KR102627819B1 - Device, method and computer program for determining personal information in context using artificial intelligence - Google Patents

Abstract

Devices, methods, and computer programs for determining personal information in context using artificial intelligence are provided. The device includes a memory that stores one or more instructions, and one or more processors that execute the one or more instructions stored in the memory, wherein the one or more processors execute the one or more instructions, thereby generating an input containing two or more sentences. Detect an entity name in a text, extract a first sentence and one or more second sentences adjacent to the first sentence containing the detected entity name in the input text, and use a sentence summary neural network model to extract the first sentence and the one or more second sentences. The above second sentence is summarized into one third sentence, and a context judgment neural network model is used to infer whether the third sentence is a sentence related to personal information.

Description

Device, method, and computer program for determining personal information in context using artificial intelligence {DEVICE, METHOD AND COMPUTER PROGRAM FOR DETERMINING PERSONAL INFORMATION IN CONTEXT USING ARTIFICIAL INTELLIGENCE}

The present invention relates to an apparatus, method, and computer program for determining personal information in context using artificial intelligence.

Cases of damage caused by personal information leaks are increasing every year. In the case of online, the speed of spread of leaked personal information can increase the damage to victims of personal information leakage. Reflecting this, regulations are being strengthened, such as enacting and revising laws to protect personal information and establishing measures to ensure the safety of personal information for personal information processors.

Therefore, personal information processors must improve the personal information protection management level and system of their internal systems. For example, overall monitoring should be strengthened, including detection of distribution and misuse of personal information, identification of misuse and abuse of personal information, and detection of personal information viewing. In particular, automated monitoring of access to personal information should be performed to audit personal information distribution.

Recently, personal information is often leaked from websites, portal sites, Internet bulletin boards, online communities, and open chat rooms that are accessible to general users such as institutions, companies, and schools. Personal information processors must be able to collect and analyze information, including personal information, in such spaces to detect personal information in advance and, if necessary, prevent new information, including personal information, from being entered using various application software.

Public Patent Publication No. 10-2021-0049546 (2021.05.06.)

The problem that the present invention seeks to solve is to provide a device, method, and computer program for determining personal information in context using artificial intelligence that can more accurately determine whether personal information exists.

However, the problem to be solved by the present invention is not limited to the problems described above, and other problems may exist.

As a means to solve the above-mentioned problem, a first aspect of the present invention includes a memory storing one or more instructions, and one or more processors executing the one or more instructions stored in the memory, wherein the one or more processors are configured to: Executing one or more instructions to detect an entity name in input text containing two or more sentences, extracting from the input text one or more second sentences adjacent to a first sentence containing the detected entity name, Summarizing the first sentence and the one or more second sentences into one third sentence using a sentence summary neural network model, and inferring whether the third sentence is a sentence related to personal information using a context judgment neural network model. , provides a device for determining personal information in context using artificial intelligence.

In some embodiments, the one or more processors may determine the detected entity name to be personal information if the third sentence is inferred to be a sentence related to personal information.

In addition, the one or more processors extract one or more fourth sentences adjacent to the first sentence within the input text, and detect an identifier corresponding to the personal information within the first sentence and the one or more fourth sentences. And, when the identifier is detected, the detected entity name can be secondarily determined as personal information.

In some embodiments, the one or more processors determine, when the input text is of a first type, the first sentence including the detected entity name within the input text using the context-sensitive neural network model to determine whether the first sentence is related to personal information. You can infer whether it is a sentence or not.

As a means to solve the above-described problem, a second aspect of the present invention includes detecting an entity name in an input text containing two or more sentences, a first sentence containing the detected entity name in the input text. extracting one or more second sentences adjacent to each other, summarizing the first sentence and the one or more second sentences into one third sentence using a sentence summary neural network model, and using a context judgment neural network model Provides a method for determining personal information in context using artificial intelligence, including the step of inferring whether the third sentence is a sentence related to personal information.

In some embodiments, the method may further include determining the detected entity name as personal information if the third sentence is inferred to be a sentence related to personal information.

In addition, the method includes extracting one or more fourth sentences adjacent to the first sentence within the input text, detecting an identifier corresponding to the personal information within the first sentence and the one or more fourth sentences. A step, and when the identifier is detected, may further include the step of secondarily determining the detected entity name as personal information.

