KR102825770B1 - System for providing intelligent law information service platform using law navigation - Google Patents

Abstract

A system providing an intelligent legal information service platform using legal navigation is provided, and includes a platform service provision server including a user terminal that inputs a query in everyday terms, receives and outputs a response to the query, a construction unit that constructs a legal knowledge base by mapping an everyday terms database (DataBase) and a legal terms database to define the relationship between everyday terms and legal terms, an analysis unit that extracts search terms from the query when a query in everyday terms is input from a user terminal and analyzes morphemes of the search terms, a transmission unit that transmits the analysis results of the morpheme analysis to the legal knowledge base, a query unit that searches for index information from previously constructed legal information index data when a legal term corresponding to the search term is extracted from the legal knowledge base, and a provision unit that returns a search result including the index information to the user terminal.

Description

{SYSTEM FOR PROVIDING INTELLIGENT LAW INFORMATION SERVICE PLATFORM USING LAW NAVIGATION}

The present invention relates to a system providing an intelligent legal information service platform using legal navigation. The system learns the relationship between everyday terms and legal terms through an artificial intelligence algorithm, thereby providing a response even when a query is made using everyday terms, and providing related keywords as legal navigation based on everyday terms.

As legal rules, statutes, and judgments increase, a lot of legal knowledge is accumulated and its form is becoming more complex. There is a need to structure a huge amount of legal information worldwide and facilitate searching, and research on ontologies in the legal field is being conducted more actively, and with the advent of the Semantic Web, legal ontology is being suggested as a means to facilitate legal information searching. Based on Natural Language Processing (NLP) technology, semantic relationships between legal concepts are extracted and made into legal ontology, which is then applied to legal search systems. Keyword-based searches, which are the mainstream of current information retrieval methods, fail to satisfy both recall and accuracy due to excessive occurrence of information, inconsistencies between legal terms and everyday terms, ambiguity of legal terms, and users’ desire for Q&A-style searches. Therefore, the emergence of a new method is expected.

At this time, a method of providing laws by setting up a system of upper and lower laws or providing results according to a searched query in the form of navigation was studied and developed. In relation to this, in the prior art, Korean Patent No. 10-2000546 (announced on July 17, 2019) and Korean Patent No. 10-1886007 (announced on August 6, 2018), when searching for building regulation information, by setting the direct relationship between individual provisions of upper and lower laws and mapping building laws, only the provisions of laws that meet the conditions can be extracted and reviewed according to the user's purpose and need, and by setting the direct relationship between individual provisions of upper and lower laws in the order of laws, enforcement decrees, enforcement rules, and ordinances, lower laws can be accurately mapped to upper laws and linked with hyperlinks so that laws can be checked immediately, and a structure for extracting a search query from information entered on a user terminal to search for legal information, and searching for the searched legal information. Each configuration is disclosed to enable users to intuitively and easily recognize searched information by providing a navigation screen visualized by arranging it in a two-dimensional or three-dimensional shape.

However, the former configuration only reveals a configuration that allows users to know the legal system, but when searching with everyday terms that users use on a daily basis, not only the legal system but also the desired search results are not displayed, and the latter configuration visualizes the search results, but also does not allow searching with everyday terms, so it only reveals a configuration that does not provide a navigation screen for the search results, let alone the search results. The current legal information retrieval system, which operates on a keyword-based search, frequently fails to provide search results for general queries entered by users because it only uses legal terms. The gap between everyday terms used by the general public and legal terms used in the law makes it difficult for the general public to obtain the legal search results they want from the current search engine that requires legal terms to be selected as keywords. Therefore, research and development of a search system that allows the general public, who are not experts in law, to search with everyday terms and obtain the desired results is required.

One embodiment of the present invention provides a system for providing an intelligent legal information service platform using legal navigation, which constructs a database for everyday terms and legal terms, and learns, verifies, and tests an artificial intelligence algorithm using each mapping relationship so that even if a query is made using everyday terms, a search term is converted into a legal term, and a search result for the search term is found and an answer is provided, so that a user can receive a desired answer even if he or she does not know the exact keyword or legal term, and visualizes related keywords as a legal navigation using a related keyword map based on the searched everyday term so that the related laws can be intuitively understood, and provides information on which law is applied preferentially by constructing a system between each law, legal article, and legal term. However, the technical task to be achieved by this embodiment is not limited to the technical task described above, and other technical tasks may exist.

