JP2006221560A - Data substitution device, data substitution method, and data substitution program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To precisely detect personal information to be substituted with other data. <P>SOLUTION: This data substitution device 1 has a reading means for reading a document data, an extraction means for analyzing the document data by syntax to extract the personal information, a substitution means for substituting the each personal information extracted by the extraction means with a data different from the personal information, and an output means for outputting the document data substituted by the substitution means. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a technique for replacing data.

In recent years, there has been an increasing interest in handling confidential information such as personal information. That is, when a document including information to be concealed, such as personal information, is disclosed, it is necessary to delete or mask (for example, black out) a portion where the information to be concealed is described. For example, Patent Document 1 describes an information concealment method in which a mask screen is embedded in information on a portion to be concealed.
JP 2003-298732 A

  By the way, in a company database, daily business reports for reporting business activities, inquiries from consumers sent to call centers, etc. are digitized as text data and accumulated in large quantities. Such text data generally includes personal information such as name, telephone number, and mail address. Therefore, when text data including personal information is disclosed outside or inside the company to share knowledge, there is a risk that personal information may be leaked.

  Therefore, when the person in charge publishes text data including personal information, the person in charge needs to visually search for a portion corresponding to the personal information, and perform mask processing on the portion of the personal information found. However, if the text data to be processed is enormous, it will be a heavy workload for the person in charge to detect the portion corresponding to the personal information and perform mask processing, and there is a risk that personal information may not be detected or missed. There is.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to extract personal information with higher accuracy and replace the extracted personal information with other data.

  In order to solve the above-described problems, the present invention provides, for example, a data replacement device, a reading unit that reads document data, an extracting unit that parses document data and extracts personal information, and an extracting unit extracts And replacing means for replacing each piece of personal information with data different from the personal information, and output means for outputting the replaced document data replaced by the replacing means.

  According to the present invention, personal information can be detected with higher accuracy and replaced with other data.

  Hereinafter, embodiments of the present invention will be described.

  FIG. 1 is a functional configuration diagram of a data replacement device 1 to which an embodiment of the present invention is applied. The illustrated data replacement device 1 is a device that extracts personal information from a document file including text data and replaces it with other data. The personal information is information relating to an individual, and is information that can identify a specific individual (including information that can identify a specific individual by collating with other information). . The personal information may be, for example, name, address (including postal code), telephone number, e-mail address, date of birth, credit number, account number, and the like.

 As shown in the figure, the data replacement device 1 of the present embodiment includes a data reading unit 11, a data extraction unit 12, a replacement processing unit 13, a data output unit 14, a data management unit 15, an input file 21, An output file 22, a user dictionary 23, a replacement rule table 24, and a syntax analysis rule 25 are included.

 The data reading unit 11 reads data of a document file stored in the input file 21. The data extraction unit 12 extracts personal information from the data read by the data reading unit 11. The replacement processing unit 13 replaces the personal information extracted by the data extracting unit 12 with other data. The data output unit 14 outputs the data in which the personal information is replaced to the output file 22 or an output device (not shown). The data management unit 15 manages (adds, updates, deletes) data in the user dictionary 23, the replacement rule table 24, and the syntax analysis rule 25.

  It is assumed that various document files are stored in the input file 21. Document files include, for example, daily business reports for reporting business activities, inquiry information from consumers sent to the call center, emails sent back to customers, data written on bulletin board systems, external vendors, etc. Consignment data to the customer or HTML data created for each customer can be considered. Such document files generally include personal information such as names and telephone numbers.

  Note that the document file may include data other than text data (image data, control codes such as tabs and line feeds). The input file 21 may be a database in which such a document file is stored. Further, when the document file stored in the input file 21 is input / created using an input unit (not shown) of the data replacement device 1, a storage device (such as an MO (Magneto-Optical disk) or a CD-ROM) ( For example, when data of a storage medium is input or when a file is transferred from another computer system (not shown) via a network.

