Legaltech-based personal data de-identification solution providing system

Abstract

A system for providing a legal tech-based personal information de-identification solution is provided, and includes a de-identification service providing server comprising: a user terminal that registers litigation documents and outputs a result in which personal information within the litigation documents is de-identified; a receiving unit that receives litigation documents from the user terminal; an extraction unit that extracts personal information of parties and representatives included in the mandatory entry area within the litigation documents; a party storage unit that stores the personal information of parties and representatives in a party RDB (Relational Database) by mapping it to pre-set item identifiers and identification numbers; a party de-identification unit that replaces the personal information of parties or representatives within the litigation documents by assigning item identifiers and identification numbers when such information is found; a detection unit that detects personal information of the general public excluding parties or representatives within the litigation documents using a pre-established NER (Named-Entity Recognition) model; a general public de-identification unit that replaces the personal information of the general public by mapping it to pre-set item identifiers and identification numbers; and a general public storage unit that stores the personal information of the general public in a general public RDB by mapping it to pre-set item identifiers and identification numbers.

Description

Technology Field

[0001] The present invention relates to a legal tech-based personal information anonymization solution provision system. By detecting, verifying, storing, and anonymizing personal information contained in litigation documents, the system anonymizes all documents across different levels of litigation using the same identifier and enables restoration to the original text when necessary. Background Technology

[0002] Court judgments, as the products of judicial decision-making, possess a strong public character and play a pivotal role in reinforcing legal stability and the accountability of judicial institutions. The disclosure of judgments holds value as important legal resources that not only enhance the transparency of legal application and interpretation but also significantly contribute to legal research, education, and the formation of social trust. In particular, recent advancements in generative artificial intelligence technology have opened up possibilities for utilizing judgment data from new perspectives, providing innovative opportunities for convergence research between law and technology. However, the current judgment disclosure system faces structural limitations. The disclosure rate for civil, criminal, and administrative cases is merely about 45%, and fees incurred for viewing applications, along with processing times of up to one month, severely restrict access to judgments. This stems from the conflict between the protection of personal information and the public's right to know; due to concerns about privacy infringement, the judiciary mandates the anonymization of personal information prior to disclosure, but this process requires substantial human and material resources.

[0003] At this time, de-identification technology for personal information protection was researched and developed. In this regard, prior art Korean Registered Patent No. 10-2791363 (published April 7, 2025) and Korean Registered Patent No. 10-2785954 (published March 25, 2025) disclose a configuration in which a de-identification policy by type of personal information is pre-set for original PDF electronic documents stored in a DB, and when a user requests to view the document, personal information in the electronic document is detected, and real-time de-identification is performed using deletion, masking, and string replacement methods according to the policy, and then provided to the user; and a configuration in which, when a law firm uploads litigation documents, text is extracted using format conversion and intelligent OCR, and an electronic record is automatically produced through LLM-based summarization and vector DB generation, and then cases are searched and recommended using natural language and keywords.

[0004] However, in the former case, only configurations are disclosed where personal information is not detected in non-structured data, such as unstructured data, which can lead directly to personal information leakage; similarly, in the latter case, only the configuration for database construction and search for LLM-based searching is disclosed, but configurations for anonymization to protect personal information are not disclosed. Furthermore, personal information detection and anonymization technologies reveal several limitations as they are not specialized for litigation documents. While pattern definition methods are effective for structured items like resident registration numbers, they lack scalability and comprehensiveness for litigation documents, which frequently contain unstructured personal expressions or addresses with numerous forms such as road names and lot numbers, resulting in missed detections. NER-based machine learning methods have the advantage of capturing various patterns, but they frequently exhibit contextual false positives, such as detecting personal information even in phrases that contain no personal information at all. Although LLM fine-tuning methods offer a relatively high detection rate, they face limitations such as the difficulty in securing large-scale training data, the requirement of significant training time, and the challenge of sufficiently training the expressions unique to litigation documents. Therefore, research and development of systems to detect personal information contained in litigation documents are required. The problem to be solved

[0005] One embodiment of the present invention provides a legal tech-based personal information de-identification solution system that, in order to detect personal information included in litigation documents, extracts personal information of parties and their representatives from mandatory entry areas and stores it in a party RDB (Relational Database), searches for, verifies, and replaces the personal information of parties and their representatives from litigation documents to de-identify the personal information, and de-identifies personal information of the general public other than the parties by detecting and verifying it using a Named-Entity Recognition (NER) model, storing it in a general public RDB, and replacing it. This allows for an increased detection rate even if the personal information is unstructured data, enables verification and de-identification processing for all documents produced as case records—such as cases of different levels of court proceedings and related cases—as well as allows for restoration to the original state if necessary after de-identification. However, the technical problem that this embodiment aims to solve is not limited to the technical problem described above, and other technical problems may exist. means of solving the problem