As a technical means for achieving the above-described technical task, one embodiment of the present invention includes a platform service providing server including a user terminal that inputs a query in everyday terms, receives and outputs a response to the query, a construction unit that constructs a legal knowledge base by mapping an everyday terms database (DataBase) and a legal terms database to define a relationship between everyday terms and legal terms, an analysis unit that extracts a search term from the query when a query in everyday terms is input from the user terminal and analyzes the morphemes of the search term, a transmission unit that transmits the analysis result of the morphemes analyzed to the legal knowledge base, a query unit that searches for index information from previously constructed legal information index data when a legal term corresponding to the search term is extracted from the legal knowledge base, and a provision unit that returns a search result including the index information to the user terminal.

According to any one of the problem solving means of the present invention described above, by constructing a database of everyday terms and statutory terms, and learning, verifying and testing an artificial intelligence algorithm using each mapping relationship, even if a query is made using everyday terms, a search term is converted into a statutory term, and a search result for the search term is found and an answer is provided, so that a user can receive a desired answer even if he or she does not know the exact keyword or statutory term, and by visualizing related keywords as a statutory navigation using a related keyword map based on the searched everyday term, the related laws can be intuitively understood, and by constructing a system between each statute, statutory article and statutory term, information can be provided as to which law is preferentially applied.

FIG. 1 is a drawing for explaining a system for providing an intelligent law information service platform using law navigation according to one embodiment of the present invention.
Figure 2 is a block diagram for explaining a platform service providing server included in the system of Figure 1.
FIGS. 3 and 4 are drawings for explaining an example of an intelligent law information service platform service implemented using law navigation according to one embodiment of the present invention.
FIG. 5 is a flowchart illustrating a method for providing an intelligent law information service platform service using law navigation according to one embodiment of the present invention.

Hereinafter, with reference to the attached drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily practice the present invention. However, the present invention may be implemented in various different forms and is not limited to the embodiments described herein. In addition, in order to clearly describe the present invention in the drawings, parts that are not related to the description are omitted, and similar parts are assigned similar drawing reference numerals throughout the specification.

Throughout the specification, when a part is said to be "connected" to another part, this includes not only the case where it is "directly connected" but also the case where it is "electrically connected" with another element in between. Also, when a part is said to "include" a component, this should be understood to mean that it may further include other components, unless specifically stated to the contrary, and does not preclude the presence or possibility of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

The terms "about," "substantially," and the like used throughout the specification are used in a meaning close to or at the numerical value when manufacturing and material tolerances inherent in the stated meaning are presented, and are used to prevent unscrupulous infringers from unfairly using the disclosure in which exact or absolute numerical values are mentioned in order to help understanding of the present invention. The terms "step of doing ~" or "step of ~" used throughout the specification of the present invention do not mean "step for ~."

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

Some of the operations or functions described as being performed by a terminal, apparatus or device in this specification may instead be performed by a server connected to the terminal, apparatus or device. Similarly, some of the operations or functions described as being performed by a server may also be performed by a terminal, apparatus or device connected to the server.

In this specification, some of the operations or functions described as terminal and mapping or matching may be interpreted to mean mapping or matching the terminal's unique number or personal identification information, which is the terminal's identifying data.

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

FIG. 1 is a drawing for explaining an intelligent law information service platform providing system using law navigation according to one embodiment of the present invention. Referring to FIG. 1, an intelligent law information service platform providing system (1) using law navigation may include at least one user terminal (100), a platform service providing server (300), and at least one national law information providing server (400). However, since the intelligent law information service platform providing system (1) using law navigation of FIG. 1 is only one embodiment of the present invention, the present invention is not limited by FIG. 1.

At this time, each component of Fig. 1 is generally connected via a network (Network, 200). For example, as shown in Fig. 1, at least one user terminal (100) can be connected to a platform service providing server (300) via a network (200). And, the platform service providing server (300) can be connected to at least one user terminal (100) and at least one national law information providing server (400) via the network (200). In addition, at least one national law information providing server (400) can be connected to the platform service providing server (300) via the network (200).

Here, a network means a connection structure that enables information exchange between each node, such as multiple terminals and servers, and examples of such networks include a local area network (LAN), a wide area network (WAN), the Internet (WWW: World Wide Web), wired and wireless data communication networks, telephone networks, and wired and wireless television communication 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, the Internet, LAN (Local Area Network), Wireless LAN (Wireless Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), RF (Radio Frequency), Bluetooth (Bluetooth) network, NFC (Near-Field Communication) network, satellite broadcasting network, analog broadcasting network, and DMB (Digital Multimedia Broadcasting) network.

