute information as a file search condition;
For each registered file, a search step for searching for a file by calculating similarity using weights from the attribute information of the search condition and the registered attribute information;
A search result display step for displaying a search result of the file by the search step and generating a display state according to a browsing operation on the search result;
Browsing operation extraction step for extracting browsing operation history by the search result display step,
A weight calculation step for calculating a weight to be used for calculating the similarity in the search step from the browsing operation history, and setting the search step when performing the search again;
Is executed.

  Here, the attribute information includes a plurality of attribute items, and the search step calculates the file similarity as a sum of the similarity of each attribute item and the weight of each attribute item.

The search result display step can display a preview of the file as a search result by browsing operation.
The browsing operation extraction step calculates the frequency of each attribute item of the previewed file or files and extracts it as a browsing operation history,
The weight calculation step calculates the weight for each attribute item according to the frequency of each attribute item in the browsing operation history.

  Specifically, the browsing operation extraction step counts the common attribute value included in each attribute item of the plurality of previewed files and extracts the maximum frequency of the attribute value for each attribute item, and the weight calculation step includes the attribute item The weight of each attribute item is calculated as a normalized value obtained by dividing the maximum frequency for each preview number.

The search result display step can rearrange and display the search result files in ascending or descending order according to the similarity of the selected attribute item for any one of the plurality of attribute items.
The browsing operation extraction step calculates a value corresponding to the rank of each attribute item selected for sorting and extracts it as a browsing operation history,
In the weight calculation step, the weight for each attribute item may be calculated according to the rank of each attribute item in the browsing operation history.

  Specifically, in the browsing operation extraction step, for the rank of each attribute item selected for sorting, a score that becomes a higher value in the order of the sorting operation is calculated and stored as a browsing operation history, and weight calculation is performed. The step calculates the weight of each attribute item as a normalized value obtained by dividing the score for each attribute item by the number of times of sorting.

Furthermore, the search result display step can display a preview of the file as a search result by a browsing operation,
The browsing operation extraction step calculates the frequency and browsing time of each attribute item of one or more previewed files and extracts it as a browsing operation history,
In the weight calculation step, the weight for each attribute item may be calculated according to the frequency and browsing time of each attribute item in the browsing operation history.

  Specifically, the browsing operation extraction step counts the common attribute value included in each attribute item of the plurality of previewed files and extracts the maximum frequency of the attribute value for each attribute item, and the weight calculation step includes the attribute item The weight of each attribute item is calculated as a normalized value obtained by dividing the maximum frequency for each time by the viewing time.

The search result display step can display a preview of the file as a search result by a browsing operation, and in ascending or descending order according to the similarity of the selected attribute item with respect to any one selection operation of a plurality of attribute items The search result files can be sorted and displayed in
The browsing operation extraction step
A first extraction step of calculating a frequency of each attribute item of the previewed file or files and extracting it as a browsing operation history;
A second extraction step of calculating a value according to the rank of each attribute item selected for sorting and extracting as a browsing operation history;
The browsing operation extraction step calculates the frequency and browsing time of each attribute item of the previewed one or more files and extracts it as a browsing operation history;
Comprising at least two of
The weight calculation step corresponds to the browsing operation extraction step,
A first weight calculating step for calculating a weight for each attribute item according to the frequency of each attribute item in the browsing operation history in the first extraction step;
The weight calculation step by the second extraction step includes a second weight calculation step for calculating a weight for each attribute item according to the rank of each attribute item in the browsing operation history;
A third weight calculating step of calculating a weight for each attribute item according to the frequency and browsing time of each attribute item of the browsing operation history by the third extracting step;
You may provide at least any two of these.

  In this case, when the weight calculation step combines at least two of the first weight calculation step to the third weight calculation step, for example, a search is performed by calculating an average of weights for each attribute item calculated for each. Set to step.

  In the search condition setting step, attribute values of a plurality of attribute items input using the search work screen are set as search conditions.

  In the search condition setting step, the attribute value of the attribute item is extracted from the file specified using the search work screen, and the search condition is set.

  The attribute information includes, as attribute items, type, owner, date of creation or update, and word / phrase.

(Method)
The present invention provides a file management method. The present invention provides a file management method for searching for a target file from a file storage unit registered in association with attribute information extracted for each file.
A search condition setting step for setting attribute information as a file search condition;
For each file in the file storage unit, a search step of searching for a file by calculating similarity using weights from the attribute information of the search condition and the registered attribute information;
A search result display step for displaying a search result of the file by the search step and generating a display state according to a browsing operation on the search result;
Browsing operation extraction step for extracting browsing operation history by the search result display step,
A weight calculation step for calculating a weight to be used for calculating the similarity in the search step from the browsing operation history, and setting the search step when performing the search again;
It is provided with.

(apparatus)
The present invention provides a file management apparatus. The file management apparatus of the present invention
A file storage unit registered in association with attribute information extracted for each file;
A search condition setting section for setting attribute information as a file search condition;
For each registered file, a search unit that calculates a similarity using a weight from the attribute information of the search condition and the registered attribute information, and searches for the file;
A search result display unit that displays a search result of the file by the search unit and generates a display state according to a browsing operation on the search result;
A browsing operation extraction unit for extracting a browsing operation history by the search result display unit;
Calculating a weight to be used for calculating the similarity of the search unit from a browsing operation history, and setting a weight to the search unit when performing a search again;
It is provided with.