In some embodiments, when the input text is of a first type, the method determines whether the first sentence including the detected entity name in the input text is a sentence related to personal information by using the context judgment neural network model. A step of inferring whether or not may be further included may be included.

In some embodiments, the method may further include detecting personal information within the input text containing the two or more sentences using regular expression pattern matching prior to detecting an entity name within the input text. You can.

As a means to solve the above-described problem, a third aspect of the present invention provides a computer program stored in a computer-readable recording medium for combining with a computer and executing the above-described method.

Other specific details of the invention are included in the detailed description and drawings.

According to the present invention described above, the sentence containing the entity name and one or more adjacent sentences are summarized into one summary, and then inferred whether the summary reflecting the context contains personal information, thereby determining whether the entity name is personal information. You can judge more accurately whether it is done or not.

In addition, according to the present invention described above, it is possible to determine more definitively whether the entity name corresponds to personal information by additionally detecting an identifier corresponding to the entity name that is initially determined to be personal information in one or more adjacent sentences.

In addition, according to the present invention described above, the process of generating a single summary sentence for input text corresponding to a predetermined type is omitted, and it is inferred whether a sentence including an entity name in the input text contains personal information. By doing so, it is possible to more efficiently determine whether the entity name corresponds to personal information.

The effects of the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below.

1 is a block diagram of a device according to an embodiment of the present invention.
Figure 2 is a flowchart of a method according to another embodiment of the present invention.
Figure 3 is a diagram to explain the process of entity name detection, sentence summary, and context determination using an artificial neural network model.
FIG. 4 is a diagram illustrating a personal information determination process for an exemplary input text.
Figure 5 is a flowchart of a method for determining personal information by additionally detecting an identifier corresponding to an entity name according to another embodiment of the present invention.
Figure 6 is a flowchart of a method for detecting personal information using both regular expression pattern matching and an artificial neural network model according to another embodiment of the present invention.
Figure 7 is a flowchart of a method for omitting the process of generating one summary for input text corresponding to a predetermined type according to another embodiment of the present invention.

The advantages and features of the present invention and methods for achieving them will become clear by referring to the embodiments described in detail below along with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below and may be implemented in various different forms. The present embodiments are merely provided to ensure that the disclosure of the present invention is complete and to provide a general understanding of the technical field to which the present invention pertains. It is provided to fully inform the skilled person of the scope of the present invention, and the present invention is only defined by the scope of the claims.

The terminology used herein is for describing embodiments and is not intended to limit the invention. As used herein, singular forms also include plural forms, unless specifically stated otherwise in the context. As used in the specification, “comprises” and/or “comprising” does not exclude the presence or addition of one or more other elements in addition to the mentioned elements. Like reference numerals refer to like elements throughout the specification, and “and/or” includes each and every combination of one or more of the referenced elements. Although “first”, “second”, etc. are used to describe various components, these components are of course not limited by these terms. These terms are merely used to distinguish one component from another. Therefore, it goes without saying that the first component mentioned below may also be a second component within the technical spirit of the present invention.

Unless otherwise defined, all terms (including technical and scientific terms) used in this specification may be used with meanings commonly understood by those skilled in the art to which the present invention pertains. Additionally, terms defined in commonly used dictionaries are not interpreted ideally or excessively unless clearly specifically defined.

In describing the present invention, detailed descriptions of related known technologies will be omitted if they are obvious to those skilled in the art and are judged to unnecessarily obscure the gist of the present invention.

“Personal information” refers to information about a living individual that can identify the individual, such as name, resident registration number, or video. Personal information includes information that can be easily combined with other information to identify a specific individual, even if the information alone cannot identify the specific individual. Personal information includes information about objective facts such as name, address, contact information, income, education, grades, occupation, e-mail, video, call content, credit, debt, and Internet access IP, as well as subjective information such as a third party's opinion or evaluation of the individual. Contains information.

“Identifier” refers to a value or name uniquely assigned to an individual or an object related to an individual. For example, identifiers include unique identification information (resident registration number, passport number, etc.), name, detailed address, date information (birthday, anniversary, etc.), phone number, medical record number, bank account number, credit card number, certificate number, and automobile. It may include, but is not limited to, number, email address, etc.