  The output file 22 stores a document file after the replacement processing unit 13 replaces personal information. The user dictionary 23 is a database in which an arbitrary character string corresponding to personal information is registered. The replacement rule table 24 is a table in which rules for replacing personal information are set. The parsing rule 25 stores various data necessary for calculating the reliability of personal information. The user dictionary 23, the replacement rule table 24, and the syntax analysis rule 25 will be described later.

  The data replacement device 1 described above includes, for example, a CPU 901, a memory 902, an external storage device 903 such as an HDD, an input device 904 such as a keyboard and a mouse, and an output device such as a display and a printer as shown in FIG. A general-purpose computer system including 905 and a communication control device 906 for network connection can be used. In this computer system, each function of the data replacement device 1 is realized by the CPU 901 executing a predetermined program loaded on the memory 902. That is, each function of the data replacement device 1 is realized by the CPU 901 of the data replacement device 1 executing a program for the data replacement device 1. The input file 21, the output file 22, the user dictionary 23, the replacement rule table 24, and the syntax analysis rule 25 of the data replacement device 1 are stored in the memory 902 or the external storage device 903. Moreover, the communication control apparatus 906 shall be provided as needed.

  Next, the user dictionary 23 will be described.

  The user dictionary 23 is a database in which an arbitrary character string corresponding to personal information is registered. The user dictionary 23 of this embodiment has a group name to which the character string belongs and a replacement application flag for each registered character string. The group name is a type (category) for classifying the word indicated by the character string. For example, if the registered character string is a customer name, "customer" is the group name, if the registered character string is an employee name, the group name is "employee", and the registered character string is the business partner In the case of the company name, “business partner” is set as the group name.

  The replacement application flag is a flag for determining whether or not a character string registered in the user dictionary 23 is a replacement target in a replacement process described later. For example, if the replacement application flag is “ON”, the character string is extracted as a replacement target. On the other hand, when the replacement application flag is “OFF”, even the character string registered in the user dictionary 23 is excluded from the replacement target.

  FIG. 3A is a diagram showing an example of a setting screen (or reference screen) for the user dictionary 23. The data management unit 15 receives an instruction input by the user using the input device, and displays a setting screen illustrated in the drawing on the output device.

  On the illustrated setting screen, a check box 31, a group name 32, and a character string 33 for setting a replacement application flag are displayed for each character string registered in the user dictionary 23. In the present embodiment, when a check is input in the check box 31, the data management unit 15 sets “ON (replacement target)” in the replacement application flag of the user dictionary 23. On the other hand, when a check is not input in the check box 31, the data management unit 15 sets “OFF (not subject to replacement)” in the replacement application flag.

  By displaying the check box 31 and accepting a user input (instruction), even a character string registered in the user dictionary 23 can be individually and temporarily excluded from replacement. That is, the user can flexibly change the user dictionary 23 according to the type and attribute of the document file for which the replacement process is executed.

  In addition, an add button 34, an edit button 35, a delete button 36, an OK button 37, and a CANCEL button 38 are displayed on the setting screen shown in the figure. When the add button 34 is pressed, the data management unit 15 displays an add dialog box (dialog box) for registering a new character string. Further, when the edit button 35 is pressed, the data management unit 15 displays an edit dialog box for editing the character string selected by the cursor. Further, when the delete button 36 is pressed, the data management unit 15 deletes the character string selected with the cursor from the user dictionary 23.

  FIG. 3B is a diagram showing an example of an edit dialog box displayed when the edit button 35 is pressed. The editing dialog box has a group input field 351 using a drop-down list and a character string input field 352. In the case of the editing dialog box, the group input field 351 and the character string input field 352 are displayed in advance with the group name and character string to be edited selected on the setting screen of FIG. And The user presses the OK button 353 after inputting (after changing) the group name and the character string. The data management unit 15 accepts pressing of the OK button 353 and updates the user dictionary 23 with the input content. Further, the data management unit 15 accepts pressing of the CANCEL button 354, and transitions to the setting screen of FIG. 3A without updating the user dictionary 23.

Next, the replacement rule table 24 will be described.
FIG. 4 is a diagram showing an example of the replacement rule table 24. The replacement rule table 24 is a table in which rules for replacing personal information are set. The replacement rule table 24 shown in the figure has an extraction method 41, a rule name 42, a regular expression character pattern 43, a replacement character string 44, an ID flag 45, an application flag 46, and a priority 47.