  According to the present invention, file attribute information (including attribute items such as update date, owner, and file type) is used to calculate the similarity of search files. Further, the attribute information focused on by the user is extracted as a search viewpoint from the browsing operation history of the user on the search result display screen, and the search reflecting the user's intention is realized by feeding back to the next search.

As a method to extract the user's search viewpoint, provided on the search result screen,
(1) Preview screen,
(2) Sorting function,
Can be used.

  The preview screen is a function for displaying a summary or image of an arbitrary search result from the search results displayed in a list. The preview viewed by the user indicates that the file is focused on by the user. If multiple files are previewed, the common items of each file, such as words, file type, owner, update date and time, etc. are detected and reflected in the next search. Can be searched for.

  In addition, when combined with the sorting function, if the preview is intensively executed for attribute items with higher display priority, for example, specific file types, the search viewpoint of the user can be more reliably extracted. Can do.

  In the present invention, it is assumed that the search is performed recursively, and the search viewpoint can be extracted in the same manner between the first and second searches.

As a result, since the feature of the file that the user is looking for is automatically extracted from the operation of browsing the search result by the user, there is less burden on the user and high-accuracy search that reflects the user's search viewpoint is performed. Is possible.

Block diagram of a hardware environment in which the file search program of the present invention is executed Block diagram of another hardware configuration in which the file search program of the present invention is executed Block diagram of another hardware configuration in which the file search program of the present invention is executed The block diagram of the function structure of the file search device by this invention The block diagram which showed the detailed functional structure of the file registration process part of FIG. Explanatory drawing of the attribute information storage part of FIG. The block diagram which showed the detailed functional structure of the file search process part of FIG. Explanatory drawing of the search screen used in this embodiment Explanatory drawing of the concrete processing content of the file search processing part of FIG. Explanatory drawing of the attribute information file corresponding to the search result of FIG. Explanatory drawing of the weight table containing the operation history produced | generated based on the preview of the search result of FIG. Flow chart of search processing for calculating weights used in next search using preview as browsing operation history Explanatory drawing of the attribute information file corresponding to the search results sorted by the similarity of the attribute item "type" Explanatory drawing of the weight table containing the operation history produced | generated based on rearrangement of the search result of FIG. Explanatory drawing of the rearrangement | ranking order score table which defined the score set to an attribute item by the rearrangement order | rank of FIG. Flow chart of search processing for calculating weights to be used in the next search using sorting as browsing operation history Explanatory drawing of the weight table containing the operation history produced | generated based on the browsing time of the preview of the search result of FIG. Flow chart of search processing for calculating weight used in next search using preview time as browsing operation history Flow chart of search processing for calculating weights used in the next search using the number of previews, the number of rearrangements, and the preview time as the browsing operation history.

  FIG. 1 is a block diagram of a hardware configuration for executing a file search program of the present invention. In FIG. 1, in the present embodiment, the file search device 10 includes a CPU 12, a memory 14, an input / output unit 16, and a storage device 18. A file search program 20 and a file 22 are stored in the storage device 18. .

  The file search program 20 is executed by the CPU 12 and searches for a target file from the file 22. The input / output unit 16 includes devices such as a keyboard, a mouse, and a display, and the storage device 18 is, for example, a hard disk drive.

  FIG. 2 is a block diagram of another hardware configuration in which the file search program of the present invention is executed. In the embodiment of FIG. 2, the file search apparatus 10 is connected to the file management apparatus 24 via the network 11. The file search device 10 includes a CPU 12, a memory 14, an input / output unit 16, and a storage device 18, and a file search program 20 is stored in the storage device 18.

  On the other hand, the file management device 24 includes a CPU 26, a memory 28, an input / output unit 30, and a storage device 32, and the file 22 is stored in the storage device 32. Therefore, the file search device 10 executes the file search program 20 in the storage device 18 by the CPU 12 and searches for the target file from the file 22 in the storage device 32 of the file management device 24.

  FIG. 3 is a block diagram of another hardware configuration in which the file search program of the present invention is executed. Similar to the embodiment of FIG. 2, the file search apparatus 10 is connected to the file management apparatus 24 via the network 11. However, the file search program 20 and the file 22 are stored in the storage device 32 of the file management device 24.

  Therefore, the file search device 10 reads the file search program 20 from the storage device 32 of the file management device 24 and executes it by the CPU 26 to search for the target file from the file 22 stored in the storage device 32 of the file management device 24. To do.

  FIG. 4 is a block diagram of a functional configuration of the file search apparatus 10 according to the present invention. In FIG. 4, the file search apparatus 10 includes a file registration processing unit 34, a file search processing unit 36, a file storage unit 38, and an attribute information storage unit 40.

  FIG. 5 shows a functional configuration of the file registration processing unit 34 of FIG. 4, and the file registration processing unit 34 is provided with a file processing detection unit 42, a file registration unit 44, and an attribute information extraction unit 46.

  The file search processing unit 36 in FIG. 4 detects a file storage process from a file write event in the file system, for example, and notifies the file registration unit 44 and the attribute information extraction unit 46 in FIG. 5 of the detection of the storage process. The file registration unit 44 stores the actual state of the file in the file storage unit 38 arranged in a storage device such as a hard disk using the file writing function of the file system.