In the following, the term at least one is defined as a term including the singular and the plural, and it will be obvious that even if the term at least one does not exist, each component can exist in the singular or the plural and can mean the singular or the plural. In addition, it will be said that whether each component is provided in the singular or the plural may change depending on the embodiment.

At least one user terminal (100) may be a user terminal that searches for everyday terms as a query word using a web page, app page, program or application related to an intelligent law information service platform service using law navigation, or inputs a sentence including an everyday term as a query word to search for results. At this time, everyday terms mean terms that most ordinary people use in their daily lives instead of law terms, although there are law (legal) terms. For example, there is a law term called brokerage fee, but everyday terms such as dubbing fee or brokerage fee are often used. At this time, dubbing fee or brokerage fee are everyday terms, and brokerage fee is a law term. When receiving a response to a query from a platform service providing server (300), the user terminal (100) may be a terminal that outputs law navigation in which keywords included in the query and related keywords are diagrammed in a graph format. At this time, the legal navigation refers to a map in the form of a graph in which each related keyword included in the query has a node and is connected by edges according to the degree of similarity with the keyword.

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

The platform service providing server (300) may be a server that provides an intelligent law information service platform service web page, app page, program or application using law navigation. In addition, the platform service providing server (300) may be a server that maps and stores the relationship between the everyday term DB and the law term DB in a law knowledge base, builds a dataset of [Everyday term-Law term], and allows an artificial intelligence algorithm to learn this relationship so that when an everyday term is entered as a query, a law term is output. In addition, the platform service providing server (300) may be a server that extracts keywords from a query entered from a user terminal (100) using a text analysis technique, extracts keywords related to the extracted keywords, creates a related keyword map in which edges are connected to a network according to the similarity between the keywords and related keywords, and then provides the map as law navigation. At this time, since there are cases where a legal term exists across multiple legal provisions and laws, the platform service providing server (300) may be a server that stores and learns the relationship between [legal term-legal provision-law] using an artificial intelligence algorithm, and also stores and learns the relationship between [law-law] using an artificial intelligence algorithm. Accordingly, the platform service providing server (300) may be a server that, when multiple laws are applied to one keyword, gives priority to multiple laws according to the relationship between general laws and special laws and lists them in the law navigation.

Here, the platform service providing server (300) may be implemented as a computer that can connect to a remote server or terminal via a network. Here, the computer may include, for example, a notebook, desktop, laptop, etc. equipped with a navigation system and a web browser.

At least one national law information provision server (400) may be a server that provides sources of various laws, precedents, interpretation examples, etc. when the platform service provision server (300) provides a response to an inquiry using a web page, app page, program, or application related to an intelligent law information service platform service using law navigation. Here, at least one national law information provision server (400) may be implemented as a computer that can connect to a remote server or terminal through a network. Here, the computer may include, for example, a notebook, desktop, or laptop equipped with a navigation system and a web browser.

FIG. 2 is a block diagram for explaining a platform service providing server included in the system of FIG. 1, and FIGS. 3 and 4 are diagrams for explaining an embodiment of an intelligent law information service platform service implemented using law navigation according to one embodiment of the present invention.

Referring to FIG. 2, the platform service providing server (300) may include a construction unit (310), an analysis unit (320), a transmission unit (330), a query unit (340), a provision unit (350), a diagram unit (360), a legal system visualization unit (370), a basis presentation unit (380), a ranking presentation unit (390), an automatic response unit (391), an anti-hallucination unit (393), a case-based question composition unit (394), a component requirement extraction unit (395), a burden of proof guidance unit (396), a legal reasoning unit (397), and an expert matching unit (398).

When the platform service providing server (300) according to one embodiment of the present invention or another server (not shown) operating in conjunction with it transmits an intelligent law information service platform service application, program, app page, web page, etc. using law navigation to at least one user terminal (100) and at least one national law information providing server (400), the at least one user terminal (100) and the at least one national law information providing server (400) may install or open the intelligent law information service platform service application, program, app page, web page, etc. using law navigation. In addition, the service program may be driven on at least one user terminal (100) and at least one national law information providing server (400) 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 written in HTML (Hyper Text Mark-up Language), and includes, for example, Chrome, Edge (Microsoft Edge), Safari, Firefox, Whale, UC Browser, etc. In addition, an application refers to an application program on a terminal, and includes, for example, an app running on a mobile terminal (smartphone).