“Sensitive information”, contrary to dictionary or legal definitions, refers to personal information that is located adjacent to an identifier and can identify a specific individual.

“Named Entity” refers to one or more words, words, or morphemes, such as nouns or numerical expressions, that have a specific meaning. For example, an entity name may include, but is not limited to, a person, institution, organization, place, location, animal, plant, product, event, creative work, date, time, currency, ratio, or other quantitative expressions.

Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings.

1 is a block diagram of a device according to an embodiment of the present invention.

Referring to Figure 1, device 100 includes memory 110 and processor 120.

The memory 110 stores various data, programs, or applications for driving and controlling the device 100. A program or application stored in memory 110 includes one or more instructions. A program or application stored in the memory 110 may be executed by the processor 120.

In some embodiments, memory 110 may store one or more data, programs, or applications that make up an artificial neural network. Alternatively, memory 110 may store one or more data, programs, or applications that control an artificial neural network.

The processor 120 executes an operating system (Operation System) and various programs or applications stored in the memory 110. Processor 120 may include one or more processors including single core, dual core, triple core, quad core, and multiple cores thereof. The processor 120 may include a main processor and a subprocessor operating in a sleep mode.

As the processor 120 executes one or more instructions stored in the memory 110, the device 100 can execute methods according to various embodiments, which will be described later.

Device 100 may be implemented in the form of a server.

Alternatively, the device 100 may be a smartphone, tablet PC, PC, smart TV, mobile phone, personal digital assistant (PDA), laptop, media player, micro server, global positioning system (GPS) device, e-book reader, or digital broadcasting terminal. , navigation, kiosks, MP3 players, digital cameras, home appliances, and other mobile or non-mobile computing devices, but are not limited to these. Additionally, the device 100 may be implemented as a wearable device such as a watch, glasses, hair band, or ring equipped with communication and data processing functions.

The device 100 may further include other components not shown in FIG. 1, such as a communication interface, an input device, and a display unit.

Figure 2 is a flowchart of a method according to another embodiment of the present invention.

Referring to FIG. 2, in step S210, the device 100 detects an entity name in input text containing two or more sentences.

The entity name detected within the input text is a noun or numeric expression with a specific meaning, so it is a candidate for personal information that is suspected to be personal information. Personal information can be determined by executing the following steps for a sentence containing an entity name corresponding to a personal information candidate.

In some embodiments, device 100 may receive input text from a user. A user may enter input text into device 100 using an input device or another computing device connected to device 100. Alternatively, the device 100 may directly extract the input text from a random address space that may or may not be specified in advance on the Internet.

In some embodiments, an entity name detection method using a rule-based dictionary may be used to detect entity names. Alternatively, a machine learning-based artificial neural network model can be used to detect entity names. Various known methods can be used for entity name detection.

The input text contains two or more sentences. In principle, the number or length of sentences in the input text is not limited. However, in order to improve the speed or accuracy of entity name detection, the number or length of sentences can be limited to a predetermined size.

Although not clearly shown, step S210 may further include processes such as sentence separation and morpheme analysis.

The device 100 executes the following steps for each entity name detected in the input text.

In step S220, the device 100 extracts one or more second sentences adjacent to the first sentence containing the entity name detected in the input text.

The first sentence refers to a sentence containing the entity name detected in step S210. The second sentence refers to a sentence adjacent to the first sentence. For example, the second sentence may be n sentences before and after the first sentence, but is not limited thereto.

In some embodiments, the two sentences preceding or following the first sentence or the four sentences preceding or following the first sentence may be extracted as the second sentence. Alternatively, a larger number of sentences can be extracted as the second sentence in the same manner. When extracting the second sentence, the number of sentences located before the first sentence may not be the same as the number of sentences located after the first sentence. This takes into account cases where the number of sentences located before the first sentence is relatively insufficient compared to the number of sentences located after the first sentence, or vice versa. Alternatively, when extracting the second sentence, the sentence located before the first sentence or the sentence located after the first sentence may not be extracted at all. This considers the case where the first sentence is the first or last sentence of the input text.

In step S230, the device 100 summarizes the first sentence and the one or more second sentences into one third sentence using an artificial neural network model.