[0006] As a technical means for achieving the technical task described above, one embodiment of the present invention includes a de-identification service providing server comprising: a user terminal that registers a litigation document and outputs a result in which personal information within the litigation document is de-identified; a receiving unit that receives the litigation document from the user terminal; an extraction unit that extracts personal information of a party and an agent included in the mandatory entry area within the litigation document; a party storage unit that stores the personal information of a party and an agent in a party RDB (Relational Database) by mapping it to a pre-set item identifier and identification number; a party de-identification unit that replaces the personal information of a party or an agent within the litigation document by assigning an item identifier and identification number when the personal information of a party or an agent within the litigation document is searched; a detection unit that detects personal information of a general person excluding a party or an agent within the litigation document using a pre-established NER (Named-Entity Recognition) model; a general person de-identification unit that replaces the personal information of a general person by mapping it to a pre-set item identifier and identification number; and a general person storage unit that stores the personal information of a general person in a general person RDB by mapping it to a pre-set item identifier and identification number. Effects of the invention

[0007] According to any one of the means for solving the problem of the present invention described above, by automatically detecting various forms of personal information, such as names, addresses, resident registration numbers, and phone numbers included in litigation documents, and then verifying them with a human and storing them in a database, consistency is ensured so that even if the same person appears in multiple litigation documents, court levels, or related cases, they can always be de-identified in the same way. Furthermore, it guarantees entity-level interpretability and connectivity between documents, which were impossible with special character masking or single-document unit numbering methods. Since the mapping between item identifiers, identification numbers, and personal information is accumulated on a case-by-case basis during the de-identification process, de-identified documents can be stably recovered during the subsequent record production stage. Additionally, by training a classification model and an NER model based on verified party search history information and general public search history information, the effect of continuously reducing detection omissions and false positives is achieved. Through this structure, it is possible to establish a personal information management system that encompasses the entire case without damaging the meaning of the document while safely protecting personal information during subsequent utilization processes based on legal tech, such as correction, summarization, and search of litigation documents. Brief explanation of the drawing

[0008] FIG. 1 is a diagram illustrating a legal tech-based personal information de-identification solution providing system according to one embodiment of the present invention. Figure 2 is a block diagram illustrating a de-identification service provider server included in the system of Figure 1. FIGS. 3 and 4 are drawings for illustrating an embodiment in which a legal tech-based personal information de-identification solution according to an embodiment of the present invention is implemented. FIG. 5 is a flowchart illustrating a method for providing a legal tech-based personal information de-identification solution according to an embodiment of the present invention. Specific details for implementing the invention

[0009] Embodiments of the present invention are described below with reference to the attached drawings so that those skilled in the art can easily implement the invention. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals.

[0010] Throughout the specification, when a part is described as being "connected" to another part, this includes not only cases where they are "directly connected" but also cases where they are "electrically connected" with other elements interposed between them. Furthermore, when a part is described as "including" a component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components, and it should be understood that this does not preclude the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0011] Terms such as “about,” “substantially,” etc., used throughout the specification, are used to mean at or near the stated value when inherent manufacturing and material tolerances are presented in the stated meaning, and are used to prevent unscrupulous infringers from unfairly exploiting the disclosure in which precise or absolute values ​​are mentioned to aid in understanding the invention. Terms such as “step” or “step of” used throughout the specification of the invention do not mean “step for”.

[0012] In this specification, the term "part" includes a unit realized by hardware, a unit realized by software, and a unit realized using both. Additionally, one unit may be realized using two or more pieces of hardware, and two or more units may be realized by one piece of hardware. Meanwhile, "part" is not limited to software or hardware, and "part" may be configured to reside in an addressable storage medium or configured to run on one or more processors. Accordingly, as an example, "part" includes components such as software components, object-oriented software components, class components, and task components, as well as processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functions provided within the components and "parts" may be combined into a smaller number of components and "parts" or further separated into additional components and "parts." In addition, the components and '~parts' may be implemented to play one or more CPUs within the device or secure multimedia card.

[0013] Some of the operations or functions described herein as being performed by a terminal, device, or device may instead be performed by a server connected to said terminal, device, or device. Likewise, some of the operations or functions described as being performed by a server may also be performed by a terminal, device, or device connected to said server.

[0014] In this specification, some of the operations or functions described as mapping or matching with a terminal may be interpreted as meaning mapping or matching the terminal's unique number or personal identification information, which is the terminal's identifying data.

[0015] The present invention will be described in detail below with reference to the attached drawings.

[0016] FIG. 1 is a diagram illustrating a legal tech-based personal information de-identification solution providing system according to an embodiment of the present invention. Referring to FIG. 1, the legal tech-based personal information de-identification solution providing system (1) may include at least one user terminal (100) and a de-identification service providing server (300). However, since the legal tech-based personal information de-identification solution providing system (1) of FIG. 1 is merely an embodiment of the present invention, the present invention is not to be interpreted as being limited by FIG. 1.

