Text anonymization device and text anonymization method

Abstract

To realize flexible and efficient anonymization according to text data to be processed.SOLUTION: A text anonymization device includes a storage unit configured to store text data to be processed and a processing unit configured to perform processing to conceal a specific character string included in the text data to anonymize the text data to be processed. When a selection of a character string to be concealed is accepted from a user, the processing unit presents a character string related to the selected character string to be concealed, accepts a selection as to whether the related character string needs to be concealed, and anonymizes the text data based on the selected character string to be concealed and the selection result of whether the related character string needs to be concealed.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a text anonymization device and a text anonymization method. [Background technology]

[0002] Various methods have been proposed for anonymizing structured data. For example, the k-anonymization method is known, which abstracts the original data so that there are at least k pieces of data with the same conditions for a specified data item. On the other hand, for text data such as consultation records, nursing records, and questionnaire records in the welfare field, which may contain a large amount of personal information, no structure is defined, and the attributes, contents, and units of the information contained are not self-evident. Therefore, since anonymization methods for structured data, which are the mainstream of anonymization methods, cannot be applied, anonymization methods unique to text data such as the records have been proposed.

[0003] For example, Patent Document 1 proposes a text anonymization device that determines an anonymization level based on information about the usage environment of an electronic document, determines an anonymous expression for anonymizing proper nouns in the electronic document according to the anonymization level, and replaces proper nouns in the electronic document to be processed with the anonymous expression. Patent Document 1 states that "the anonymization level adjustment means 21 inputs a set of document IDs of documents to be accessed and conditions of the usage environment of the electronic document. The anonymization level adjustment means 21 determines an anonymization level according to the user's attributes, the usage location, and the usage method conditions such as viewing and copying. The proper name anonymization means 22 determines an anonymous expression for anonymizing proper names according to the anonymization level determined by the anonymization level adjustment means 21. Then, the document output means 24 replaces the proper names in the document with the anonymous expression determined by the proper name anonymization means 22 and outputs the electronic document to be accessed." [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP2006-185311A Summary of the Invention [Problem to be solved by the invention]

[0005] However, in general, in text data such as consultation records, nursing records, and questionnaire records in the welfare field, it is possible to infer individuals or anonymized words from contexts other than proper nouns and proper expressions, and anonymizing only proper nouns is insufficient. For example, consultation records in the welfare field, such as welfare assistance and child abuse, contain a lot of sensitive information about individuals, information about privacy, and family structure, so even if proper names and personal identification codes such as personal names, addresses, and phone numbers are anonymized, it is still possible to infer individuals from the context.

[0006] Furthermore, the anonymization method in Patent Document 1 describes a method that includes a storage means for storing proper names and their anonymous expressions in advance, and replaces proper names identified from a text to be processed with the anonymous expressions. However, since text data generally has a high degree of free description, it is difficult to comprehensively determine in advance proper names and categories and items of information to be anonymized that may be included in the text data. For example, in the case of consultation records in the welfare field, the categories, items, and contents of information that the recorder should write vary depending on the recorder and the person seeking consultation, so the categories of information included in the record contents must be continually checked or updated for each consultation case.

[0007] In view of the above problems, an object of the present invention is to realize flexible and efficient anonymization according to text data to be processed. [Means for solving the problem]

[0008] In order to achieve the above-mentioned object, one representative text anonymization device of the present invention includes a memory unit that stores text data to be processed, and a processing unit that anonymizes the text data to be processed by processing the text data to conceal a predetermined character string contained in the text data, and when the processing unit receives a selection from a user about a character string to be concealed, it presents character strings related to the character string selected to be concealed, receives a selection as to whether the related character strings need to be concealed, and anonymizes the text data based on the character string selected to be concealed and the selection result of whether the related character strings need to be concealed. Furthermore, one representative text anonymization method of the present invention is a text anonymization method for anonymizing text data to be processed by performing processing to conceal a predetermined character string included in the text data to be processed, characterized in that a text anonymization device accepts a user's selection of a character string to be anonymized, presents character strings related to the character string selected as the character string to be anonymized, accepts a selection as to whether the related character strings need to be anonymized, and anonymizes the text data based on the character string selected as the character string to be anonymized and the selection result of whether the related character strings need to be anonymized. Effect of the Invention

[0009] According to one aspect of the present invention, flexible and efficient anonymization can be achieved according to the text data to be processed. [Brief description of the drawings]