In step S240, the device 100 uses an artificial neural network model to infer whether the third sentence is a sentence related to personal information. “Sentence related to personal information” may mean, for example, a sentence containing personal information.

In step S250, if the third sentence is inferred to be a sentence related to personal information, the device 100 determines the detected entity name to be personal information.

As described above, the entity name is a personal information candidate that is suspected to be personal information, but it is difficult to accurately determine whether or not it is personal information based on the entity name alone without considering the context.

On the other hand, determining whether a sentence containing an entity name contains personal information allows for a relatively accurate judgment due to the large amount of information compared to judging the entity name alone. The method of Figure 2 considers not only the sentence containing the entity name but also surrounding sentences adjacent to the sentence to determine whether the sentence contains personal information, thereby enabling more accurate determination of personal information. In particular, rather than simply using multiple sentences as the unit of personal information judgment to expand the scope of information, it is characterized by determining whether a summary that summarizes multiple sentences contains personal information. As a result, sentences are refined by removing meaningless noisy information, and personal information can be determined more quickly by observing the context and at the same time streamlining the amount of information.

Figure 3 is a diagram to explain the process of entity name detection, sentence summary, and context determination using an artificial neural network model.

Referring to FIG. 3, the device 100 can perform entity name detection, sentence summary, and context determination processes using three artificial neural network models 310, 320, and 330.

The entity name detection neural network model 310 receives input text containing two or more sentences, detects and outputs the entity name within the input text.

In Figure 3, an artificial neural network model is used to detect entity names, but the method is not limited thereto. As described above, an entity name detection method using a rule-based dictionary can be used to detect entity names. In this case, a rule-based entity name detection model can receive input text, detect the entity name within the input text, and output it.

The sentence summary neural network model 320 receives the first sentence including the entity name detected by the entity name detection neural network model 310. Additionally, the sentence summary neural network model 320 receives the first sentence and one or more second sentences adjacent to each other as input. As described above, the second sentence may be n sentences before or after the first sentence, but is not limited thereto. The sentence summary neural network model 320 summarizes the first sentence and the one or more second sentences into one third sentence and outputs it.

In some embodiments, the sentence summary neural network model 320 may be implemented based on a Transformer model. For example, the sentence summary neural network model 320 uses a language model based on a transformer model to embed data in the form of two or more sentences as numerical vectors, and uses the embedded vectors as a summary layer consisting of one or more layers ( It may be implemented as a structure that is input into a summarization layer and summarized as a single sentence, but is not limited to this.

The context judgment neural network model 330 receives the third sentence summarized by the sentence summary neural network model 320. The context judgment neural network model 330 infers whether the third sentence is a sentence related to personal information and outputs an inferred value.

In some embodiments, the context judgment neural network model 330 may also be implemented based on a transformer model. For example, the context judgment neural network model 330 uses a language model based on a transformer model to embed data in the form of a single sentence as a numerical vector, and the embedded vector is a classification layer composed of two layers. It can be implemented as a structure that classifies whether or not personal information is included in the sentence by entering it in , but is not limited to this.

In some embodiments, as described above, a language model other than the Transformer model may be used. A detailed description of the language model, including the known transformer model, is omitted because it may unnecessarily obscure the gist of the present invention.

FIG. 4 is a diagram illustrating a personal information determination process for an exemplary input text.

Referring to FIG. 4 , a series of conversations are input to the entity name detection neural network model 310 as example input text 410 . The input text 410 may include more sentences than the three dialogue sentences shown in FIG. 4. For convenience of explanation, Figure 4 shows only sentences used in the summary.

The entity name detection neural network model 310 can detect “November” and “Younghee” as entity names in the example input text 410, respectively. However, the following process will only explain the follow-up process for “November”.

Next, “Shall we celebrate Younghee’s birthday in advance in November?” is extracted as the first sentence containing the entity name “November” in the input text 410. And one or more second sentences adjacent to the first sentence above are extracted together with "I can't look up BL. Is it because it's mail?" and "I don't know when I'll see it." In the example of FIG. 4, since there is no sentence located before the first sentence, only the sentence located after the first sentence was extracted as the second sentence. And the two sentences located after the first sentence were extracted as the second sentence. However, it is not limited to this, and even in the same example, the second sentence can be extracted in various ways and in various numbers.