[0017] At this time, each component of FIG. 1 is generally connected through a network (Network, 200). For example, as shown in FIG. 1, at least one user terminal (100) can be connected to a de-identification service provider server (300) through the network (200). And, the de-identification service provider server (300) can be connected to at least one user terminal (100) through the network (200).

[0018] Here, a network refers to a connection structure capable of exchanging information among individual nodes, such as multiple terminals and servers. Examples of such networks include Local Area Networks (LANs), Wide Area Networks (WANs), the World Wide Web (WWW), wired and wireless data networks, telephone networks, and wired and wireless television networks. Examples of wireless data communication networks include, but are not limited to, 3G, 4G, 5G, 3GPP (3rd Generation Partnership Project), 5GPP (5th Generation Partnership Project), 5G NR (New Radio), 6G (6th Generation of Cellular Networks), LTE (Long Term Evolution), WIMAX (World Interoperability for Microwave Access), Wi-Fi, Internet, LAN (Local Area Network), Wireless LAN (Wireless Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), RF (Radio Frequency), Bluetooth network, NFC (Near-Field Communication) network, satellite broadcasting network, analog broadcasting network, DMB (Digital Multimedia Broadcasting) network, etc.

[0019] In the following, the term "at least one" is defined as a term including both singular and plural forms, and it will be obvious that even if the term "at least one" does not exist, each component may exist in a singular or plural form and may mean singular or plural. Furthermore, whether each component is provided in a singular or plural form may be changed according to the embodiment.

[0020] At least one user terminal (100) may be a terminal of a de-identification manager that uploads litigation documents using a web page, app page, program, or application related to a legal tech-based personal information de-identification solution, verifies the results of personal information detection, and outputs the results of personal information de-identification.

[0021] Here, at least one user terminal (100) may be implemented as a computer capable of connecting to a remote server or terminal via a network. Here, the computer may include, for example, a navigation system, a laptop equipped with a web browser, a desktop, a laptop, etc. At this time, at least one user terminal (100) may be implemented as a terminal capable of connecting to a remote server or terminal via a network. At least one user terminal (100) may include all kinds of handheld-based wireless communication devices, such as navigation, PCS (Personal Communication System), GSM (Global System for Mobile communications), PDC (Personal Digital Cellular), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), Wibro (Wireless Broadband Internet) terminal, smartphone, smartpad, tablet PC, etc.

[0022] The de-identification service providing server (300) may be a server that provides a legal tech-based personal information de-identification solution web page, app page, program, or application. Additionally, the de-identification service providing server (300) may be a server that, when a lawsuit document is input from a user terminal (100), extracts parties, representatives, and the general public using a classification model and an NER model, verifies them on the user terminal (100), and stores them in the parties' RDB and the general public's RDB so that they can be used for subsequent recovery or in subsequent court cases. Furthermore, the de-identification service providing server (300) may be a server that, when de-identification is performed after verification, stores the de-identified results in the parties' search history information and the general public's search history information. Here, the de-identification service providing server (300) may be implemented as a computer capable of accessing a remote server or terminal via a network. Here, the computer may include, for example, a laptop, desktop, laptop, etc. equipped with a navigation system and a web browser.

[0023] FIG. 2 is a block diagram for explaining a de-identification service providing server included in the system of FIG. 1, and FIG. 3 and FIG. 4 are drawings for explaining an embodiment in which a legal tech-based personal information de-identification solution according to an embodiment of the present invention is implemented.

[0024] Referring to FIG. 2, the de-identification service providing server (300) may include a reception unit (310), an extraction unit (320), a party storage unit (330), a party de-identification unit (340), a detection unit (350), a general public de-identification unit (360), a general public storage unit (370), a party history management unit (380), a party verification unit (390), a general public history management unit (391), a general public verification unit (393), and a relationship building unit (395).

[0025] When a de-identification service providing server (300) or another server (not shown) operating in conjunction with one embodiment of the present invention transmits a legal tech-based personal information de-identification solution application, program, app page, web page, etc. to at least one user terminal (100), the at least one user terminal (100) may install or open the legal tech-based personal information de-identification solution application, program, app page, web page, etc. Additionally, a service program may be operated on at least one user terminal (100) using a script executed in a web browser. Here, a web browser refers to a program that enables the use of web (WWW: World Wide Web) services and receives and displays hypertext described in HTML (Hyper Text Mark-up Language), and includes, for example, Chrome, Microsoft Edge, Safari, Firefox, Whale, UC Browser, etc. In addition, "application" refers to an application on a terminal, and includes, for example, an app running on a mobile terminal (smartphone).

[0026] Referring to FIG. 2, the receiving unit (310) can receive a lawsuit document from the user terminal (100). The user terminal (100) can register a lawsuit document. At this time, the lawsuit document may be a document for civil, criminal, or administrative cases, but the type is not limited thereto.