Referring to FIG. 2, the construction unit (310) can construct a legal knowledge base that stores a legal term DB and a legal term DB by mapping them to define the relationship between the legal terms and legal terms. At this time, the legal knowledge base may be something that stores the relationship between the legal terms and the legal terms by mapping them. In addition, the legal knowledge base may be a database that stores which legal provisions the legal terms are located in, which laws the legal provisions belong to, and what the priority is when the legal terms belong to multiple laws.

The analysis unit (320) can extract search terms from the query and analyze the morphemes of the search terms when a query is entered in everyday language from the user terminal (100). The user terminal (100) can input a query in everyday language. For example, when the user terminal (100) inputs the question, "What should I do when the landlord asks me to raise the key money excessively?" as a query on a site such as Fig. 4f, i.e., the National Law Information Center page of the Ministry of Government Legislation (https://www.law.go.kr/LSW/main.html), the search terms (keywords) [landlord], [key money], and [raise] can be extracted from the query as in Fig. 4d, and the morphemes of these three search terms can be analyzed.

The transmission unit (330) can transmit the analysis result of morpheme analysis to the legal knowledge base. At this time, since [everyday term-legal term] is mapped in the legal knowledge base, [landlord] can be changed to [landlord] or [rental business operator], [rental deposit] can be changed to [rental deposit], and [raise] can be changed to [increase] or [increase], [increase], [increase], etc. As shown in Fig. 4d, the search unit (340) can search for index information from the previously constructed legal information index data when a legal term corresponding to a search word is extracted from the legal knowledge base. At this time, the node in the yellow circle is the search word, the legal term related to it is the node in the blue box, and the everyday term linked to the legal term is the node in the white circle.

<Index>

The basic concept of indexing is explained with reference to Figure 4g. Indexing refers to the process of preprocessing a document set for efficient retrieval, and the indexing process consists of ① extracting text from a binary document file (Text Extraction), ② tokenization, ③ stop-word removal, ④ normalization, and ⑤ inverted indexing.

① Text extraction Extract pure text from documents with various formats ② Tokenization Tokenize text to make it easier to search ③ Remove stop words Remove unnecessary idioms and stop words that do not make sense ④ Normalization Change the term with meaning to its base form and create a stem ⑤ Reverse file index Find specific documents from words through an index rather than searching for words in documents.

At this time, the blue part of Fig. 4g, [index module-reverse file index-search module], goes through the processes of Figs. 4h to 4j. To this end, [morphological analysis-part-of-speech tagging-stop word removal-compound noun decomposition-index word extraction] is performed in the index module as in Fig. 4h, and as in Fig. 4i, the reverse file index identifies in which document each index word appears, and then as in Fig. 4j, the search module provides which document it has the highest similarity to. At this time, the source of Figs. 4g to 4j is https://jiwondev.tistory.com/38. It is made clear that this site and related drawings are only for explaining the index and are not related to an embodiment of the present invention. In addition, since the indexing and search processes are not limited thereto, if another method for indexing exists, it may be possible to use that method.

The provision unit (350) can return search results including index information to the user terminal (100). The Ministry of Government Legislation is currently operating the Easy-to-Find Living Law Information (https://www.easylaw.go.kr/CSP/Main.laf) site, and various questions and answers are already databased on this site. Accordingly, the automatic response unit (391) can directly respond in real time from the national law information provision server (400) as described below, but can also provide answers using the [question-response] database already built in the living law information.

Referring to FIG. 4b, the user terminal (100) can receive and output a response to a query as shown in Table 2.

question My landlord wants me to increase my deposit. What should I do? response If the landlord asks you to raise the deposit,
First, according to Article △△ of the Special Act on Private Rental Housing, “When a rental business operator requests an increase in rent during the rental period, the increase rate prescribed by Presidential Decree must not exceed 5% of the rent, taking into account the housing cost index, the rate of change in rent in the surrounding area, the number of rental housing units, etc. This means that there are certain standards and restrictions on rent increases.”
In addition , according to Article ○○ of the Enforcement Decree of the Housing Lease Protection Act, “when renewing a lease agreement, the request for an increase in rent or deposit cannot exceed 1/20 of the agreed rent or deposit.”