[0010] [Figure 1] FIG. 1 shows the overall configuration of the system. [Diagram 2] FIG. 4 is an explanatory diagram of a basic information table. [Diagram 3] FIG. 13 is an explanatory diagram of a consultation record table. [Figure 4] FIG. 13 is an explanatory diagram of a selected word history table. [Diagram 5] FIG. 13 is an explanatory diagram of a thesaurus master table. [Figure 6] FIG. 13 is an explanatory diagram of a co-occurrence word master table. [Figure 7]Processing flow of the text anonymization device. [Figure 8] An example of the results of anonymizing proper nouns. [Figure 9] Example data showing the locations of selected words. [Figure 10] Example of selected words history table before update. [Figure 11] An example of an updated selected words history table. [Figure 12] Examples of specific locations where synonyms occur. [Figure 13] An example of identifying the location of co-occurring words. [Figure 14] An example of past word selection localization. [Figure 15] An example of updating the selected word history. [Figure 16] An example of the anonymization results of selected words, etc. [Figure 17] An example of the screen that appears when you select the word "depression" as the word you want to anonymize in a consultation record in which only proper nouns have been anonymized. [Figure 18] An example of the screen that appears when "developmental disorder" is selected as the word to be anonymized in a consultation record in which only proper nouns have been anonymized. [Figure 19] An example of a screen displaying previously selected words. [Figure 20] An example of a screen requesting whether or not synonyms should be anonymized. [Figure 21] An example of a screen requesting whether or not to anonymize co-occurring words. [Figure 22] An example of a screen requesting whether or not previously selected words should be anonymized. [Diagram 23] An example screen showing anonymization results. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, the embodiments will be described with reference to the drawings. In the following examples, anonymizing text data refers to processing a sentence describing a specific individual so that it is impossible to infer who the sentence is about. In other words, the data is processed so that it does not directly contain the name of a specific individual, and does not contain any character strings (words, phrases, etc.) that would make it easy to match an individual. In the following embodiments, the anonymization of words and the like refers to processing (substitution, deletion, and the like) the words to conceal them, thereby contributing to the anonymization of text data. The anonymization of text data also takes into account inferences from the context. For example, when anonymizing a word, the system considers not only the word itself but also its combination with other words and its position in the sentence to determine whether to conceal the word. On the other hand, when anonymizing text data, it is necessary to preserve as much of the content of the text as possible. If too many words are concealed, the content of the text itself will be lost, so it is necessary to determine the words to be concealed so that the content of the text can be understood while avoiding personal inferences. EXAMPLES

[0012] 1 is a diagram showing an overall configuration of a system according to the present embodiment. The system shown in FIG. The text anonymization device 100 includes an I / O interface 101 , a memory 102 , a communication device 103 , a CPU (Central Processing Unit) 104 , and an auxiliary storage device 105 . The CPU 104 loads a program stored in the auxiliary storage device 105 into the memory 102 to perform overall control of the system itself, and also performs various types of judgment, calculation and control processing. The I / O interface 101 connects to input and output devices. The communication device 103 is connected to a network 140 and communicates with other devices.

[0013] The auxiliary storage device 105 stores a basic information database 106 , a record database 107 , a selected word history database 108 , a synonym database 109 , and a co-occurrence word database 110 . The auxiliary storage device 105 also stores programs that realize a user portal function 111, a synonym location specification function 112, a co-occurring word location specification function 113, a selected word specification function 114, an anonymization function 115, and a proper noun anonymization function 116.

[0014] The user client terminal 120 includes an I / O interface 121 , a memory 122 , a communication device 123 , a CPU 124 , and an auxiliary storage device 125 . The CPU 124 performs various processes by loading programs stored in the auxiliary storage device 125 into the memory 122 and executing them. The I / O interface 121 connects to input and output devices. The communication device 123 is connected to a network 140 and communicates with other devices.

[0015] The CPU 104 and the CPU 124 are arithmetic units that execute programs loaded into the memory 102 and the memory 122 to operate as corresponding functional units. The arithmetic units may include one or more processor cores, CPUs, or GPUs (Graphics Processing Units). The processes and calculations described below are executed by the CPU 104 or 124.

[0016] The user client terminal 120 performs operations such as logging in to the text anonymization device 100, inputting the entire text data to be anonymized, inputting words to be anonymized in the text data, and displaying related sentences and anonymization records received from the text anonymization device 100. The user client terminal 120 is operated by a user who receives consultations from clients, creates records, and inputs them.

[0017] The text anonymization device 100 receives text data from the user client terminal 120 and words and related sentences selected by the user to be anonymized via the user portal function 111, and performs anonymization. Details will be described later.