[0027] The extraction unit (320) can extract personal information of the parties and their representatives included in the mandatory entry area within the litigation document. At this time, the mandatory entry area may be an area where the personal information of the parties and their representatives is recorded. The mandatory entry area refers to a fixed and standardized section where core information that must be recorded in a litigation document or public document is located, and it is an area where accurate mechanical extraction is possible because the document layout is consistent.

[0028] Mandatory entry area 1. Information on parties such as plaintiff / defendant / applicant / respondent, etc. 2. Representative (Attorney) Information 3. Case Number / Case Name 4. Court Name 5. Date of sentencing 6. Address / Location (if entered in a fixed format)

[0029] This refers to an area that is nearly identical in location and format across all litigation documents and can be definitively extracted using regular expressions or templates. The mandatory entry area is important because it contains essential personal information with very high accuracy; therefore, if the parties' personal information is extracted from this mandatory entry area and stored in a database, it becomes unnecessary to rely on model predictions, unlike the NER model for extracting general public personal information described later. Furthermore, since basic personal information related to a case appears repeatedly in all relevant documents, extracting and storing it in the parties' RDB (described later) makes it easier to distinguish from names and other elements appearing in subsequent NER detection sections, thereby reducing the error rate in identifying personal information.

[0030] The party storage unit (330) can store the personal information of the party and their representative in the party RDB (Relational Database) by mapping it to a pre-set item identifier and identification number. For example, if it is in the form of [Personal Information-Item Identifier-Identification Number], it can be recorded as [Hong Gil-dong-Name-1]. Once stored in the party RDB in this way, the name [Hong Gil-dong] is subsequently replaced in the form of [Name_1], that is, [Item Identifier_Identification Number]. At this time, a verification step is performed at the user terminal (100) before being stored in the party RDB, so that mechanical errors are reduced and errors do not occur in the party RDB, which serves as the source for replacement. The data stored in this way serves as the same standard data for replacement even if the level of appeal changes or litigation documents are added or modified later, thereby performing the role of ensuring that the same name is subsequently replaced in the same form ([Item Identifier_Identification Number]). Through this, the de-identified personal information can be re-identified.

[0031] Token Original_Value Actor_Type Case_ID Item Identifier_Identification Number personal information Types of parties Case number Name_1 Kim Young-hee manuscript 2025Gahap12345

[0032] The party de-identification unit (340) can replace personal information of a party or representative within a litigation document by assigning an item identifier and an identification number when such information is found. For example, let's assume that there is a term [Kim Young-hee (Plaintiff 1)] in Document A, [Young-hee] in Document B, [Mr. Kim] in Document C, and [Daughter of Plaintiff 2] in Document D. In this case, if [Kim Young-hee = Name_1] is not stored in the party DB, a different token is attached to each litigation document, and the de-identification result appears differently for each document, so that one person appears as multiple people throughout the entire document. Therefore, it must be stored as [Kim Young-hee → [Name_1]] so that de-identification can be performed identically in all documents. In this case, when [Kim Young-hee] is mapped to [Mr. A_1], whenever personal information [Kim Young-hee] appears, it is replaced with [Mr. A_1], and people viewing the litigation document or trying to understand it can accurately understand the case without confusion. Even if you try to summarize or search for litigation documents using LLM, LLM will be able to search for documents and provide results without confusing each party or the general public.

[0033] The detection unit (350) can detect personal information of ordinary people, excluding parties or representatives within the litigation document, using a pre-established Named-Entity Recognition (NER) model. For example, general personal information candidates other than basic personal information, such as names of excluded persons, witnesses, and institutions, can be automatically detected by running a Korean NER model (KoELECTRA, KoBERT, KLUE-BERT, etc.) on a Hugging Face Transformers-based inference server (FastAPI or Spring), and simultaneously perform supplementary detection using a regular expression rule set (rule file + Admin UI management). The NER model results and regular expression detection results are integrated in a scoring logic or rule engine to generate a final candidate list, and this integration process can be executed on a GPU or CPU node in a Docker / Kubernetes-based serving environment.

[0034] For the generated candidates, the user terminal (100) checks whether they are personal information, their type, and their relationship in the verification UI. Only verified information is stored in the [General Public Search History Information DB], and candidates confirmed not to be personal information are stored in the [Detection Exclusion / Learning DB] for future model improvement. A token is assigned to the verified personal information to be used consistently throughout the entire incident, and the de-identification processor replaces the corresponding string within the document with the token. This replacement process can be performed in a Python text pipeline or a Java de-identification module. A mapping table between the token and the original personal information is stored in the [General Public RDB], enabling accurate regeneration of the original text upon future recovery.

[0035] The general public de-identification unit (360) can replace the personal information of the general public by mapping it to a pre-set item identifier and identification number. The user terminal (100) can output a result in which the personal information in the litigation document has been de-identified.