The first sentence and one or more second sentences 420 extracted in this way are input to the sentence summary neural network model 320. The sentence summary neural network model 320 combines the first sentence and one or more second sentences 420 into one third sentence 430, such as "I'm going to celebrate Younghee's birthday in November, but I don't know when I'll see it because it's by mail." Summarize.

And when the third sentence 430, which is a summary sentence, is input to the context judgment neural network model 330, the context judgment neural network model 330 infers whether the sentence is related to personal information. For a given example, the context judgment neural network model 330 will output an inference result indicating that it corresponds to personal information.

Figure 5 is a flowchart of a method for determining personal information by additionally detecting an identifier corresponding to an entity name according to another embodiment of the present invention. Redundant description of steps S510 to S540, which are the same as steps S210 to S250 of the method of FIG. 2, will be omitted.

Referring to FIG. 5, in step S550, if the third sentence is inferred to be a sentence related to personal information, the device 100 first determines that the entity name detected in step S510 is personal information. “Making the first judgment” means that the judgment result of step S550 is verified in the subsequent step S570. That is, in step S550, a temporary or provisional judgment is made.

In step S560, the device 100 detects an identifier corresponding to the entity name detected within the input text, that is, an identifier corresponding to the entity name first determined to be personal information.

In some embodiments, device 100 may detect identifiers for all of the input text.

In some embodiments, device 100 may detect identifiers for only a portion of the input text. The device 100 may extract one or more fourth sentences adjacent to the first sentence containing the detected entity name from the input text. The device 100 may extract the fourth sentence in the same or similar manner as the method for extracting the second sentence described above. For example, the range for extracting the fourth sentence may be wider than the range for extracting the second sentence, or vice versa. The number of sentences extracted as the fourth sentence may be greater than the number of sentences extracted as the second sentence, or vice versa. The device 100 may detect an identifier corresponding to the entity name within the first sentence and one or more fourth sentences.

In some embodiments, device 100 may detect an identifier for one or more entity names detected in step S510. When two or more entity names are detected in the input text, an identifier corresponding to the entity name can be detected among one or more entity names other than the entity name subject to personal information determination.

The fact that an entity name and an identifier "correspond" to each other refers to a case where the entity name corresponds to personal information of an individual that can be identified by an identifier. If the entity name and identifier correspond to each other, the entity name is easily combined with the identifier and functions as information that allows identification of a specific individual. Therefore, in this case, the entity name becomes sensitive personal information.

Meanwhile, in step S550, even if the third sentence is inferred to be a sentence not related to personal information, the device 100 detects an identifier corresponding to the entity name detected in the input text. This is to supplementally detect personal information, taking into account that the inference of the human neural network model may be incorrect.

If the identifier is detected in step S570, the device 100 secondarily determines the entity name detected in step S510 as personal information. The secondary judgment in step S570 corresponds to the final and definitive judgment.

When an identifier associated with an entity name is detected, the entity name is determined to be personal information according to the secondary judgment result, even if it is different from the first judgment result.

Figure 6 is a flowchart of a method for detecting personal information using both regular expression pattern matching and an artificial neural network model according to another embodiment of the present invention. Redundant description of steps S630 to S680, which are the same as steps S520 to S570 of the method of FIG. 5, will be omitted.

Referring to FIG. 6, in step S610, before detecting the entity name in the input text, the device 100 detects personal information in the input text including two or more sentences using regular expression pattern matching. Regular expressions can also be called regular expressions.

The device 100 can detect personal information expressed according to a certain rule, format, or pattern within the input text using regular expression pattern matching. For example, personal information that can be detected using regular expression pattern matching includes resident registration number, passport number, name, address, birthday, phone number, bank account number, credit card number, automobile number, and email address, but is limited to this. It doesn't work.

In some embodiments, in step S620, the device 100 may not perform entity name detection for personal information detected using regular expression pattern matching. Alternatively, even if personal information detected using regular expression pattern matching is detected as an entity name, the device 100 may not execute the following steps for generating a summary. When using regular expression pattern matching, personal information can be judged relatively more accurately than when using a context judgment neural network model, so the personal information determination process is not repeated.