[0018] 2 is an explanatory diagram of a basic information table stored in the basic information database 106. In the basic information table, the client's name, date of birth, sex, place of residence, etc. are linked to the client's ID. For example, the client ID: aa is linked to name: A, date of birth: 19501020, gender: F, and place of residence: Kokubunji City, Tokyo.

[0019] 3 is an explanatory diagram of a consultation record table stored in the record database 107. The consultation record table associates consultation records created by users with the client ID who consulted, the consultation date, the record ID, and the record date. For example, the following consultation record is linked to client ID: aa, record ID: 2, along with consultation date: 20221220. "We responded to the second consultation. It appears that the reason for the family's financial difficulties is the need to care for the family's father, who collapsed due to a myocardial infarction. Currently, the family's grandparents, who live in the neighboring city of Higashi-Koganei, are caring for the family's father and taking care of his personal needs. In addition, Mr. A himself has tendencies toward developmental disorders. There is also a possibility that she may have a mild intellectual disability, which is thought to be contributing to her difficulty understanding herself. She appears to have a tendency towards anxiety and depression, but is not currently receiving medical treatment. Furthermore, it appears that she developed schizophrenia in April of last year, and was hospitalized at Sakura Municipal Hospital in Tokyo until around summer. We plan to have her come with her mother next time to discuss applying for welfare assistance."

[0020] 4 is an explanatory diagram of a selected word history table stored in the selected word history database 108. The selected word history table associates words that the user wishes to anonymize in step S105, which will be described later, with the client ID, the last selection date and time of the word, the selection frequency, and the selected word ID. In step S105, which will be described later, if there is a word that the user wants to anonymize, the selected word history table is updated. At that time, if the word is already written in the selected word history table, a common selected word ID is assigned to the word. If the word is not written in the selected word history table, a new selected word ID is assigned and stored.

[0021] 5 is an explanatory diagram of a synonym master table stored in the synonym database 109. The synonym master table associates synonyms of predefined words with synonym IDs. In step S107 of "extracting synonyms" described later, synonyms for the selected word to be anonymized are extracted from the synonym database 109. Alternatively, synonyms may be identified using the publicly available synonym database 109. Details will be described in the explanation of the processing flow.

[0022] 6 is an explanatory diagram of a co-occurrence word master table stored in the co-occurrence word database 110. The co-occurrence word master table associates co-occurrence words with co-occurrence word IDs for pre-defined words. In step S108 described later, co-occurring words of the selected word to be anonymized are extracted from the co-occurring word database 110. Alternatively, co-occurring words may be specified using a publicly available co-occurring word database. Alternatively, co-occurring words of a specific word may be searched for from past consultation records. Details will be described in the explanation of the processing flow.

[0023] FIG. 7 is a process flow illustrating the operation of the text anonymization device 100. As a premise, in this embodiment, an anonymization process performed by the text anonymization device 100 on text data will be described using a consultation record of a consultation for people in financial difficulty as an example. Here, a person who has a problem and seeks advice is called a client, and a person who receives the advice and creates a consultation record is called a recorder. The recorder becomes a user and creates a consultation record, and inputs the created consultation record into the user client terminal 120. However, the text data to be anonymized by the text anonymization device 100 is not limited to consultation records as in the present embodiment. For example, medical charts, nursing records in welfare, questionnaire data, etc. may be the text data to be anonymized by this device.

[0024] Step S101: Acquire basic information about the client The user uses the user client terminal 120 to input basic information of the client, including the client ID, name, date of birth, sex, and place of residence. The inputted basic information of the client is transmitted to the text anonymization device 100 via the network 140. The transmitted client basic information is received by the user portal function 111 in the text anonymization device 100, which generates a basic information table and stores it in the basic information database 106.

[0025] Step S102: Registration of consultation records After step S101, the user uses the user client terminal 120 to input a consultation record. The input consultation record is transmitted to the text anonymization device 100 via the network 140.

[0026] Step S118: Request for anonymization of proper nouns After step S102, the user client terminal 120 requests the text anonymization device 100 to anonymize proper nouns included in the consultation record.

[0027] Step S103: Anonymizing proper nouns The proper noun anonymization function 116 of the text anonymization device 100 receives the consultation record transmitted from the user client terminal 120, and then anonymizes only the proper nouns in response to the request in S118. The transmitted consultation record is received by the user portal function 111, which generates a consultation record table by linking it to the client ID and record ID, and stores it in the record database 107. FIG. 3 is an example of the generated consultation record table. In the following, the anonymization process performed by the text anonymization device 100 will be described for the consultation record with client ID=aa and record ID=2. Fig. 8 shows an example of the anonymization result of proper nouns in the consultation record with client ID=aa and record ID=2. A known method may be used for the anonymization. For example, as in the example of Fig. 8, the anonymization may be performed by replacing the proper nouns with *, or the anonymization process may be performed by deleting, replacing, aggregating, and abstracting the proper nouns.