  In the extraction method 41, a method for extracting a character string corresponding to personal information (including a character string that may be personal information) is set. In this embodiment, three types of extraction methods are used: extraction by syntax analysis, extraction by regular expressions, and extraction by the user dictionary 23. The extraction method by syntactic analysis is a method of analyzing a sentence structure (for example, dependency structure) by syntactic analysis and extracting a character string having a high possibility of personal information. The extraction method by syntax analysis will be described later.

  The regular expression extraction method is information that corresponds to personal information, and is a method of extracting a regular expression character string that can be identified from the numbering system, the type of character used (character pattern), and the like. As regular expression data, for example, a telephone number, a mail address (domain name), a postal code, a date of birth, a credit number, an account number, and the like are conceivable. It is assumed that the user can register an arbitrary number system and character pattern in the replacement rule table 24 using a setting screen described later. The extraction method by the user dictionary 23 is a method of matching and extracting the same character string as each character string registered in the user dictionary 23 described above.

  In the rule name 42, the type (category) of personal information and whether to replace the personal information of the type are set. In the replacement rule table 24 shown in the figure, when the extraction method 41 is syntax analysis, a rule for replacing a person name and a rule for replacing a place name are set. When the extraction method 41 is a regular expression, a rule for replacing a telephone number and a rule for replacing an e-mail address are set.

  When the extraction method 41 is a user dictionary, a rule for replacing a character string whose group name is “customer” and a rule for not replacing a character string whose group name is “employee” are set. For example, in the case of a document file that is disclosed only within the company (for example, when sharing knowledge within a predetermined department, when a supervisor evaluates the work of subordinates), there is a case where it is desired to know the employee name. In such a case, by applying a rule that does not replace the character string whose group name is “employee”, the employee name can be excluded from replacement.

  In the regular expression character pattern 43, the regular expression numbering system and the type of character used (character pattern) are set. In the illustrated regular expression character pattern 43, numbers / symbols and number systems used for telephone numbers and alphanumeric characters / symbols and name systems used for mail addresses are set. In the replacement character string 44, a character string after conversion (replacement) of the extracted character string is set. When the rule name 42 is “employee (non-replacement)”, the employee name is not replaced, and a space is set in the replacement character string 44.

  The ID flag 45 is a flag for determining whether or not to add identification information for identifying each character string to which the same rule is applied to the replacement character string 44. In the illustrated example, the ID flag 45 of the rule name 42 of “person name (replacement)” is “present”. In this state, when two character strings of “Tanaka” and “Suzuki” are extracted by syntax analysis in one document file, the replacement processing unit 13 determines the replacement character string ([person name]). Immediately after that, identification information is given. In the present embodiment, a serial number indicating the extracted order is assigned as identification information. That is, the replacement processing unit 13 replaces “Tanaka” extracted previously with “[Personal name] 1” and “Suzuki” extracted next with “[Personal name] 2”.

  The application flag 46 is a flag for determining whether or not each rule registered in the replacement rule table 24 is applied in a replacement process described later. For example, when the application flag 46 is “ON”, the replacement process is performed by applying the rule. On the other hand, when the application flag 46 is “OFF”, the replacement process is performed without applying the rule.

  The priority order 47 indicates the priority order when the same character string is extracted in duplicate by a plurality of rules. For example, it is assumed that the same name and the same name are registered in the group name “customer” and the group name “employee” in the user dictionary 23. In this case, in the illustrated replacement rule table 24, the priority order 47 of the rule name 42 “customer (replacement)” is higher than the priority order of “employee (non-replacement)”. Accordingly, the replacement processing unit 13 gives priority to the rule “customer (replacement)” and replaces the duplicate name with [customer].

  FIG. 5A is a diagram showing an example of a setting screen (or reference screen) for the replacement rule table 24. Note that the data management unit 15 receives an instruction input by the user using the input device, and displays a setting screen illustrated in the drawing on the output device.