Also, if the question is related to passports, questions and answers like those in Table 3 are also possible.

question I'm planning my first trip abroad. How do I get a passport? response In order to obtain a passport, you must first submit a passport application form and one passport photo. The passport photo must be a natural color upper body front view photo taken within 6 months prior to the passport issuance application date without a hat or other items, and must be 3.5 cm wide and 4.5 cm high, with a head length (from chin to crown) of 3.2 cm or more and 3.6 cm or less. In addition, you must also prepare other documents required for issuing a general passport, as prescribed by the Ministry of Foreign Affairs Ordinance. (Passport Act Enforcement Decree Article □□)
In addition, the application for issuance of a passport must be made in person, and if the applicant is under 18 years of age, the application must be made with the consent of a legal representative. (Passport Act Article △△)

The diagrammatic section (360) can provide a related keyword map that visualizes the relationship between everyday terms and legal terms using nodes and edges. At this time, although it is referred to as a related keyword map, a method capable of defining and expressing the relationship between each node and edge, such as a GNN (Graph Neural Network) or a knowledge graph, can also be used. The applicant of the present invention has disclosed a legal information knowledge base as shown in FIGS. 4k and 4l in public data, and is currently continuously updating terms and defining relationships and conducting final work in order to commercialize a service according to one embodiment of the present invention.

Accordingly, the diagramming unit (360) can provide a related keyword map based on the relationships between everyday terms and statutory terms that have already been constructed. The legal system visualization unit (370) can store statutory terms by mapping them to at least one statutory provision, store at least one statutory provision by mapping it to at least another statutory provision, store at least one relationship between statutes by mapping it, and reflect the relationship between statutory terms and statutory provisions, the relationship between at least one statutory provision, and the relationship between at least one statute in the related keyword map. Referring to the applicant's previously applied and registered or published Korean Patent No. 10-1011394 (announced on January 28, 2011) or Korean Patent Publication No. 2014-0131831 (published on November 14, 2014), a database for a statutory system diagram and related statutes has already been constructed, and a knowledge base is planned to be constructed and linked while constructing an intelligent statutory information service platform as shown in FIG. 4e.

Based on this, the legal system can be visualized. For example, in Fig. 4d, for the word [Housing Lease], three laws are described, and in the order of general law-special law, the [Housing Lease Protection Act], which is applied first, is printed first, followed by the [Special Act on Private Rental Housing], and lastly, the [Civil Act] can be applied. In summary, the relationship of [Everyday Terms-Legal Terms-Legal Articles-Legal Acts] is stored, and the priority or relationship between each law can be stored as [Legal Act-Legal Act], and this relationship can be displayed in a related keyword map such as Fig. 4d.

The basis presentation unit (380) can move to the page of the statutory provision corresponding to the node when a node is selected from the user terminal (100) and provide it to the user terminal (100). When the node [Rent Increase/Decrease Claim] in Fig. 4d is selected, it can move to the statutory provision linked thereto, i.e., Article 7 of the Housing Lease Protection Act. At this time, since the statutory provision is also provided by the Ministry of Government Legislation (https://www.law.go.kr/), the movement can be set by simple page linking.

The ranking presentation section (390) can guide the legal system according to the ranking between the general law and the special law when at least one statutory provision and at least one other statutory provision are in a relationship between general law and special law, and can output and guide the statutory provision according to the special law that is applied preferentially.

The automatic response unit (391) can generate a response to a query input from a user terminal (100) using a generative AI (Artificial Intelligence) based on an LLM (Large Language Model) and provide the response to the user terminal (100). Recently, with the introduction of generative AI in the legal tech field, there are increasing cases in which generative AI provides responses to queries that should be answered by various legal experts such as lawyers, tax accountants, and accountants. Using this, one embodiment of the present invention can analyze a user's query and generate a response corresponding to it using generative AI.

The hallucination prevention unit (393) can generate a response based on the national law information provision server (400) to prevent hallucinations in the generative AI, and can include the source of the response in the response. If the generative AI cannot generate a response, it can cause hallucinations and give an incorrect answer. Since the response provided by the Ministry of Government Legislation should not be an answer that is not based on the law, the source can be indicated so that only responses based on the law are generated, and only answers based on the source can be generated.

<Case-based question composition>

The case-based question composition unit (394), referring to (a) of FIG. 3b, collects case law from the national law information provision server (400), extracts the upper-level law of the article that is referenced more than a preset number of times among the articles that appear as the basis law in the case law, and constructs a dataset with pairs of questions and answers in a triplet structure based on the law of the upper-level law, but uses GPT-4 to provide actual case-based question examples in a few shots within the prompt so that the questions are composed as case-based questions based on the law.