[0028] Step S104: Extract previously selected words After step S103, the text anonymization device 100 refers to the selected word history table stored in the selected word history database 108 to extract previously selected words.

[0029] Step S119: Send the anonymization results of proper nouns and previously selected words After step S104, the text anonymization device 100 transmits to the user client terminal 120 the consultation record result in which the proper nouns have been anonymized in step S103 and the previously selected words extracted in step S104.

[0030] Step S105: Selection of words to be anonymized After step S119, the user client terminal 120 uses morphological analysis technology or the like to break down each sentence in the consultation record results in which proper nouns have been anonymized into words or parts of speech, displays them, and accepts from the user the selection of words that the user wishes to anonymize. Fig. 17 is an example of a screen when the user selects "depression" as a word that the user wants to anonymize in a consultation record in which only proper nouns have been anonymized. As shown in Fig. 17, words that the user wants to anonymize may be displayed with shading. In the example screen of Fig. 17, if the user wishes to anonymize the word "depression," for example, the user selects "depression" in the displayed consultation record and then selects the icon marked "anonymize." If the user wishes to anonymize not only "depression" but also synonyms of "depression," the user may be able to select an icon marked "synonyms." If the user wishes to anonymize not only "depression" but also co-occurring words with "depression," the user selects the icon marked "co-occurring words." Figure 18 shows an example of the screen when "developmental disorder" is selected as the word to be anonymized in a consultation record where only proper nouns have been anonymized. If you want to anonymize synonyms and co-occurring words of "developmental disorder," select the "synonyms" or "co-occurring words" icon.

[0031] Step S106: Displaying previously selected words After step S105, the user client terminal 120 displays the previously selected words extracted in step S104. 19 is an example of a screen displaying previously selected words. Of the previously selected words displayed, words that the user selects as words that he or she wishes to anonymize may be displayed with shading.

[0032] As shown in FIG. 19, if one wishes to anonymize the previously selected word "father" in a consultation record in which proper nouns have been anonymized, one selects "father" and then selects the "anonymize" icon. If one wishes to anonymize synonyms of "father", one may select "father" and then select the "synonym" icon. If one wishes to anonymize co-occurring words of "father", one may select "father" and then select the "co-occurring words" icon. In the subsequent processing in this embodiment, a case will be described in which "father" is selected from the previously selected words, and then the "anonymize icon" is selected.

[0033] The timing for displaying the previously selected words on the user client terminal 120 may be, for example, the timing when the "selection history tab" in FIG. 19 is specified. As the past selected words, a similarity between the client and the past client regarding the consultation record to be anonymized may be calculated, and the desired anonymization and selected words may be displayed for the consultation record of the similar past client. For example, the similarity may be calculated based on the age, sex, and place of residence in the basic information table as shown in FIG. 2, from the match of age between client ID=aa and a client associated with another client ID, or the match of classification of the place of residence such as region, prefecture, city, town, and village. The determination of the match when calculating the similarity may be determined depending on the method used. For example, when using cosine similarity, it may be determined that the attribute similarity matches when the value of cosine similarity is 0.5 or more, or a threshold value other than 0.5 may be used.

[0034] When displaying the past selected words, the words may be displayed in order of importance according to the importance assigned to each selected word. The importance may be preset for each word. For example, the frequency of selection in the selected word history table may be used as an index of importance, as shown in FIG. 4, and the past selected words may be displayed in order of the frequency of selection. A specific importance index may be set for each selected word instead of the frequency of selection. The importance index to be referenced may be an importance index recorded in the past selected word history table for consultation records of past clients similar to the client (ID=aa).

[0035] Step S120: Request for related words In this embodiment, a case will be described in which "depression" is selected in step S105 as shown in FIG. 17, and then "synonyms" are selected, and "developmental disorder" is selected in step S105 as shown in FIG. 18, and then "co-occurring words" are selected, and then "father" is selected in step S106 as shown in FIG. 19, and then "anonymization" is selected. In this case, for example, as shown in FIG. 9, the selected word and the appearance position of the word in the consultation record to be anonymized may be converted into data together with whether or not anonymization of each word is selected, whether or not synonyms are selected, and whether or not co-occurring words are selected, and the data may be transmitted to the text anonymization device 100 to request related words. Note that the appearance position of the selected word may be expressed as the number of characters from which the selected word is written in the consultation record to be anonymized (client ID=aa, record ID=2), for example, as shown in FIG. 9. However, as in the example of Fig. 9, the selected word "father" was selected in step S106 from the words presented on the screen example as in Fig. 17 based on the selected word history table, and therefore its appearance position may be undefined at the time of step S120. When the user selects a word to be anonymized from the previously selected words presented in step S106 according to the selected word history table, the appearance position is identified in step S109.