  The setting screen shown in the figure includes a check box 51 for setting an application flag 46 for each rule registered in the replacement rule table 24, and a display field 52 for displaying the setting contents set in the replacement rule table 24. Have. In the display column 52, a rule name, an extraction method, a regular expression character pattern, and the like are displayed.

  In this embodiment, the data management unit 15 sets “ON” in the application flag 46 when a check is input in the check box 51, and sets the application flag 46 in a case where a check is not input in the check box 51. Set “OFF”. By displaying the check box 51 and accepting a user input (instruction), even a rule registered in the replacement rule table 24 can be temporarily excluded from the application of the replacement process. That is, the user can select a rule to be applied according to the type and attribute of the document file on which the replacement process is executed.

  In addition, an add button 53, an edit button 54, a delete button 55, an OK button 56, and a CANCEL button 57 are displayed on the setting screen shown in the figure. When the add button 53 is pressed, the data management unit 15 displays an addition dialog box for newly registering a new rule. Further, when the edit button 54 is pressed, the data management unit 15 displays an edit dialog box for editing the rule selected by the cursor. Further, when the delete button 55 is pressed, the data management unit 15 deletes the rule selected with the cursor from the replacement rule table 24.

  FIG. 5B is a diagram showing an example of an editing dialog box displayed when the edit button 54 is pressed. The editing dialog box shown in the figure displays input fields, radio buttons, and check boxes for changing each item of the replacement rule table 24. That is, in the editing dialog box shown in the figure, an extraction method input field 541 using a drop-down list, a rule name input field 542, a regular expression character pattern input field 543, and radio buttons for inputting whether or not to replace A check box 545 for instructing whether or not to add an ID to the replacement character string, a replacement character string input field 546, and a priority order input field 547 are displayed.

  It is assumed that the setting contents of the rule to be edited selected on the setting screen of FIG. 5A are displayed in advance in the input fields of the editing dialog box. The user presses an OK button 548 after inputting desired setting contents (after change). The data management unit 15 accepts the pressing of the OK button 548 and updates the replacement rule table 24 with the input content. Further, the data management unit 15 accepts pressing of the CANCEL button 549, and returns to the setting screen of FIG. 5A without updating the replacement rule table 24.

  Next, the syntax analysis rule 25 will be described.

  The parsing rule 25 stores a personal name dictionary in which a person's first name and surname are registered, a place name dictionary in which names of prefectures and municipalities are registered, reliability data used for reliability calculation, and the like. In the present embodiment, a suffix indicating a person, a predicate indicating a person's action, a word that modifies a place name, a predetermined case particle, and the like are used for the reliability data. A specific example of reliability data will be described later.

  Next, a method for extracting a character string corresponding to personal information by syntax analysis will be described.

  The data extraction unit 12 extracts a character string corresponding to personal information described below using, for example, a morphological analysis engine and a syntax analysis engine. The morpheme analysis engine performs morphological analysis (Morphological Analysis), which is a technology of natural language processing, to decompose a sentence into morphemes (words) that are the minimum character strings, and specify the part of speech of each decomposed morpheme. The parsing engine analyzes a sentence structure (for example, a dependency structure) by performing syntactic analysis, which is a natural language processing technique.

  Note that morphological analysis and syntax analysis are described below, for example.

“Introduction to morphological and syntactic analysis”, [searched on January 26, 2005], Internet <URL: http://www.unixuser.org/~euske/doc/nlpintro/>
First, the data extraction unit 12 performs morphological analysis for each sentence (sentence) of a document file to be processed. That is, the data extraction unit 12 uses a morpheme analysis engine to decompose a sentence into morphemes (words) that are the minimum character strings, and specifies the part of speech of each decomposed morpheme. Then, the data extraction unit 12 extracts only the character string of the morpheme whose specified part of speech is a noun.

  Then, the data extraction unit 12 determines whether each extracted character string matches the character string registered in the personal name dictionary or the place name dictionary of the parsing rule 25. The data extraction unit 12 further extracts a character string registered in the personal name dictionary or the place name dictionary as personal information. Then, the data extraction unit 12 determines the reliability of the character strings registered in the personal name dictionary (hereinafter “personal name data”) and the character strings registered in the geographical name dictionary (hereinafter “place name data”). calculate. The reliability is an index indicating a possibility (reliability) corresponding to personal information.