[0036] The general public storage unit (370) can store the general public's personal information in the general public RDB by mapping it to a pre-set item identifier and identification number. By storing it in the general public RDB in this way, it is possible to revert the de-identified litigation document when a need arises to revert it later, or to ensure that the same de-identification process is applied to litigation documents of all levels for the same person, thereby preventing confusion when reading the de-identified litigation document.

[0037] <Example>

[0038] - Case No.: 2025Gadan12345 - Plaintiff: Lee Min-soo - Defendant: Kim Young-soo - Plaintiff's Daughter: Lee Ji-eun (Witness) - Address: 123, △△-ro, ○○-gu, Seoul - Contact: 010-1234-5678

[0039] Let us assume that for the events in Table 3, there exists a page like Table 4.

[0040] 1. Parties - Plaintiff: Lee Min-soo (Address: 123, △△-ro, ○○-gu, Seoul, Tel: 010-1234-5678) - Defendant: Kim Young-soo (Address: 999, △△-ro, ○○-gu, Seoul) 2. Claim and Grounds Plaintiff Lee Min-soo entered into a loan agreement with Defendant Kim Young-soo around March 1, 2023. Plaintiff's daughter, Lee Ji-eun, was present at the scene at the time the contract was signed.

[0041] The process for handling this may be as shown in Table 5 below.

[0042] process explanation Step 1: Uploading and Pre-processing Litigation Documents - The user uploads a litigation document PDF to the system. - The system converts PDF to text, cleans up line breaks and special characters, and performs basic spelling and spacing corrections. - Result: Machine-readable [refined text] is prepared. Step 2: Extraction of Party Information (Basic Personal Information) - First, read the party blocks from the refined text - Apply regular expressions and pattern analysis to the "Plaintiff:" and "Defendant:" lines to extract party information division name address number manuscript Lee Min-soo 123, △△-ro, ○○-gu, Seoul 010-1234-5678 defendant Kim Young-soo 999, △△-ro, ○○-gu, Seoul (doesn't exist) Step 3: Store personal information based on the party's RDB After verification via the user terminal, once verification is complete, it is stored in the party's RDB. Step 4: De-identification of the body of litigation documents (Primary: Based on the parties) While scanning the entire text of the litigation document, -"Lee Min-soo" → [Name_Plaintiff 1]-"Kim Young-soo" → [Name_Defendant 1]-"123, △△ Road, ○○ District, Seoul" → [Address_Plaintiff 1]-"010-1234-5678" → [Number_Plaintiff 1] were replaced. 1. Parties Plaintiff: [Name_Plaintiff 1] (Address: [Address_Plaintiff 1], Phone: [Number_Plaintiff 1]) Defendant: [Name_Defendant 1] (Address: [Address_Defendant 1]) 2. Claim and Grounds Plaintiff [Name_Plaintiff 1] entered into a loan agreement with Defendant [Name_Defendant 1] around March 1, 2023. Plaintiff's daughter, Lee Ji-eun, was present at the scene at the time the contract was concluded. Up to this point, Lee Ji-eun remains an "ordinary person" who still needs to be processed in the NER / relationship extraction stage. Step 5: Detection of General Persons (NER + Relationships) and Additional De-identification - Run NER / Relationship model on a sentence- "The plaintiff's daughter, Lee Ji-eun, ... attended." → Recognize Lee Ji-eun as a PERSON → Extract relationship based on the expression "plaintiff's daughter" → Determine Lee Ji-eun as "Plaintiff 1's child" Add the following to the general public RDB target_id case_id type original_value token T6 2025Gadan12345 Name_Witness_Plaintiff 1's Child Lee Ji-eun [Name_Witness_Plaintiff 1_Child 1] Now, replace "Lee Ji-eun" in the text with [Name_Witness_Plaintiff1_Child1]. De-identification Result (Updated) Plaintiff's daughter [Name_Witness_Plaintiff 1_Child 1] was present at the scene at the time the contract was concluded. Step 6: Reuse in levels, related cases, and subsequent documents - Assume that this case later proceeds to the appellate court (High Court), generating dozens of additional documents such as preparatory briefs, opinions, and summaries of arguments. 1. Query the basic personal information database using the case ID. 2. Match candidate names / relationships. 3. Verify once manually, or if the automatic matching rate is high, reuse the token (Case-Level Token) immediately. ▶ Consequently: The same token—[Name_Plaintiff1], [Name_Defendant1], [Name_Witness_Plaintiff1_Child1]—is used in any document.

[0043] There may be moments when recovery is required, for example, when the court wants to view the original text for internal review because the non-identified state is inconvenient to view, or when the original text needs to be printed for record preservation, the system can retrieve the non-identified text, scan tokens such as [Name_Plaintiff1], [Address_Plaintiff1], [Number_Plaintiff1], and [Name_Witness_Plaintiff1_Child1], query the DetectionTarget table of the RDB for each token, and then restore the original text by replacing it with Original_Value. At this time, the DetectionTarget table of the parties' RDB may be as shown in Table 6 below.