[0036] Step S107: Updating the selected word history The text anonymization device 100 updates the selected word history in accordance with a request for related words transmitted from the user client terminal 120 (step S120). Figure 15 is an example of updating the selected word history in accordance with a related word request, for example, as in Figure 9. Specifically, "depression" and "developmental disorder," which are included in Figure 9 but not listed in the selected word history in Figure 4, are linked to new selected word IDs and added to the selected word history table. The appearance position of the selected word may be expressed as the number of characters from which the selected word appears in the consultation record to be anonymized (client ID=aa, record ID=2), for example, as in Figure 9.

[0037] Step S108: Identifying the location of synonyms The text anonymization device 100, in response to a request for related words transmitted from the user client terminal 120 (step S120), identifies synonyms of the selected word and their positions of appearance from the consultation record to be anonymized. For example, when a synonym for the word "depression" is requested as shown in Fig. 17, the synonym location identifying function 112 identifies synonyms for the word "depression" from the synonym master table stored in the synonym database 109. Next, the synonym location identifying function identifies the location where the identified synonym for the word "depression" appears in the consultation record to be anonymized.

[0038] The method of identifying synonyms may be to use a synonym master table stored in a synonym database 109 as shown in FIG. 5, or to use a publicly available thesaurus, or to search for synonyms in an external database to obtain synonyms of the selected word.

[0039] Fig. 12 shows an example in which, when the consultation record to be anonymized has client ID=aa and record ID=2 and the word "depression" for which a synonym is requested is selected, the appearance position of "schizophrenia" in the consultation record is specified as a synonym listed in the synonym master table for the selected word "depression." The appearance position of the selected word may be expressed as the number of characters from which the selected word is written in the consultation record to be anonymized, as shown in Fig. 9.

[0040] Step S109: Identifying the location of co-occurring words The text anonymization device 100, in response to a request for related words transmitted from the user client terminal 120 (step S120), identifies co-occurring words of the selected word and their occurrence positions from the consultation record to be anonymized.

[0041] For example, when co-occurrence words of the word "developmental disorder" are requested as shown in Fig. 18, the co-occurrence word position identifying function 113 identifies co-occurrence words of the word "developmental disorder" from the co-occurrence word master table stored in the co-occurrence word database 110. The co-occurrence word position identifying function 113 identifies the positions where the co-occurrence words of the identified word "developmental disorder" appear in the consultation record to be anonymized.

[0042] Co-occurrence words may be identified using a co-occurrence word master table stored in a co-occurrence word database 110 as shown in FIG. 6, or a publicly available co-occurrence word dictionary may be used, or co-occurrence words may be obtained by searching for co-occurrence words in an external database.

[0043] Fig. 13 shows an example in which, when the consultation record to be anonymized has client ID=aa and record ID=2 and the word "developmental disorder" that requests co-occurrence words is selected, the appearance position of "intellectual disability" in the consultation record is specified as a co-occurrence word in the co-occurrence word master table of the selected word "developmental disorder." The appearance position may be expressed as the number of characters from which the word is written in the consultation record to be anonymized.

[0044] Step S110: Identifying the location where the anonymized selected word appears In response to a request for related words sent from the user client terminal 120 (step S120), the text anonymization device 100 identifies the positions at which words selected as words to be anonymized from the history of past selected words appear in the consultation record to be anonymized. For example, when "father" is selected from among previously selected words as shown in FIG. 9, the selected word identification function 114 identifies the position where "father" appears in the consultation record to be anonymized.

[0045] Fig. 14 shows an example of identifying the appearance position of the selected word "father" in the consultation record to be anonymized when the consultation record to be anonymized has client ID=aa and record ID=2 and "father" has been selected as a previously selected word. Note that the appearance position may be expressed as the number of characters from which the word is written in the consultation record to be anonymized.