  FIG. 6 is a process flow diagram for calculating the reliability of personal name data and place name data. Note that the data extraction unit 12 parses a sentence including personal name data or place name data using a parsing engine before performing the illustrated processing flow.

  First, the data extraction unit 12 inputs target data (person name data or place name data) to be processed, and sets the reliability of the initial value (S21). In the present embodiment, the reliability of the initial value is 50%. Then, the data extraction unit 12 refers to a sentence including the target data, and determines whether or not a suffix indicating a person (for example, “sama” or “san”) is added immediately after the target data. (S22). Note that a suffix indicating a person is stored in the reliability data of the parsing rule 25.

  If the suffix is added (S22: YES) and the target data is personal name data, the data extraction unit 12 adds 25% to the reliability of the initial value. If the target data is place name data, the data extraction unit 12 subtracts 25% from the reliability of the initial value (S23). On the other hand, when no suffix is added (S22: NO), the data extraction unit 12 proceeds to S24.

  Then, the data extraction unit 12 identifies a predicate having a dependency relationship with the target data by parsing a sentence including the target data. And it is discriminate | determined whether the predicate in dependency relation with object data is a predicate which shows a person's action (for example, "say", "speak", etc.) (S24). It is assumed that the predicate indicating the human action is stored in the reliability data of the parsing rule 25.

  If the predicate indicating a human action is in a dependency relationship (S24: YES), and the target data is personal name data, the data extraction unit 12 adds 25% to the reliability. If the target data is place name data, the data extraction unit 12 subtracts 25% from the reliability (S25). On the other hand, when the predicate indicating the human action is not in the dependency relationship (S24: NO), the data extraction unit 12 proceeds to S26.

  Then, the data extraction unit 12 refers to a sentence including the target data, and determines whether or not a word that modifies the place name (for example, “origin”, “bred”) is added immediately after the target data. (S26). It is assumed that the word that modifies the place name is stored in the reliability data of the parsing rule 25. If a word that modifies the place name is added (S26: YES) and the target data is personal name data, the data extraction unit 12 subtracts 25% from the reliability. If the target data is place name data, the data extraction unit 12 adds 25% to the reliability (S27). On the other hand, when the word which modifies a place name is not added (S26: NO), the data extraction part 12 progresses to S28.

  Then, the data extraction unit 12 parses a sentence including the target data to determine whether or not a word having a different case particle depends on the target data depending on whether it represents a human action or a place. It discriminate | determines (S28). Examples of words with different case particles include “go”. That is, when the case particle of “go” is “ga” (A is “go”), A is a person. In addition, when the case particle of “go” is “ni” (“go” to A), A is a place name. It should be noted that such words with different case particles and the case particles (case particles indicating persons, case particles indicating place names) of the words are stored in the reliability data of the parsing rule 25. . When words having different case particles are not in a dependency relationship with the target data (S28: NO), the data extraction unit 12 proceeds to S32.

When a word having a different case particle has a dependency relationship with the target data (S28: YES), the data extraction unit 12 determines whether or not the case particle of the word is a case particle indicating a human action (S29). . If it is a case particle indicating a human action (S29: YES), and the target data is personal name data, the data extraction unit 12 adds 25% to the reliability. If the target data is place name data, the data extraction unit 12 subtracts 25% from the reliability (S30). On the other hand, if it is a case particle indicating a place name (S29: NO) and the target data is personal name data, the data extraction unit 12 subtracts 25% from the reliability. If the target data is place name data, the data extraction unit 12 adds 25% to the reliability (S31).
Then, the data extraction unit 12 determines whether or not the reliability calculated by the processing from S21 to S31 exceeds 100% (S32). If it exceeds 100% (S32: YES), the data extraction unit 12 sets the reliability of the target data to 100% (S33). On the other hand, when it is 100% or less (S32: NO), the extraction unit 100 sets the reliability calculated by the processing from S21 to S31 as the reliability of the target data.