  The attribute information extraction unit 46 extracts attribute information associated with the file, for example, attribute information such as a file name, a file type, an owner, a storage location, a date of creation or update, and a document phrase.

  In the item of attribute information extracted by the attribute information extraction unit 46, information other than file information can be obtained from a general file system. The attribute items in the file can be read from the actual state of the file. The attribute items constituting the attribute information listed here are examples, and all information that can be acquired from the file is included in the attribute information used in the search processing of the present invention.

  The attribute information extracted by the attribute information extraction unit 46 is stored in the attribute information storage unit 40 in association with the file.

  FIG. 6 is an explanatory diagram of the attribute information storage unit 40 of FIG. In FIG. 6, in this example, four attribute items “attribute a”, “attribute b”, “attribute c”, and “attribute d” are defined corresponding to the “file ID” to be associated with the file. = “A1, b1, c1,... D4” are stored as “attribute a” to “attribute d” corresponding to # 1 to # 4, which are attribute values corresponding to the respective attributes a to d Represents.

  FIG. 7 is a block diagram showing a detailed functional configuration of the file search processing unit 36 of FIG. 7, the file search processing unit 36 includes a search condition setting unit 48, a search unit 50, a search result output unit 54, a browsing operation extraction unit 56, an operation history information storage unit 58, a weight calculation unit 60, and a weight storage unit 62. ing.

  The search condition setting unit 48 sets attribute information as a search condition for a target file. As attribute information set as a search condition, an attribute value corresponding to the attribute item of the attribute information storage unit 40 of FIG. 6 created at the time of file registration by the file registration processing unit 34 of FIG. 5 is set.

  Setting of attribute information as a search condition by the search condition setting unit 48 is specified by a method in which a user directly inputs a value of an attribute item by searching a search operation screen, and a file close to a target file is specified. One of the methods for automatically extracting and setting attribute information as a search condition from a file can be selected.

  The search unit 50 targets the files stored in the file storage unit 38 by using weights from the attribute information of the search conditions and the attribute information registered corresponding to the files of the attribute information storage unit 40 for each file. Similarity) is calculated, and a file having a high similarity is searched.

  The similarity calculation performed by the similarity calculation unit 52 provided in the search unit 50 is calculated as a sum obtained by multiplying the similarity of each attribute item set as a search condition by the weight of each attribute item. The search result output unit 54 displays the search result of the file by the search unit 50 and sets a display state according to a browsing operation such as preview or rearrangement for the search result.

  The browsing operation extraction unit 56 extracts browsing operation history such as preview and rearrangement in the search result output unit 54 and stores it in the operation history information storage unit 58. Further, the weight calculation unit 60 calculates weights used for calculating the similarity by the similarity calculation unit 52 provided in the search unit 50 from the browsing operation history stored in the operation history information storage unit 58, and performs the search again. Set for similarity calculation of the search unit 50.

  FIG. 8 is an explanatory diagram of a search screen used in this embodiment. In FIG. 8, a search condition setting operation unit 66, a rearrangement operation unit 68, and a search result display unit 72 are arranged on the search screen 64 from the upper side.

  The search condition setting operation unit 66 is provided with a type input frame 74, an owner input frame 76, a date and time input frame 78, and a word / phrase input frame 80 in order to set attribute items as search conditions used for searching for a target file. It has been. Further, a path input frame 82 for designating a search condition file for automatically extracting and setting attribute items as search conditions and a reference button 84 referred to from the search target path are provided.

  Further, on the right side of the search condition setting operation unit 66, there are a search button 86 for starting the search in a state where the operation items and the search conditions by specifying the path have been input, and the operation item or path that has already been input in the re-search. An initialization button 88 for clearing to the initial state is provided.

  In this example, the search result display section 72 displays the five search file information 98-1 to 98-5 in an array in descending order of similarity calculated by the similarity calculation at the time of search.

  In the search file information 98-1 to 98-5, taking the search file information 98-1 as an example, “file # 1” is displayed as the file ID and corresponds to the attribute items of type, owner, date / time, and phrase "Abc", "2005/11/01 17:25", "User U1", and "Summary of file # 1" are displayed as the attribute values.

  Further, preview buttons 100-1 to 100-5 are provided on the right side of the search file information 98-1 to 98-5, and by operating any of the preview buttons 100-1 to 100-5, The preview can be superimposed and displayed as a previous screen for the search screen 64.

  Therefore, when the operator wants to confirm the contents while looking at the search file information 98-1 to 98-5 in the search result display section 72, the preview can be previewed as required by the preview buttons 100-1 to 100-5. It is possible to browse and determine whether the target file has been searched.

  Further, a sorting operation unit 68 is provided on the search result display unit 72, and in this case, sorting is executed according to the priority of the attribute item, so that a type sorting button 90 and an owner sorting button 92 are provided. A date / time sorting button 94 and a phrase sorting button 96 are arranged.

  When any sort button of the sort operation unit 68 is selected, the search file information 98-1 to 98-5 according to the similarity of the selected attribute item is sorted, and the sort is completed. Thus, it is possible to appropriately determine whether or not the target file can be searched by browsing the preview while paying attention to the file having the higher similarity.

  In the search result display unit 72 of FIG. 8, five search results are displayed. However, the search result display unit 72 can display more search results by scroll operation or screen switching operation.