[0044] DetectionTarget table (for token mapping) target_id case_id type original_value token T1 2025Gadan12345 Name_Manuscript Lee Min-soo [Name_Manuscript 1] T2 2025Gadan12345 Name_Defendant Kim Young-soo [Name_Defendant 1] T3 2025Gadan12345 address 123, △△-ro, ○○-gu, Seoul [Address_Manuscript1] T4 2025Gadan12345 phone number 010-1234-5678 [No._Manuscript1] T5 2025Gadan12345 address 999, △△-ro, ○○-gu, Seoul [Address_Defendant 1]

[0045] The party history management unit (380) can register the results of searching and de-identifying the personal information of the party and their representative into the party search history information database. The party verification unit (390) can, when the party de-identification unit searches for the personal information of the party or their representative within the litigation document, transmit the search results to the user terminal (100) to request verification, and when verification is completed from the user terminal (100), allow the verification results to be stored in the party search history information database.

[0046] The general public history management unit (391) can register the results of searching and de-identifying the general public's personal information in the general public search history information database. The general public verification unit (393) can, when the detection unit detects the general public's personal information, transmit the detection result to the user terminal (100) to request verification before the general public de-identification unit replaces the general public's personal information, and when verification is completed from the user terminal (100), the verification result can be stored in the general public search history information database.

[0047] As such, the search history database contains a history of what information the classification model and NER model found, which of this information was verified by humans as personal information, and how the personal information was changed as a result of the verification, so it can be used to train the classification model and NER model.

[0048] The relationship building unit (395) may set each party, agent, and general public as a node and set the relationship between the party, agent, and general public as an edge so as to build the relationship between the party, agent, and general public into at least one relationship model. The relationship model may be composed of any one or at least a combination of GRU (Gated Recurrent Unit), GNN (Graph Neural Network), and KG (Knowledge Graph). That is, if the personal information of the party and general public is de-identified, it may be de-identified but it may be difficult to understand the relationship between each party and general public. In this case, if a relationship model is used, the structure of the person within the incident can be captured together, resulting in a much better structure.

[0049] For example, in cases involving many characters or multiple terms referring to the characters, it is often ambiguous as to who the current subject refers to. Therefore, when conducting case studies, it is common to diagram the event by designating each character as A, B, and C and drawing lines. Taking this into account, by structuring the data by setting each character as a node and establishing the relationships between characters as edges, it is possible to explain the event graphically based on this structure, as shown in Fig. 4 (d), when the LLM searches or summarizes the case.

[0050] Layer role tool or model Floor 1: Classification Model OR NER Model Entity recognition such as people, organizations, and addresses BERT·ELECTRA-based Token Classification + CRF Floor 2: Extract Relationships (Local) Sentence-level relationship labeling BiGRU / BiLSTM + Attention, or BERT-based RE (Relation Extraction) head Layer 3: Graph Refinement Relationship correction in the overall event character graph GNN (GAT, GraphSAGE, etc.) 4th Floor: KG·SNA KG storage, centrality / community analysis, multi-event linkage Neo4j / JanusGraph + SNA library (NetworkX, etc.)

[0051] Similar to a two-layer structure, GRU / BiGRU is a model optimized for extracting relationships within sentences or segments; therefore, it can be utilized to extract relationships within such sentences or segments. For example, since it is suitable for extracting sentence-level relationships such as “Plaintiff A’s daughter B is…” or “Defendant’s friend C is…”, one can first define human entity spans (Plaintiffs A, B, C) using an NER model, slice surrounding sentences around each entity pair, pass those sentences through [BiLSTM / BiGRU + Attention], and classify the relationship between [AB] as [Parent-Child] and the relationship between [Defendant-C] as [Friend]. Similar to a three-layer structure, GNN can refine the relationship structure across the entire event. Since GRU-based systems are sentence-level, they have limitations in integrating relationships across the entire document or event; in such cases, a GNN can be layered on top.

[0052] Graph composition ideas Node Person (plaintiff, defendant, witness, third party, etc.), institution, place Edge (Initial) - Co-occurrence within a sentence / paragraph - Explicit relationships such as "daughter / parent / spouse / cohabitant / lover" - "Potential relationship signals" such as the same address / same contact information / authorization by the same law firm GNN role - For the "initial graph" created under the above conditions, relationship labels are corrected and supplemented via message passing. - Ambiguous relationships are inferred by observing surrounding structures. ▶ Example: If A and B consistently appear together in multiple documents, C is identified as B's 'daughter', and C's address is the same as A's, then → The probability that A and B are "spouses" increases.

[0053] Cases that were ambiguous at the sentence level can be corrected within the context of the entire event, and the relationship structure in major events or events involving multiple people can be obtained more accurately.