  In accordance with the reliability value described above, the data extraction unit 12 finally extracts personal name data and place name data that are highly likely to be personal information. That is, the data extraction unit 12 determines that personal name data and place name data having a reliability level equal to or higher than a predetermined value (for example, 50%) as personal information, and extracts the data for replacement with other data.

  Next, replacement processing performed by the data replacement device 1 will be described.

Before performing this process, the user uses the setting screens (see FIGS. 3 and 5) to set the user dictionary 23 and the replacement rule table 24 in consideration of the type and attributes of the document file to be processed. Change to
FIG. 7 is a flowchart of the replacement process.

  First, the data reading unit 11 reads a document file stored (accumulated) in an input file (S41). Then, the data extraction unit 12 performs extraction of personal information and calculation of reliability (see FIG. 6) by the above-described syntax analysis (S42).

  Then, the data extraction unit 12 extracts personal information using a regular expression and sets the reliability (S43). That is, the data extraction unit 12 extracts a character string corresponding to one of the regular expression character patterns 43 registered in the replacement rule table 24 as personal information. Then, the data extraction unit 12 sets a predetermined value (for example, 99%) for the reliability of the character string extracted by the regular expression.

  Then, the data extraction unit 12 extracts personal information by the user dictionary 23 and sets the reliability (S44). That is, the data extraction unit 12 extracts a character string that matches any character string registered in the user dictionary 23 as personal information. Even in the case of a character string registered in the user dictionary 23, the data extraction unit does not extract the character string when the group name is not replaced with “employee”. Then, the data extraction unit 12 sets a predetermined value for the reliability of each character string extracted using the user dictionary 23. Since the character string registered in the user dictionary 23 corresponds to personal information, a high reliability (for example, 100%) is set.

  Note that the order of the processing of S42, S43, and S44 is not limited to this, and may be performed in any order.

  Then, the replacement processing unit 13 replaces a character string whose reliability exceeds a predetermined value in the character string extracted by the data extracting unit 12 (S45). That is, based on the reliability calculated by the data extraction unit 12, the replacement processing unit 13 replaces only a character string that is highly likely to be personal information with other data (character string). As a result, a character string that is highly likely to correspond to personal information is masked, and the risk of leakage of personal information can be prevented even when the replaced document file is disclosed.

  Further, the replacement processing unit 13 refers to the replacement rule table 24 and specifies a replacement character string 44 corresponding to the rule name 42 of the extracted character string. Then, the replacement processing unit 13 replaces the extracted character string with the specified replacement character string 44. Further, the replacement processing unit 13 refers to the corresponding ID flag 45 of the replacement rule table 24, and adds “identification information” (for example, serial number) to the replacement character string 44 when “Yes”.

  Then, the data output unit 14 outputs the replaced document file subjected to the replacement process by the replacement processing unit 13 to the output file 22 (S46). Then, the data reading unit 11 determines whether or not an unprocessed document file exists in the input file 21 (S47). If there is an unprocessed document file (S47: YES), the data reading unit 11 returns to S41 and reads the unprocessed document file. If there is no unprocessed document file (S47: NO), the data reading unit 11 ends the replacement process.

  Next, the processing of FIG. 7 will be specifically described by taking the document file data shown in FIG. 8A as an example. 8B is a document file after replacement of the document file of FIG. Also, the replacement process is performed using the replacement rule table 24 shown in FIG. 4 and the user dictionary 23 shown in FIG.

  First, the data reading unit 11 reads the illustrated document file (S41). Then, the data extraction unit 12 performs the processes of S42, S43, and S44, and extracts a character string that may correspond to personal information.

  FIG. 9 is a diagram illustrating an example of an extracted character string table in which character strings extracted from the document file illustrated by the data extraction unit 12 are set. The data extraction unit 12 creates an extracted character string table as necessary and temporarily stores it in a memory or an external storage device. The extracted character string table shown in the figure includes an extracted character string 61, a position 62 of the extracted character string, an applied rule name 63, an extraction method 64, a reliability 65, and a replacement character string 66.