  FIG. 9 is an explanatory diagram of specific processing contents of the file search processing unit 36 of FIG. 9, in this example, the attribute value “a2” or “a3” is designated for the attribute a as the search condition setting unit 48, the attribute value “b2” or “b3” is designated for the attribute b, and the attribute c is the attribute value. The value “c2” is designated, and the attribute value “d1” is designated for the attribute d.

  Based on the specification of the attribute value of each attribute item by the search condition setting unit 48, the search unit 50 uses the specified attribute value a2, a3, b2, b3, c2, d1 to determine the attribute value for each file in the attribute information storage unit 40. Similarity calculation is performed between the two items to calculate the similarity sa, sb, sc, sd for each of the attribute items a to d.

  In this example, the similarity of each attribute item is calculated as one point when the attribute value of the specified condition matches the attribute value of the attribute information storage unit 40. For example, the attribute value “d1” for the file “# 1” is calculated. ”Only for the file“ # 2, ”the attribute values a2, b2, and c2 match for the file“ # 2, ”and the attribute values a3, b3 for the file“ # 3 ”match. In addition, the file “# 4” is set to 0 because there is no matching attribute value.

On the other hand, a weight storage unit 62 is provided for the search unit 50, and weights Wa, Wb, Wc, and Wd are set for each of the attributes a to d. The similarity S is calculated using the similarities Sa, Sb, Sc, Sd for the attribute items a to d in the search unit 50 and the weights Wa, Wb, Wc, Wd for the attribute items a to d in the weight storage unit 62. To do.
Similarity S = Wa · a + Wb · b + Wc · c + Wd · d (1)

  Here, assuming that the initial value is “1” as the weights Wa to Wd by the weight storage unit 62, the similarity of the file # 1 is 1 point and the similarity of the file # 2 is 3 points according to the above equation (1). The similarity of # 3 is 2 points and the score of file # 4 is 0, and the search result output unit 54 outputs and displays the search results arranged in descending order of similarity.

  In the actual calculation of the similarity S, the weights obtained by normalizing the appearance probabilities of the attribute items so that the sum of the weights of all the attribute items is “1” are used. The normalized initial values of the four weights Wa, Wb, Wc, and Wd are “0.25”.

  Therefore, in the search processing of the present invention, the search viewpoint that is the feature of the file that the user is looking for is automatically extracted from the operation history such as preview and rearrangement on the search result of the user, and the search viewpoint is weighted. Thus, by reflecting in the weight used for the calculation of the similarity in the next search process, it is possible to perform a highly accurate search reflecting the user's search intention without imposing a burden on the user.

  In other words, the user operates the preview buttons 100-1 to 100-5 of the search file information 98-1 to 98-5 as necessary on the search result display unit 72 shown on the search screen 64 of FIG. Whether the target file is to be previewed or sorted according to the similarity of the selected attribute item by selecting one of the attribute items such as type, owner, date / time, or phrase in the sort operation unit 68 A browsing operation is performed to determine whether.

  A common attribute item is extracted from the browsing result for such a search result from the preview or rearrangement order intentionally browsed by the user, and the attribute item that the user is paying attention to is the weight of the formula (1). By adjusting the weight calculation so as to increase the search, a search reflecting the user's search viewpoint can be performed for the next search.

In the present invention, any of the following can be performed as a method of automatically extracting the user's search viewpoint and adjusting the weight.
(1) A method of calculating a weight by extracting a user's search viewpoint from a preview browsing operation;
(2) A method of calculating a weight by extracting a user's search viewpoint by a user rearrangement operation;
(3) A method of calculating a weight by extracting a user's search viewpoint from a browsing time by a preview browsing operation;
(4) A method of calculating a weight by combining at least two of the above (1), (2), and (3);
First, a method of calculating a weight by extracting a user's search viewpoint by a preview browsing operation will be described.

  FIG. 10 is a search result list table 101 of the search result display unit 72 in the search screen 64 of FIG. 8, and represents type 104, owner 106, date 108 and phrase 110 as attribute items corresponding to the file ID 102. . Assume that previews are executed for file # 1, file # 3, and file # 5 on such search result list display 101 as indicated by the arrows 115-1, 115-2, and 115-3 on the left side.

  A common attribute value is counted for each of the attribute item type 104, the owner 106, the date / time 108, and the word / phrase 110 for the files # 1, # 3, and # 5 on which the preview is executed, and the maximum frequency of the attribute value is calculated. A value obtained by dividing the maximum frequency by the number of previews represents the probability that the attribute value has appeared, and it can be said that the larger the value indicating the probability, the more the user is paying attention to the attribute value.

  If the value obtained by dividing the maximum frequency by the number of previews is set as the weight of the equation (1), a search that reflects the search viewpoint that is the attribute that the user is interested in in the next search can be realized.

  FIG. 11 is an explanatory diagram of the weight table 112 including the browsing operation history generated based on the files # 1, # 3, and # 5 in which the search result list table 101 of FIG. 9 is executed.

  In the weight table 112, type, owner, date, and phrase attribute values are set as the number of searches 114, the number of previews 116, and the weight 118. Here, the weight 118 is divided into two rows, an upper row and a lower row, the upper row is the appearance probability of each attribute, and the lower row is the weight obtained by normalizing the appearance probability of each attribute.

  The search result list table 101 in FIG. 10 is used for weight calculation when the number of previews 116 in the weight table 112 in FIG. 11 is “first” and the number of previews 116 is “3”.