[0054] In a four-layer structure, a KG is used to store and retrieve common relationships between events and all events. This is because a KG structure is required to use relationships extracted by GNN or GRU for the long term. The items to be stored in the KG may be, for example, as shown in Table 9 below.

[0055] Store in KG Node Event ID, person (non-identification token), organization, address, etc. Edge - Relationships such as [PARENT_OF], [CHILD_OF], [SPOUSE_OF], [LOVER_OF], [FRIEND_OF], [CO_WORKER_OF], [NEIGHBOR_OF], etc. - Event / job relationships such as [PARTY_IN_CASE], [WITNESS_IN_CASE], [REPRESENTED_BY], [WORKS_AT]

[0056] By establishing a KG in this way, consistent non-identifiable tokens can be assigned when the same person or family appears across multiple events, and it is possible to identify how many levels of a relationship network a particular individual is connected to “within a single event” (finding key figures). Furthermore, it can be expanded to advanced services in the future, such as recommending similar events or pattern analysis.

[0057] In the case of the same 4-layer SNA, which is an indicator that quantifies relationship structures, it enables the identification of degree centrality (who is connected to the most people) and betweenness centrality (who acts as a hub or broker within the network) for the graph of individuals within an event; it also allows for the identification of family-unit or group-unit clusters through community detection. In other words, it can automatically detect the central figure of an event and capture structures where multiple events are intertwined around a single victim or defendant. Furthermore, upon de-identification, it becomes possible to group individuals belonging to the same community into tokens of the same category.

[0058] Currently, only simple notations such as [Name_Witness 1] and [Name_Witness 2] are possible; however, by applying a relationship model, details such as [Name_Plaintiff 1_Spouse], [Name_Plaintiff 1_Child 1], and [Name_Defendant 1_Friend 1] can be identified. Additionally, addresses can be categorized as [Address_Plaintiff 1_Family Shared Residence] or [Address_Simple Workplace Address]. Consequently, even if the individuals involved (parties and the general public) become unrecognizable due to anonymization, the structure remains much more intact when reconstructing and analyzing the case later (in an anonymized state). Furthermore, the error rate is significantly reduced when attempting to summarize or analyze the case using LLM.

[0059] Hereinafter, the operation process according to the configuration of the anonymization service providing server of FIG. 2 described above will be explained in detail with reference to FIG. 3 and FIG. 4. However, it is obvious that the embodiment is merely one of the various embodiments of the present invention and is not limited thereto.

[0060] Referring to FIG. 3, (a) a de-identification service providing server (300) receives a litigation document uploaded from a user terminal (100), then extracts the parties and their representatives from the mandatory entry area as in (b), verifies them with the user terminal (100), and stores them in the party RDB as in (c), thereby storing them as reference values ​​so that they can be used when performing recovery after de-identification, or so that the same de-identification can be performed on other cases of different levels or related documents. Then, the de-identification service providing server (300) can provide the de-identification service to the user terminal (100) after performing de-identification as in (d).

[0061] After searching for parties and agents in this way, the next step is to search for the general public. As shown in (a) of FIG. 4, the personal information of the general public is detected using a NER model, and after verification is received on the user terminal (100) regarding the detected results, the results are stored in the general public RDB and de-identified once verification is complete. Then, this history is stored in the general public search history information DB and used to retrain the NER model. Similarly, (b) the party search history information DB can also be used to retrain the classification model. (c) Furthermore, litigation documents can be recovered using the party RDB and the general public RDB, and as shown in (d), the relationships between each party and the general public can be structured using various relationship models. This allows for providing diagrammatic results when others understand the case while reading the de-identified litigation documents, or using relationships to improve the accuracy of answers when the LLM summarizes the case of the litigation documents or provides answers to search results.

[0062] As for matters not described in the legal tech-based personal information de-identification solution provision method of FIGS. 2 to 4, they are identical to or can be easily inferred from the content described above regarding the legal tech-based personal information de-identification solution provision method through FIG. 1, so further explanation will be omitted.

[0063] FIG. 5 is a diagram illustrating the process of data transmission and reception between each component included in the legal tech-based personal information de-identification solution providing system of FIG. 1 according to an embodiment of the present invention. Hereinafter, an example of the process of data transmission and reception between each component will be described through FIG. 5, but the present invention is not to be interpreted as being limited to such an embodiment, and it is obvious to those skilled in the art that the process of data transmission and reception shown in FIG. 5 may be changed according to various embodiments described above.

[0064] Referring to FIG. 5, the de-identification service provider server receives a lawsuit document from a user terminal (S5100).

[0065] Then, the de-identification service provider extracts the personal information of the parties and their representatives included in the mandatory entry area within the litigation document (S5200), maps the personal information of the parties and their representatives to pre-set item identifiers and identification numbers, and stores it in the parties' RDB (Relational Database) (S5300).