  The data extraction unit 12 extracts, for example, “Suzuki” 81 from the illustrated document file as a character string of personal name data by syntax analysis, and sets the extracted character string 61 a. Then, the data extraction unit 12 sets the extracted position of “Suzuki” and the corresponding rule name and extraction method of the replacement rule table 24 in the extracted character string table.

  Then, the data extraction unit 12 calculates the reliability of “Suzuki” by the process of FIG. That is, the data extraction unit 12 sets an initial value (50%) of personal name data (S21). Then, the data extraction unit 12 refers to the illustrated document file and adds “25%” to the initial value (50%) because there is a suffix (“san”) indicating a person immediately after “Suzuki” 81 ( S23). Then, the data extraction unit 12 parses the sentence including “Suzuki” 81 and adds 25% to the reliability because the predicate indicating the action (“listening”) has a dependency relationship with “Suzuki” 81. (S25). Therefore, the data extraction unit 12 sets the reliability of “Suzuki” 81 to 100%.

  In addition, the data extraction unit 12 extracts a character string of “Taro Yamamoto” 82 that matches the user dictionary 23 (group name: customer) from the document file illustrated, and sets the extracted character string 62b. Then, the data extraction unit 12 sets the extracted position, the corresponding rule name in the replacement rule table 24, and the extraction method in the extracted character string table. Further, the data extraction unit 12 sets a predetermined reliability (for example, 100%) of the user dictionary in the extracted character string table.

  The character string “Jiro Sudo” 83 in the illustrated document file matches the user dictionary 23 (group name: employee). However, in the case of the replacement rule table 24 shown in FIG. 4, the employee (non-replacement) rule is applied, so the data extraction unit 12 does not extract “Jiro Sudo” 83. “Sudo” and “Jiro” are extracted as personal name data by parsing. However, in the replacement rule table 24 shown in FIG. 4, the employee (non-replacement) rule has a higher priority than the person name (replacement) rule. Therefore, the data extraction unit 12 does not extract “Sudo” and “Jiro”.

  In addition, the data extraction unit 12 extracts the character string of the telephone number 84 that matches the regular expression character pattern in the replacement rule table 24 from the document file shown in the figure, and sets the extracted character string 62c. Then, the data extraction unit 12 sets the extracted position, the corresponding rule name in the replacement rule table 24, and the extraction method in the extracted character string table. Further, the data extraction unit 12 sets a predetermined reliability (for example, 99%) of the regular expression in the extracted character string table.

  Then, the replacement processing unit 13 sets the corresponding replacement character string in the replacement rule table 24 in the extracted character string table. Then, the replacement processing unit 31 converts the extracted character string 61 having a reliability 65 of a predetermined value (for example, 50% or more) into a replacement character string 66. The replacement processing unit 13 refers to the ID flag of the replacement rule table 24 and adds identification information (serial number) to the replacement character string 66 of [person name]. In addition, when there is a duplicated character string (“Taro Yamamoto” 62b, “Yamamoto” / “Taro” 62d), the replacement processing unit 13 determines that the replacement character string is highly reliable ([customer]). Replace.

  Then, the data output unit 14 outputs the replaced document file (FIG. 8B) to the output file 22.

  The embodiment of the present invention has been described above.

  The data extraction unit 12 of the present embodiment performs syntax analysis of the document file and calculates the reliability of the extracted character string (data) based on the analyzed sentence structure (see FIG. 6). Thereby, personal information can be extracted with higher accuracy. In other words, it is possible to reduce the leakage of personal information and extraction errors, and further reduce the risk of leakage of personal information.

  In the present embodiment, a character string that is highly likely to correspond to personal information is replaced (masked) with another character string. Thereby, even when the document file after replacement is disclosed, the personal information can be concealed.

  In the present embodiment, a character string that is highly likely to correspond to personal information is converted into a replacement character string corresponding to the type / attribute (person name, place name, telephone number, etc.) of the character string. Thereby, even when personal information is replaced (hidden), the type or attribute of the original personal information can be grasped.