For this weight calculation, the appearance probability for each attribute item is the appearance probability = (maximum frequency of counting common attribute values) / (number of previews)
(2)
Calculate as

For example, taking the type 104 in the attribute item of FIG. 10 as an example, the maximum frequency of counting common attribute values is the frequency “2” of the attribute value “abc”. In this case, the appearance probability Pa of the type 104 based on the browsing operation history is Pa = 2/3.
It is obtained as

Since the maximum frequency of the common attribute value of the next owner 106 is the frequency “2” of “user U1”, the appearance probability Pb of the owner is Pb = 2/3.
It becomes. For the date and time 108, since all attribute values are different, the maximum frequency of the common attribute value is “1”, and the appearance probability Pc is Pc = 1/3.
It becomes.

Further, for the word / phrase 110 as the attribute value, since three words / phrases are registered in the files # 1, # 3, and # 5 after previewing, the total number is nine, and the words # 1, # 3, and # 5 Since the maximum frequency “2” of the commonly included attribute value “apple” is “2”, the appearance probability Pd is Pd = 2/9.
It becomes.

  Thus, for the appearance probabilities 2/3, 2/3, 1/3, and 2/9 for each attribute item calculated from the operation history of the number of previews, the sum of the appearance probabilities of the attribute items is “1”. To obtain the weights Wa, Wb, Wc, and Wd of each attribute item.

Normalization is calculated by calculating the sum 17/9 of the appearance probabilities 2/3, 2/3, 1/3, and 2/9 of each attribute item, and multiplying each appearance probability by the reciprocal (9/17) of the sum. It becomes as follows.
Type weight Wa = 6/17
Owner weight Wb = 6/17
Weight of date and time Wc = 3/17
Word weight Wd = 2/17
When this is represented by a decimal number, it becomes 0.35, 0.35, 0.18, and 0.12, as shown in FIG. 11, and the total is “1”.

  The weight table 112 of FIG. 11 stores the case where the number of searches 114 is repeated from the first time to the fifth time, and the preview times from the first time to the fifth time are 3, 5, 2, 4 times. , This indicates that it has been executed six times. For the second and subsequent times, the weight is calculated assuming that three words / phrases are registered in all the files # 1 to # 5.

  For the appearance probability of each attribute item after the second time, the sum obtained by adding the maximum frequency of the previous search to the current search is also obtained from the following equation as the sum total of the number of previews up to the previous time. Here, n is the current number of searches.

For example, when the appearance probability of the “second time” type in the search number 114 is viewed, the maximum frequency of the common attribute value in the second preview number 5 is “4”. The maximum frequency “2” is added to “2 + 4 = 6”. The total number of previews is “3 + 5 = 8” by adding the first and second times. Therefore, the appearance probability Pa of the second type is Pa = 6/8
It becomes.

  When the weight for each attribute item finally obtained by the execution of the preview by repeating the number of searches from the first time to the fifth time is seen, the highest normalized weight is the type “0”. .47 ”, and it can be seen from the execution of the user preview that the user's attention level is high for the attribute item“ type ”, and this is reflected in the search result by weight calculation.

  FIG. 12 is a flowchart of a search process according to the present invention that employs the method of extracting the user's search viewpoint and calculating the weight by the preview browsing operation shown in FIGS.

  In FIG. 12, an attribute item is set as a search condition in step S1, and then the similarity is calculated for all files from the equation (1) in step S2. The weights as initial values in this case are all the same value for each attribute item.

  Subsequently, in step S3, the search result is displayed as in the search screen of FIG. 8 according to the similarity. Subsequently, in step S4, whether or not the preview is executed is checked. If the preview is performed, the process proceeds to step S5, and the common attribute values of the preview file are counted.

  If there is no preview in step S4, whether or not the file is the target file is checked based on the user's input instruction in step S6. 8 is determined based on whether or not the initialization button 88 provided in the search condition setting operation unit 66 is operated.

  If the initialization button 88 is operated, it is determined that recursive search is executed, and the process proceeds to step S8. As shown in the weight table 112 of FIG. 11, the weight for each attribute item is determined from the preview browsing operation history for the current search result. Then, the process returns to step S1 again, and the next search process is repeated. At this time, the weight based on the preview operation history information calculated in step S8 is used to calculate the similarity in step S2.

  Next, a method for calculating a weight by extracting a user's search viewpoint by a rearrangement operation will be described.

  Here, the rearrangement operation is any of the type rearrangement button 90, the owner rearrangement button 92, the date / time rearrangement button 94, and the phrase rearrangement button 96 in the switching operation unit 68 provided on the search screen 64 of FIG. The search file information 98-1 to 98-5 in the search result display portion is selected in ascending order from the highest degree of similarity, or conversely, from the lowest degree of similarity, according to the similarity of the selected operation item. This is an operation to sort the search results in some descending order to make the search results easier to see.

  For example, the search result display unit 72 in the search screen 64 of FIG. 8 becomes the search result list table 101 of FIG. 10 before the rearrangement. If the attribute item “type” is selected and rearranged, the search of FIG. The result list table 101-1 is obtained.

  The sorting operation can be executed by selecting a plurality of attribute items. For example, when sorting is performed in the order of “type” and “owner”, sorting is performed first according to the similarity of “owner”, and then “type” is applied to the result having the same owner. ”Is executed.