[0046] Step S121: Request whether related words need to be anonymous The text anonymization device 100 transmits to the user client terminal 120 the appearance positions of synonyms of the selected word identified in step S107, the appearance positions of co-occurring words of the selected word identified in step S108, and the appearance positions of previously selected words identified in step S109. For example, data such as that shown in Figure 12 regarding extracted synonyms and their positions of appearance, data such as that shown in Figure 13 regarding extracted co-occurring words and their positions of appearance, and data such as that shown in Figure 14 regarding extracted previously selected words and their positions of appearance may be sent to the user client terminal 120.

[0047] Step S111: Selection of whether or not to keep synonyms anonymous The user client terminal 120 receives the request in step S121, and based on the synonyms of the selected word and their positions, displays the synonyms of the selected word and their positions in the consultation record in which only the proper nouns have been anonymized, and accepts from the user the choice of whether or not to anonymize.

[0048] Fig. 20 is an example of a screen that requests the user client terminal 120 to display synonyms of the selected word and whether or not they should be anonymized, based on the synonyms and their positions of appearance. For example, as in the example screen of Fig. 20, the synonym "schizophrenia" identified in step S108 for the selected word "depression" and its positions of appearance may be displayed, and the user may select whether or not it should be anonymized. The positions of appearance of the synonym "schizophrenia" may be displayed with shading, as in Fig. 20. In this example, if one wishes to anonymize "schizophrenia", one selects the "anonymization icon". If one wishes to anonymize synonyms of "schizophrenia" again, one selects the "synonym icon", and if one wishes to anonymize co-occurring words, one selects the "co-occurring word icon". In the subsequent processing in this embodiment, a case will be described in which the "anonymization icon" is selected for "schizophrenia", which is displayed as a synonym of "depression".

[0049] Step S112: Selection of whether or not to anonymize co-occurring words The user client terminal 120 receives the request in step S121, and based on the co-occurring words of the selected word and their positions, displays the co-occurring words of the selected word and their positions in the consultation record in which only the proper names have been anonymized, and accepts from the user the choice of whether or not to anonymize.

[0050] Fig. 21 is an example of a screen for requesting whether or not to anonymize the co-occurring words displayed by the user client terminal 120. For example, as in the example screen of Fig. 21, the co-occurring word "intellectual disability" identified in step S109 with respect to the selected word "developmental disability" and its occurrence position may be displayed, and the user may select whether or not to anonymize it. In this example, if one wishes to anonymize "intellectual disability", one selects the "anonymization icon". If one wishes to anonymize synonyms of "intellectual disability", one selects the "synonym icon", and if one wishes to anonymize co-occurring words, one selects the "co-occurring word icon". In the subsequent processing in this embodiment, a case will be described in which the "anonymization icon" is selected for "intellectual disability", which is displayed as a co-occurring word with "developmental disability".

[0051] Step S113: Selection of whether previously selected words should be kept anonymous The user client terminal 120 receives the request in step S121, displays the previously selected words based on the previously selected words and their appearance positions, and accepts a selection from the user as to whether or not the previously selected words need to be anonymized. Fig. 22 is an example of a screen for requesting whether or not to anonymize the previously selected word displayed by the user client terminal 120. For example, as in the example of the screen in Fig. 22, the appearance position of the previously selected word "father" identified in step S110 may be displayed, and the user may select whether or not to anonymize it. In this example, if one wishes to anonymize "father", one selects the "anonymization icon". If one wishes to anonymize synonyms of "father" again, one selects the "synonym icon", and if one wishes to anonymize co-occurring words, one selects the "co-occurring word icon". In the subsequent processing in this embodiment, a case will be described in which the "anonymization icon" is selected for the displayed "father".

[0052] Step S122: Request for anonymization The user client terminal 120 transmits to the text anonymization device 100 the word to be anonymized selected in step S105, the synonyms to be anonymized selected in step S111, the co-occurring words to be anonymized selected in step S112, and whether or not the previously selected words require anonymity selected in step S113. The timing for making the anonymization request to the text anonymization device 100 may be, for example, when the “Anonymization Result” tab is selected among the “Consultation Record,” “Selection History,” and “Anonymization Result” tabs displayed in the example screen 23. For example, as shown in FIG. 15 , the following may be transmitted to the text anonymization device 100: “depression” and “developmental disorder” selected in step S105, the past word “father” selected in step S106, “schizophrenia”, a synonym of “depression” selected in step S111 as a word to be anonymized, “intellectual disability” selected as a co-occurring word with “developmental disorder” in step S112, “father” selected as a past selected word in step S113, and the positions at which each of the targets for anonymization appears in the consultation record, as well as whether or not anonymization was selected, whether or not a synonym was selected, and whether or not a co-occurring word was selected.