The context judgment neural network model has the advantage of being able to detect a wide range of personal information in a free form that does not follow certain rules, which cannot be detected using regular expression pattern matching. In addition, even though it is personal information, it is possible to increase the accuracy of personal information detection for numerical expressions written in random patterns without following rules or formats.

The method of Figure 6 can accurately and completely determine personal information in the input text by complementary use of rule-based regular expression pattern matching and machine learning-based artificial neural network model.

Figure 7 is a flowchart of a method for omitting the process of generating one summary for input text corresponding to a predetermined type according to another embodiment of the present invention. Redundant description of steps that are the same as those of the method in FIG. 7 will be omitted.

In step S730, after detecting the entity name in the input text, the device 100 determines whether the input text is of the first type. For example, the first type refers to general writing other than conversation. Alternatively, the device 100 may evaluate the sentence length or completeness of the input text and classify the type. The device 100 may classify the input text as the first type when the length of the sentence is longer than a predetermined length or the completeness of the sentence is greater than a predetermined value.

If the input text is the first type, the device 100 omits the summary text generation process of steps S740 to S750.

In this case, in step S760, the device 100 uses a context judgment neural network model to infer whether the first sentence including the previously detected entity name is a sentence related to personal information. In other words, since the summary sentence was not created, inference using the context judgment neural network model is not performed on the third sentence, which is the summary sentence.

If the input text is a conversation, especially in the case of an online chat conversation, the sentences are short and contain many emoticons, onomatopoeia, and mimetic words with little meaning, so there is a possibility of inaccurate results if you judge whether or not personal information is included on a sentence-by-sentence basis. high. In this case, it is effective to generate a summary including adjacent sentences through the summary generation process of steps S740 to S750, and determine whether the summary containing the context of the conversation contains personal information. However, generating a summary every time is not efficient in terms of time, cost, and resources, so the method of Figure 7 omits the process of generating a summary if the input text is a general text type with high sentence completeness.

The steps of the methods according to the embodiment of the present invention may be further divided into additional steps or combined into fewer steps, depending on the implementation of the present invention. Additionally, some steps may be omitted or the order between steps may be changed as needed.

The method of determining personal information in context using artificial intelligence according to the embodiment of the present invention described above can be implemented as a computer program (or application) to be executed in combination with a computer as hardware and stored in a computer-readable recording medium. there is.

The above-mentioned program is C, C++, JAVA, Ruby, and It may include code coded in a computer language such as Python machine language. These codes may include functional codes related to functions that define the necessary functions for executing the methods, and include control codes related to execution procedures necessary for the computer's processor to execute the functions according to predetermined procedures. can do. In addition, these codes may further include memory reference-related codes that indicate at which location (address address) in the computer's internal or external memory additional information or media required for the computer's processor to execute the above functions should be referenced. there is. In addition, if the computer's processor needs to communicate with any other remote computer or server to execute the above functions, the code uses the computer's communication module to determine how to communicate with any other remote computer or server. It may further include communication-related codes regarding whether communication should be performed and what information or media should be transmitted and received during communication.

The storage medium refers to a medium that stores data semi-permanently and can be read by a device, rather than a medium that stores data for a short period of time, such as a register, cache, or memory. Specifically, examples of the storage medium include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc., but are not limited thereto. That is, the program may be stored in various recording media on various servers that the computer can access or on various recording media on the user's computer. Additionally, the medium may be distributed to computer systems connected to a network, and computer-readable code may be stored in a distributed manner.

The description of the present invention described above is for illustrative purposes, and those skilled in the art will understand that the present invention can be easily modified into other specific forms without changing the technical idea or essential features of the present invention. will be. Therefore, the embodiments described above should be understood in all respects as illustrative and not restrictive. For example, each component described as unitary may be implemented in a distributed manner, and similarly, components described as distributed may also be implemented in a combined form.

The scope of the present invention is indicated by the claims described below rather than the detailed description above, and all changes or modified forms derived from the meaning and scope of the claims and their equivalent concepts should be construed as being included in the scope of the present invention. do.