  FIG. 14 shows a weight table 120 in which the user's search viewpoint is extracted from the search result list table 101-1 by the rearrangement operation of FIG. 13 and the weight is calculated. The weight table 120 has a search count 122 and a sort count 124. And weight 126 for each type, owner, date, and phrase.

The browsing operation history of rearrangement in the weight table 120 of FIG. 14 is executed “4 times” as the rearrangement count 124 for “first” in the search count 122. This four-time sorting is performed by selecting attribute items in the following order.
1st time: Word 2nd time: Type 3rd time: Owner 4th time: Date and time Here, the higher the score, the higher the score is set for the order of the number of times of sorting for one search result.

  FIG. 15 is an explanatory diagram of the sort order score file 128 for setting scores according to the rank with respect to the sort count in one search result. For the first time in FIG. 14, the number of times of rearrangement is four, so according to the order of execution of rearrangement on the time axis, from the first time to the fourth time, the newest fourth score from the present is the highest score of 4 points, The older third score is 3, the older second score is 2, and the oldest first score is 1. The score according to this rearrangement order is shown as a number in parentheses on the right side of the upper stage of the weight in FIG.

  Therefore, the weights of the attribute items “type”, “owner”, “date / time”, and “phrase” associated with the number of sorting times 4 for the first search result are “1/10”, “2/10”, “3 / 10 ”,“ 4/10 ”, and the weight of each attribute item is obtained by normalization so that the sum of the weights becomes 1, and“ 0.20 ”,“ 0.30 ”, 0.40”, “0” .10 "is obtained.

  In the case of FIG. 14, the number of times of sorting for the first to fifth searches is 4, 4, 1, 2, and 3. For the second and subsequent times, the weight of each attribute item is Ask for.

In the second search, the number of sorts is 2,
First time: Type second time: Owner, therefore, referring to the number of times of rearrangement from the rearrangement order score file 128 in FIG. 15, 3 points for the first time and 4 points for the second time are set.

  A value obtained by dividing the score obtained by the rearrangement by the total score is cumulatively added each time the search is repeated to calculate a weight, and used as a weight normalized to the formula (1). Search according to the search viewpoint extracted from the rearrangement operation accompanying the result is possible.

  In the case of FIG. 14, since the weight of the attribute item “owner” calculated from the reordering for the fifth search result is “0.49”, the user finally has the attribute item “owner”. It turns out that interest is high.

  FIG. 16 is a flowchart of a search process using a process of extracting a user's search viewpoint by a sorting operation on the search results shown in FIGS. 13 to 15 and calculating a weight.

  In FIG. 16, attribute items are set as search conditions in step S1, similarity is calculated for all files in accordance with equation (1) in step S2, and search results are displayed in accordance with the similarity in step S3.

  In step S4, it is determined whether or not reordering has been performed on the search result. When the reordering is performed, in step S5, a score corresponding to the reordering attribute value, specifically, the reordering order is counted.

  If reordering is not performed, it is determined in step S6 whether or not a user is determined whether the file is the target file. If it is not the target file, it is determined whether or not recursive search is performed in step S7. If it discriminate | determines, it will progress to step S8, will calculate the weight of each attribute item from the score of the order | rank obtained from the frequency | count of rearrangement as shown in the weight table 120 of FIG. 14, will return to step S1, and will perform step S2 in the next search. The weight calculated in step S8 is reflected in the similarity calculation.

  Next, a method for calculating the weight by extracting the user's search viewpoint from the preview time associated with the preview browsing operation for the search result will be described.

  In the extraction of the search viewpoint using the preview operation shown in FIGS. 10 to 12, the weight is calculated using the number of previews. Alternatively, the user's search viewpoint can be determined even using the preview time. The weight can be calculated by extraction. In other words, it can be determined that the longer the user's preview time, the closer to the search result the user is seeking.

  The method for calculating the weight based on the preview time is as shown in the weight table 130 of FIG. The weight table 130 of FIG. 17 includes the number of searches 132, the preview time 134, and the weight 136 for each attribute item of type, owner, date, and phrase.

  The weight table 130 for extracting the user's search viewpoint using the preview time is obtained by replacing the preview count 116 in the weight table 112 for extracting the user's search viewpoint from the preview count shown in FIG. It is.

  For example, in the first search, “30 seconds” is obtained as the preview time, and the maximum frequency of the common attribute value in the previewed file is obtained by this preview time, and this is divided by the preview time. Become. That is, the appearance frequency of each attribute item per unit preview time is given by the following equation.

It becomes. A value obtained by normalizing the appearance frequency per unit preview time of each loss correction item is used as a weight.

  FIG. 18 is a flowchart of a search process using a method for extracting a user's search viewpoint from a preview time and calculating a weight. In FIG. 18, an attribute item as a search condition is set in step S1, similarity is calculated for the above expression (1) for all files in step S2, and search results are displayed in accordance with the similarity in step S3.

  Subsequently, when it is determined in step S4 that the preview is to be executed, attributes common to the preview and the viewing time are counted in step S5. If the preview is not executed, it is checked in step S6 whether it is the target file. If it is not the target file, it is determined whether recursive search is executed in step S7.

  If it is a recursive search execution, the process proceeds to step S8, the attribute weight calculation based on the preview time as shown in the weight table 130 of FIG. 17 is performed, the next search is performed from step S1, and the next search process is performed in step S2. The weight of the attribute corresponding to the preview time calculated in step S8 is reflected in the calculation of the similarity.