[0066] Additionally, when the de-identification service provider server detects personal information of a party or representative within a litigation document, it assigns an item identifier and an identification number to replace it (S5400), and detects personal information of a general person excluding the party or representative within the litigation document using a pre-established NER (Named-Entity Recognition) model (S5500).

[0067] Then, the de-identification service provider server replaces the personal information of the general public by mapping it to a pre-configured item identifier and identification number (S5600), and stores the personal information of the general public in the general public RDB by mapping it to the pre-configured item identifier and identification number (S5700).

[0068] The order of the steps described above (S5100–S5700) is merely an example and is not limited thereto. That is, the order of the steps described above (S5100–S5700) may vary, and some of these steps may be executed simultaneously or deleted.

[0069] As for matters not described in the method of providing a legal tech-based personal information de-identification solution of Fig. 5, they are identical to or can be easily inferred from the content described in Figs. 1 to 4, so further explanation will be omitted.

[0070] A method for providing a legal tech-based personal information de-identification solution according to one embodiment described through FIG. 5 may also be implemented in the form of a recording medium containing computer-executable instructions, such as an application or program module executed by a computer. A computer-readable medium may be any available medium accessible by a computer and includes both volatile and non-volatile media, as well as removable and inseparable media. Additionally, a computer-readable medium may include all computer storage media. Computer storage media include both volatile and non-volatile, removable and inseparable media implemented by any method or technique for storing information, such as computer-readable instructions, data structures, program modules, or other data.

[0071] The method for providing a legal tech-based personal information de-identification solution according to one embodiment of the present invention described above may be executed by an application basically installed on a terminal (which may include a program included in a platform or operating system, etc., basically installed on the terminal), or by an application (i.e., a program) directly installed by a user on a master terminal through an application providing server, such as an application store server, an application, or a web server related to the service. In this sense, the method for providing a legal tech-based personal information de-identification solution according to one embodiment of the present invention described above may be implemented as an application (i.e., a program) that is basically installed on a terminal or directly installed by a user, and may be recorded on a computer-readable recording medium such as a terminal.

[0072] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will understand that other specific forms can be easily modified without altering the technical spirit or essential features of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form.

[0073] The scope of the present invention is defined by the claims set forth below rather than by the detailed description above, and all modifications or variations derived from the meaning and scope of the claims and equivalent concepts thereof should be interpreted as being included within the scope of the present invention.

Claims

Claim 1 A user terminal that registers a litigation document and outputs a result in which personal information within the said litigation document has been anonymized; A de-identification service providing server comprising: a receiving unit that receives a litigation document from the user terminal; an extraction unit that extracts personal information of a party and their representative included in a mandatory entry area within the litigation document; a party storage unit that stores the personal information of the party and their representative in a party RDB (Relational Database) by mapping it to a pre-set item identifier and identification number; a party de-identification unit that replaces the personal information of the party or their representative by assigning the item identifier and identification number when the personal information of the party or their representative is searched within the litigation document; a detection unit that detects personal information of a general public excluding the party or their representative within the litigation document using a pre-established NER (Named-Entity Recognition) model; a general public de-identification unit that replaces the personal information of the general public by mapping it to a pre-set item identifier and identification number; and a general public storage unit that stores the personal information of the general public in a general public RDB by mapping it to the pre-set item identifier and identification number; wherein the de-identification service providing server registers the results of the search and de-identification of the personal information of the party and their representative into a party search history information database. A legal tech-based personal information de-identification solution providing system further comprising: a party history management unit; and a party verification unit that, when the party de-identification unit searches for personal information of the party or their representative within the litigation document, transmits the search result to the user terminal to request verification, and when verification is completed from the user terminal, stores the verification result in the party search history information database. Claim 2 delete Claim 3 delete Claim 4 A legal tech-based personal information de-identification solution providing system, characterized in that, in claim 1, the de-identification service providing server further includes a general public history management unit that registers the results of searching and de-identifying the general public's personal information as a general public search history information database. Claim 5 A legal tech-based personal information de-identification solution providing system, characterized in that, in the case where the de-identification service providing server detects the personal information of the general public in the detection unit, before the general public de-identification unit replaces the personal information of the general public, the general public verification unit transmits the detection result to the user terminal to request verification, and when verification is completed from the user terminal, the verification result is stored in the general public search history information database. Claim 6 A legal tech-based personal information de-identification solution providing system, characterized in that, in claim 1, the de-identification service providing server further comprises a relationship building unit that sets each party, agent, and general public as a node and sets the relationship between the party, agent, and general public as an edge, so as to establish the relationship between the party, agent, and general public as at least one relationship model. Claim 7 A legal tech-based personal information de-identification solution providing system, characterized in that, in claim 6, the relationship model comprises any one or at least a combination of GRU (Gated Recurrent Unit), GNN (Graph Neural Network), and KG (Knowledge Graph).

Images