Claims (10)

A memory that stores one or more instructions; and
Comprising one or more processors executing the one or more instructions stored in the memory,
The one or more processors execute the one or more instructions
Detect personal information expressed according to predetermined rules, formats, or patterns within input text containing two or more sentences using regular expression pattern matching,
Detecting entity names within the input text containing two or more sentences,
Define the detected entity name within the input text as a personal information candidate,
Extracting one or more second sentences adjacent to a first sentence containing the detected entity name from the input text,
Summarizing the first sentence and the one or more second sentences into one third sentence using a sentence summary neural network model,
Inferring whether the third sentence, which is a summary of the first sentence and the one or more second sentences, is a sentence related to personal information using a context judgment neural network model,
If the third sentence is inferred to be a sentence related to personal information, the entity name detected in the first sentence is determined to be personal information,
Detecting entity names within the input text uses a rule-based dictionary or a machine learning-based entity name detection neural network model,
The sentence summary neural network model uses a language model based on the Transformer model to embed data in the form of two or more sentences into numerical vectors, and the embedded vectors are divided into a summary layer consisting of one or more layers. It is implemented in a structure that is entered into and summarized in one sentence,
The context judgment neural network model uses a language model based on a transformer model to embed data in the form of a single sentence into a numerical vector, and inputs the embedded vector into a classification layer consisting of one or more layers to obtain personal information within the sentence. Characterized by being implemented in a structure that classifies inclusion or non-inclusion,
A device for determining personal information in context using artificial intelligence. delete According to paragraph 1,
The one or more processors:
Extracting one or more fourth sentences adjacent to the first sentence within the input text,
Detecting an identifier corresponding to the personal information within the first sentence and the one or more fourth sentences,
When the identifier is detected, the detected entity name in the first sentence is secondarily determined as personal information,
The entity name corresponds to the personal information of an individual that can be identified by the identifier, and the entity name functions as information that enables identification of a specific individual by combining with the identifier,
A device for determining personal information in context using artificial intelligence. According to paragraph 3,
The one or more processors:
When the input text is the first type,
Inferring whether the first sentence including the detected entity name in the input text is a sentence related to personal information using the context judgment neural network model,
A device for determining personal information in context using artificial intelligence. Detecting personal information expressed according to a predetermined rule, format, or pattern within an input text containing two or more sentences using regular expression pattern matching;
detecting an entity name within the input text containing two or more sentences;
defining the detected entity name within the input text as a personal information candidate;
extracting one or more second sentences adjacent to a first sentence containing the detected entity name from the input text;
summarizing the first sentence and the one or more second sentences into one third sentence using a sentence summarizing neural network model;
inferring whether the third sentence, which is a summary of the first sentence and the one or more second sentences, is a sentence related to personal information using a context judgment neural network model; and
If the third sentence is inferred to be a sentence related to personal information, determining the detected entity name in the first sentence as personal information,
Detecting entity names within the input text uses a rule-based dictionary or a machine learning-based entity name detection neural network model,
The sentence summary neural network model uses a language model based on the Transformer model to embed data in the form of two or more sentences into numerical vectors, and the embedded vectors are divided into a summary layer consisting of one or more layers. It is implemented in a structure that is entered into and summarized in one sentence,
The context judgment neural network model uses a language model based on a transformer model to embed data in the form of a single sentence into a numerical vector, and inputs the embedded vector into a classification layer consisting of one or more layers to obtain personal information within the sentence. Characterized by being implemented in a structure that classifies inclusion or non-inclusion,
A method of determining personal information in context using artificial intelligence. delete According to clause 5,
extracting one or more fourth sentences adjacent to the first sentence within the input text;
detecting an identifier corresponding to the personal information within the first sentence and the one or more fourth sentences; and
When the identifier is detected, further comprising the step of secondarily determining the detected entity name in the first sentence as personal information,
The entity name corresponds to the personal information of an individual that can be identified by the identifier, and the entity name functions as information that enables identification of a specific individual by combining with the identifier,
A method of determining personal information in context using artificial intelligence. In clause 7,
When the input text is the first type,
Further comprising the step of inferring whether the first sentence including the detected entity name in the input text is a sentence related to personal information using the context judgment neural network model,
A method of determining personal information in context using artificial intelligence. delete Combined with a computer,
A computer program stored in a computer-readable recording medium to execute the method of any one of claims 5, 7, or 8.