  FIG. 19 is a flowchart of search processing according to the present invention using a method of calculating weights used in the next search using the preview count, the sort count, and the preview time as the browsing history.

  In this search process, an attribute item is set as a search condition in step S1, and after calculating similarity with the above equation (1) for all files in step S2, search results are obtained according to the similarity in step S3. indicate.

  When a browsing operation is executed for the search result in step S4, if it is a browsing operation such as preview or rearrangement in step S5, the common attribute value of the preview file is counted in step S5, and step S6. In step S7, the common attributes of the preview file and the browsing time are counted, and in step S7, the score corresponding to the rearrangement order is counted for each attribute item.

  If the browsing operation is not executed in step S4, the presence or absence of the target file is checked in step S8. If it is not the target file, whether or not the recursive search is executed is checked in step S9. Select a combination.

As the weight calculation, at least two combinations of (1) weight calculation reflecting the number of previews, (2) weight calculation reflecting the preview time, and (3) weight calculation reflecting the number of rearrangements are selected.

  Subsequently, for the combination selected in step S11, the weight of the attribute item is calculated from each history operation information, and for each weight calculation result, an average of the two weights is calculated for the same attribute item, and the next search is performed. The weight used for calculating the similarity is set, and the next search process is resumed from step S1.

  In the flowchart of FIG. 19, weight calculation is performed by selecting at least two combinations of (1) to (3), but in addition to this, (1) and (2), You may make it involve the weight calculation which fixedly set (2) and (3) or (1) (2) (3).

  Further, the present invention provides a file search program executed by a computer, and the file search program of the present invention has the processing contents shown in the flowchart of FIG. 12, FIG. 16, FIG. 18 or FIG. .

  The present invention also provides a storage medium such as a computer storing a file search program. This storage medium is a card-type storage such as a CD-ROM, floppy (R) disk, DVD disk, magneto-optical disk, or IC card. In addition to a storage device such as a hard disk provided inside or outside the medium or the computer system, it includes a database for holding a program via a line, another computer system and its database, and a transmission medium on the line.

  The present invention includes appropriate modifications that do not impair the object and advantages thereof, and is not limited by the numerical values shown in the above embodiments.

Claims (20)