[0053] Step S114: Updating the selected word history The text anonymization device 100 newly registers in the selected word history table the synonyms and co-occurring words to be anonymized requested by the user based on a request for anonymization from the user client terminal 120 (step S122), i.e., based on whether anonymity is required for the synonyms selected in step S111, whether anonymity is required for the co-occurring words selected in step S112, and whether anonymity is required for the previously selected words selected in step S113. 10 and 11 show examples of a selected word history table updated by linking "schizophrenia," which is displayed as a synonym of "depression" and selected as anonymization target, and "developmental disorder," which is displayed as a co-occurrence word of "intellectual disability" and selected as anonymization target, to the client ID and selected word ID. The updated selected word history table is stored in the selected word history database 108.

[0054] Step S115: Anonymous processing The anonymization function 115 of the text anonymization device 100 anonymizes the consultation record in which only the proper nouns have been anonymized in step S103, based on a request for anonymization from the user client terminal 120 (step S122). For example, Fig. 16 shows an example of anonymization of a consultation record in which only proper names have been anonymized as in Fig. 8, in accordance with an anonymization request as in Fig. 15 received by the text anonymization device 100. In this embodiment, when "depression," "developmental disorder," "father," "schizophrenia," and "intellectual disability" are selected as the words desired for anonymization as in Fig. 15, the words selected as the words desired for anonymization included in the consultation record to be anonymized, client ID=aa and record ID=2, are replaced with *. The anonymization method is performed according to predetermined anonymization rules, and may involve replacing the character string to be anonymized with a specific character string, as in this embodiment, or may involve deleting, replacing, aggregating, and abstracting the character string to be anonymized.

[0055] Step S116: Sending the anonymization result The text anonymization device 100 transmits the consultation record that has been anonymized in step S115 to the user client terminal 120.

[0056] Step S117: Displaying the anonymization result The user client terminal 120 displays the anonymization result of the consultation record received from the text anonymization device 100. Fig. 23 is an example of the anonymization result of the consultation record as shown in Fig. 16, displayed on the user client terminal 120. As in this example screen, a scroll bar may be provided.

[0057] The anonymization result of the consultation record shown in Fig. 16 may be stored as data separate from the consultation record table shown in Fig. 3. For example, the data of the anonymization result can be stored in a server separate from the text anonymization device 100 and provided so as to be accessible to persons other than the person who recorded the text. In addition, access permissions may be set for the same data, with an authority level allowing access to the consultation records, which are the original data, and an authority level allowing access only in the anonymized state.

[0058] As described above, the text anonymization device 100 disclosed in the embodiments includes a memory unit (auxiliary storage device 105) that stores text data to be processed, and a processing unit (CPU 104) that anonymizes the text data to be processed by processing the text data to conceal a predetermined character string contained in the text data. When the processing unit receives a selection from a user about a character string to be concealed, it presents a character string related to the character string selected to be concealed, receives a selection as to whether the related character string needs to be concealed, and anonymizes the text data based on the character string selected to be concealed and the selection result of whether the related character string needs to be concealed. As an example, when the processing unit receives a user's selection of a word to be concealed, it presents synonyms and / or co-occurring words of the word selected to be concealed as related words, receives a selection as to whether the related words need to be concealed, and anonymizes the text data based on the word selected to be concealed and the selection result as to whether the related words need to be concealed. With this configuration and operation, flexible and efficient anonymization can be achieved according to the text data to be processed. It also reduces the time users spend searching for words they want to keep secret from within text data, shortening the time it takes to anonymize text data.

[0059] As an example, the text data to be processed is a sentence written about a specific individual, and the processing unit presents, as a proven string, a character string that was to be concealed when other text data about the individual was anonymized, accepts a selection as to whether or not the proven string needs to be concealed, and anonymizes the text data based on the selection result as to whether or not the related string needs to be concealed. Therefore, it is possible to efficiently perform anonymization by using the anonymization results of text data previously created for the same client.

[0060] As an example, the text data to be processed is a sentence written about a specific individual, and the processing unit presents, as a proven string, a character string that has been subject to concealment when anonymizing text data about other individuals similar to the individual, accepts a selection as to whether or not to conceal the proven string, and anonymizes the text data based on the selection result as to whether or not to conceal the related string. Therefore, anonymization can be performed efficiently by using the anonymization results of text data previously created for similar clients.

[0061] As an example, the text data to be processed is a sentence written about a specific individual, and the memory unit stores information that identifies the specific individual, a proven string that is to be kept secret, and the importance of the proven string in association with each other, and when presenting the proven string as a candidate for defendant, the processing unit prioritizes the proven string with the higher importance. For example, the importance level is increased as the number of times a character string has been selected as a target for anonymity in the past increases. This makes it possible to effectively utilize past performance and efficiently select words that should be kept secret.