  Further, the replacement rule table 24 of the present embodiment has an ID flag 45. Thereby, even when the personal information is replaced with the replacement character string, it is possible to identify whether or not the character string before the replacement is the same. For example, if personal information of the same character string appears multiple times in one document file, it is converted to a replacement character string with the same identification information added to the replaced document file, so the same personal name ( Alternatively, it can be recognized that the same place name is repeatedly described.

  In addition, this invention is not limited to said embodiment, Many deformation | transformation are possible within the range of the summary. For example, in the above embodiment, the personal information is replaced with a character string corresponding to the type of the personal information. However, the personal information may be converted into image data other than characters. It is also possible to perform a masking process such as deleting or blacking out personal information.

  In this embodiment, the data output unit 14 outputs the replaced document file to the output file 22 (FIG. 7: S46). However, the data output unit 14 may output the replaced document file to the output device. Thereby, the user can confirm the document file after replacement sequentially. Further, when outputting the replaced document file, the data output unit 14 may output the extracted character string table as shown in FIG. 9 to the output device together with the replaced document file. Thus, the user can replace the character string with low reliability by sequentially inputting a replacement instruction while referring to the replaced document data.

It is a figure which shows the structure of the data replacement apparatus with which one Embodiment of this invention was applied. It is a figure which shows the hardware structural example of a data replacement apparatus. It is a figure which shows an example of a user dictionary setting screen. It is a figure which shows an example of a replacement rule table. It is a figure which shows an example of the setting screen of a replacement rule table. It is a processing flowchart which calculates the reliability of parsing. It is a processing flow figure of substitution processing. It is a figure which shows an example of a document file. It is a figure which shows an example of an extraction character string table.

Explanation of symbols

1: data replacement device, 11: data reading unit, 12: data extraction unit, 13: replacement processing unit, 14: data output unit, 15: data management unit, 21: input file, 22: output file, 23: user dictionary 24: Replacement rule table 25: Parsing rule

Claims (6)

A data replacement device,
Reading means for reading document data;
Extracting means for parsing the document data and extracting personal information;
Replacement means for replacing each piece of personal information extracted by the extraction means with data different from the personal information;
Output means for outputting post-replacement document data replaced by the replacement means. The data replacement device according to claim 1, wherein
A storage means for storing a personal name dictionary and a place name dictionary;
The extraction means determines whether each character string of the document data matches the personal name dictionary or the place name dictionary, and whether the character string that matches the personal name dictionary or the place name dictionary corresponds to personal information. The reliability indicating whether or not is calculated based on the structure of the sentence analyzed by the syntax analysis,
The data replacing device, wherein the replacing means replaces a character string having the reliability exceeding a predetermined value with data different from the character string. The data replacement device according to claim 1, wherein
The replacing unit replaces the personal information extracted by the extracting unit with predetermined data according to the type of the personal information, and uses identification information for identifying the personal information for each type as the predetermined data. A data replacement device characterized by being added to the above. The data replacement device according to claim 1, wherein
A storage means for storing a user dictionary in which at least one character string corresponding to personal information is registered, and a regular expression list in which a predetermined numbering system or character pattern is registered;
The extraction means determines a character string that matches the character string registered in the user dictionary as personal information, extracts it from the document data,
A data replacement device characterized in that a character string that matches a numbering system or a character pattern registered in the regular expression list is identified as personal information and extracted from the document data. A data replacement method performed by an information processing apparatus,
The information processing apparatus includes a storage unit and a processing unit,
The processor is
A reading step of reading the document data stored in the storage unit;
An extraction step of parsing the document data to extract personal information;
A replacement step of replacing each piece of personal information extracted in the extraction step with data different from the personal information;
An output step of outputting the replaced document data replaced in the replacing step. A data replacement program executed by the information processing apparatus,
The information processing apparatus includes a storage unit and a processing unit,
In the processing unit,
A reading step of reading the document data stored in the storage unit;
An extraction step of parsing the document data to extract personal information;
A replacement step of replacing each piece of personal information extracted in the extraction step with data different from the personal information;
An output step of outputting the replaced document data replaced in the replacing step is executed.

Images