In a computer with a file storage unit that is registered in association with attribute information extracted for each file,
A search condition setting step for setting attribute information as a file search condition;
For each file in the file storage unit, a search step of searching for a file by calculating similarity using a weight from the attribute information of the search condition and the registered attribute information;
A search result display step of displaying a search result of the file by the search step and generating a display state according to a browsing operation for the search result;
A browsing operation extraction step of extracting a browsing operation history by the search result display step;
A weight calculating step of calculating a weight used for calculating the similarity of the search step from the browsing operation history, and setting the weight calculated in the search step when performing a search again;
A file search program characterized by causing
2. The file search program according to claim 1, wherein the attribute information includes a plurality of attribute items, and the search step is a sum obtained by multiplying the similarity of a file by the similarity of each attribute item and the weight of each attribute item. The file search program characterized by calculating as follows.
In the file search program according to claim 2,
The search result display step can display a preview of the file as a search result by a browsing operation,
The browsing operation extraction step calculates the frequency of each attribute item of one or more previewed files and extracts it as a browsing operation history,
The file weight calculation step calculates a weight for each attribute item according to the frequency of each attribute item in the browsing operation history.
In the file search program according to claim 3,
The browsing operation extraction step counts the common attribute value included in each attribute item of the plurality of previewed files and extracts the maximum frequency of the attribute value for each attribute item,
The weight calculation step calculates the weight of each attribute item as a normalized value obtained by dividing the maximum frequency for each attribute item by the number of previews.
In the file search program according to claim 2,
The search result display step can rearrange and display the search result files in ascending order or descending order according to the similarity of the selected attribute item for any one of the plurality of attribute items.
The browsing operation extraction step calculates a value according to the rank of each attribute item selected for sorting and extracts it as a browsing operation history,
The file calculation program characterized in that the weight calculation step calculates a weight for each attribute item according to a rank of each attribute item in the browsing operation history.
In the file search program according to claim 5,
In the browsing operation extraction step, for the rank of each attribute item selected for sorting, a score that becomes a higher value in order of the sorting operation is calculated and extracted as a browsing operation history,
The file weight calculation step calculates the weight of each attribute item as a normalized value obtained by dividing the score for each attribute item by the number of times of sorting.
In the file search program according to claim 2,
The search result display step can display a preview of the file as a search result by a browsing operation,
In the browsing operation extraction step, the frequency and browsing time of each attribute item of one or more previewed files are calculated and extracted as a browsing operation history,
The file calculation program characterized in that the weight calculation step calculates a weight for each attribute item according to the frequency and browsing time of each attribute item in the browsing operation history.
In the file search program according to claim 7,
The browsing operation extraction step counts the common attribute value included in each attribute item of the plurality of previewed files and extracts the maximum frequency of the attribute value for each attribute item,
The file weight calculation step calculates the weight of each attribute item as a normalized value obtained by dividing the maximum frequency for each attribute item by the browsing time.
In the file search program according to claim 2,
The search result display step is capable of displaying a preview of the file as a search result by a browsing operation, and in ascending order according to the similarity of the selected attribute item with respect to any one of the plurality of attribute items. The search result files can be sorted and displayed in descending order,
The browsing operation extraction step includes:
A first extraction step of calculating a frequency of each attribute item of the previewed file or files and extracting it as a browsing operation history;
A second extraction step of calculating a value according to the rank of each attribute item selected for sorting and extracting as a browsing operation history;
The browsing operation extraction step calculates a frequency and a browsing time of each attribute item of one or a plurality of files that have been previewed, and extracts the browsing operation history as a third extraction step;
Comprising at least two of
The weight calculation step corresponds to the browsing operation extraction step,
A first weight calculating step of calculating a weight for each attribute item according to the frequency of each attribute item of the browsing operation history in the first extracting step;
The weight calculating step by the second extracting step includes a second weight calculating step of calculating a weight for each attribute item according to a rank of each attribute item in the browsing operation history;
A third weight calculating step of calculating a weight for each attribute item according to the frequency and browsing time of each attribute item of the browsing operation history by the third extracting step;
A file search program comprising at least any two of the above.
10. The file search program according to claim 9, wherein the weight calculation step is performed for each attribute item calculated at the time of combining at least two of the first weight calculation step to the third weight calculation step. A file search program characterized in that an average of weights is calculated and set in the search step.
2. The file search program according to claim 1, wherein the search condition setting step sets attribute values of a plurality of attribute items input using a search screen as search conditions.
2. The file search program according to claim 1, wherein the search condition setting step extracts the attribute value of the attribute item from the file specified by using the search surface and sets the search condition. File search program.
13. The file search program according to claim 1, wherein the attribute information includes a type, an owner, a creation or update date and time, and a phrase as attribute items.
In a file search method for searching for a target file from a file storage unit that registers and associates attribute information extracted for each file,
A search condition setting step for setting attribute information as a file search condition;
For each file in the file storage unit, a search step of searching for a file by calculating similarity using a weight from the attribute information of the search condition and the registered attribute information;
A search result display step of displaying a search result of the file by the search step and generating a display state according to a browsing operation on the search result;
A browsing operation extraction step of extracting a browsing operation history by the search result display step;
Calculating a weight to be used for calculating the similarity in the search step from the browsing operation history, and setting a weight in the search step when performing a search again;
A file search method characterized by comprising:
15. The file search method according to claim 14, wherein the attribute information includes a plurality of attribute items, and the search step is a sum obtained by multiplying the similarity of a file by the similarity of each attribute item and the weight of each attribute item. A file search method characterized by calculating as follows.
The file search method according to claim 15, wherein
The search result display step can display a preview of the file as a search result by a browsing operation,
The browsing operation extraction step calculates the frequency of each attribute item of one or more previewed files and extracts it as a browsing operation history,
The file calculation method according to claim 1, wherein the weight calculation step calculates a weight for each attribute item according to a frequency of each attribute item in the browsing operation history.
The file search method according to claim 15, wherein
The search result display step can rearrange and display the search result files in ascending order or descending order according to the similarity of the selected attribute item for any one of the plurality of attribute items.
The browsing operation extraction step calculates a value according to the rank of each attribute item selected for sorting and extracts it as a browsing operation history,
The file calculation method according to claim 1, wherein the weight calculation step calculates a weight for each attribute item according to a rank of each attribute item in the browsing operation history.
The file search method according to claim 15, wherein
The search result display step can display a preview of the file as a search result by a browsing operation,
In the browsing operation extraction step, the frequency and browsing time of each attribute item of one or more previewed files are calculated and extracted as a browsing operation history,
The weight calculation step calculates a weight for each attribute item according to the frequency and browsing time of each attribute item in the browsing operation history.
The file search method according to claim 15, wherein
The search result display step is capable of displaying a preview of the file as a search result by a browsing operation, and in ascending order according to the similarity of the selected attribute item with respect to any one of the plurality of attribute items. The search result files can be sorted and displayed in descending order,
The browsing operation extraction step includes:
A first extraction step of calculating a frequency of each attribute item of the previewed file or files and extracting it as a browsing operation history;
A second extraction step of calculating a value according to the rank of each attribute item selected for sorting and extracting as a browsing operation history;
The browsing operation extraction step calculates a frequency and a browsing time of each attribute item of one or a plurality of files that have been previewed, and extracts the browsing operation history as a third extraction step;
Comprising at least two of
The weight calculation step corresponds to the browsing operation extraction step,
A first weight calculating step of calculating a weight for each attribute item according to the frequency of each attribute item of the browsing operation history in the first extracting step;
The weight calculating step by the second extracting step includes a second weight calculating step of calculating a weight for each attribute item according to a rank of each attribute item in the browsing operation history;
A third weight calculating step of calculating a weight for each attribute item according to the frequency and browsing time of each attribute item of the browsing operation history by the third extracting step;
A file search method comprising at least any two of the above.
A file storage unit registered in association with attribute information extracted for each file;
A search condition setting section for setting attribute information as a file search condition;
For each file in the file storage unit, a search unit that searches for a file by calculating similarity from the attribute information of the search condition and registered attribute information using a weight;
A search result display unit that displays a search result of the file by the search unit and generates a display state according to a browsing operation on the search result;
A browsing operation extraction unit for extracting a browsing operation history by the search result display unit;
Calculating a weight to be used for calculating the similarity of the search unit from the browsing operation history, and setting a weight to the search unit when performing a search again;
A file search apparatus comprising:

Images