[0062] The processing unit also identifies a position of the related character string in the text data, presents the related character string together with the identified position, and accepts a selection as to whether or not the related character string needs to be concealed. This makes it easy to check the relationships between words in the context, and ensures reliable anonymity while preserving the content of the sentence.

[0063] The processing unit also performs processing to conceal predetermined proper nouns, and then receives a selection from a user regarding the character string to be concealed. By pre-processing words that can be automatically anonymized in this way, the workload of users can be significantly reduced.

[0064] The present invention is not limited to the above-mentioned embodiment, and various modifications are included. For example, the above-mentioned embodiment is described in detail to easily explain the present invention, and is not necessarily limited to the embodiment having all the described configurations. Moreover, the present invention is not limited to the deletion of the configurations, and it is also possible to replace or add the configurations. [Explanation of symbols]

[0065] 21: anonymization level adjustment means, 22: proper name anonymization means, 24: document output means, 100: text anonymization device, 101: I / O interface, 102: memory, 103: communication device, 104: CPU, 105: auxiliary storage device, 106: basic information database, 107: record database, 108: selected word history database, 109: synonym database, 110: co-occurrence word database, 111: user portal function, 112: synonym location identification function, 113: co-occurrence word location identification function, 114: selected word identification function, 115: anonymization processing function, 116: proper noun anonymization function, 120: user client terminal, 121: I / O interface, 122: memory, 123: communication device, 124: CPU, 125: auxiliary storage device, 140: network

Claims

1. A storage unit for storing text data to be processed; a processing unit for anonymizing the text data by performing processing to conceal a predetermined character string included in the text data, The processing unit includes: When a user selects a string to be concealed, Present a character string related to the character string selected to be concealed; Accept a selection as to whether or not the related character string needs to be kept secret; A text anonymization device comprising: a character string selected as an object to be anonymized; and an anonymization result of whether or not the related character string needs to be anonymized.

2. The text anonymization device according to claim 1, The processing unit includes: When a user's selection of words to be concealed is accepted, presenting synonyms and / or co-occurring words of the word selected as the object of concealment as related words; Accepting a selection as to whether or not the related words need to be kept secret; A text anonymization device comprising: a text anonymization device that anonymizes the text data based on a selection result of whether the words selected as targets of anonymization and the related words need to be anonymized or not.

3. The text anonymization device according to claim 1, The text data to be processed is a document describing a specific individual, The processing unit includes: A character string that has a proven track record and that was concealed when other text data about the individual is anonymized is presented as a proven character string; Accepting a selection as to whether or not the achievement character string needs to be kept secret; The text anonymization device anonymizes the text data based on a selection result of whether or not the related character string needs to be concealed.

4. The text anonymization device according to claim 1, The text data to be processed is a document describing a specific individual, The processing unit includes: Presenting a proven character string that has been concealed when anonymizing text data of other individuals similar to the individual as a proven character string; Accepting a selection as to whether or not the achievement character string needs to be kept secret; The text anonymization device anonymizes the text data based on a selection result of whether or not the related character string needs to be concealed.

5. The text anonymization device according to claim 1, The text data to be processed is a document describing a specific individual, The storage unit stores information identifying the specific individual, a record character string that is a character string with a record to be concealed, and an importance of the record character string in association with each other; The text anonymization device is characterized in that, when presenting the track record character strings as candidates for defendant targets, the processing unit prioritizes track record character strings with high importance.

6. The text anonymization device according to claim 5, The text anonymization device is characterized in that the importance level has a higher value for a character string that has been selected as anonymized more frequently in the past.

7. The text anonymization device according to claim 1, The processing unit identifies a position of the related character string in the text data, presents the related character string together with the identified position, and accepts a selection as to whether or not to conceal the related character string.

8. The text anonymization device according to claim 1, The text anonymization device is characterized in that the processing unit performs processing to conceal predetermined proper nouns and then accepts a selection from a user regarding the character string to be concealed.

9. A text anonymization method for anonymizing text data to be processed by processing the text data to be processed to conceal a predetermined character string included in the text data, the method comprising: The text anonymizer Accept the user's selection of the character string to be concealed, Present a character string related to the character string selected to be concealed; Accept a selection as to whether or not the related character string needs to be kept secret; A text anonymization method, comprising: anonymizing the text data based on a selection result of whether the character string selected to be anonymized and the related character string need to be anonymized.

Images