The Korean legal question-answering datasets are provided by AI-Hub, LegalQA, and EasyLaw, respectively, but ① they do not construct case-based questions, and ② they do not connect questions with supporting laws. In other words, LegalQA and EasyLaw construct case-based questions but do not connect questions with supporting laws, and AI-Hub connects questions with supporting laws, but the other two do not. Accordingly, in one embodiment of the present invention, in the first step, the distribution of upper laws can be identified based on case laws provided by the National Law Information Provision Server (400), i.e., the Ministry of Government Legislation. The reason for this is to consider the ratio of upper laws frequently cited in the real world.

For example, in the case law dataset crawled by the Ministry of Government Legislation, only upper-level laws that have been referenced more than N times can be extracted from the articles that appear as the basis of the law. Then, in the second step, when constructing a dataset of pairs of questions and answers based on the laws of the upper-level laws, ① questions can be generated based on cases, and ② a dataset with a triplet structure of [question-law-answer] can be constructed to connect questions and basis laws. At this time, question examples can be provided in few shots so that questions can be generated based on actual cases.

At this time, few-shot learning is a method in which LLM-based generative AI learns to generate answers to new questions even after seeing a small number of examples. For example, when asking LLM-based generative AI to generate legal questions, if you simply say "make a legal question," LLM-based generative AI may not understand it properly. Therefore, by providing several samples including real cases, LLM-based generative AI is induced to learn the pattern. At this time, the few-shot examples provided in the prompt can be, for example, as in Table 4.

[Example 1]
Question: I lent money to a friend without interest, but I suddenly want to cancel the contract. Is that possible?
Answer: According to Article 601 of the Civil Code, an interest-free consumer loan contract can be terminated at any time, but any damages incurred by the other party must be compensated. [Example 2]
Question: I bought something from a second-hand store, but it was counterfeit. Can I get a refund?
Answer: According to Article 17 of the Electronic Commerce Act, consumers can cancel their contract within 7 days after signing the contract. Therefore, if a counterfeit product is confirmed, you can request a refund from the seller. [Example 3]
Question: I was unfairly fired from my company. What should I do?
Answer: According to Article 23 of the Labor Standards Act, you cannot be fired without just cause, and if you are unfairly dismissed, you can apply for relief to the Labor Relations Commission.

The reason for providing a few examples is to encourage the LLM-based generative AI to generate realistic questions rather than simply listing legal provisions when creating legal questions. When the LLM-based generative AI sees several examples like the above, it will learn "how to create questions that fit real cases based on legal provisions." After that, the LLM-based generative AI can create new realistic questions on its own for given legal provisions, so it will be possible to build the LLM-based generative AI in a form that is closer to more realistic legal consultation.

<Extracting the elements of the law>

Referring to (b) of FIG. 3B, the component extraction unit (395) extracts and classifies the component elements of laws included in the facts when the user terminal (100) inputs the facts in everyday terms, and guides them to the user terminal (100), when performing instruction tuning on an LLM (Large Language Model)-based generative AI (Artificial Intelligence), a dataset of instructions, facts, and component elements is constructed, and instruction tuning is performed on the LLM-based generative AI. In the inference step of the LLM-based generative AI, when providing the inference result, similar case data including the facts and component elements can be searched using RAG (Retrieval-Augmented Generation), and then the inference result can be provided together with the similar case data.

In order to determine the facts, information must be extracted from the input data and legal analysis must be performed, which can be done through a process such as Table 5.

step explanation example Technology used Preprocessing input data LLM-based generative AI analyzes legal data entered in various ways to organize facts.
- User question analysis → Decompose the legal consultation questions entered by users
- Processing legal documents (judgments, contracts) → Extracting key information within the documents
-Analysis of case records → Organizing facts in police records, litigation documents, etc.
If you ask, "A lent money to B, but B doesn't pay it back. What should I do?", you can extract key legal concepts such as "money lending contract," "default," and "compensation for damages." Natural Language Processing (NLP)
Example models: OpenAI GPT-4, Google BERT, LexiLaw Information Extraction (Entity & Fact Extraction) Automatically identify and classify key elements of an incident
- Named Entity Recognition (NER) → Extract elements such as people (A, B), dates, and amounts
-Relation Inference (Relation Extraction) → Recognize the ‘money lending’ relationship from “A lent money to B” "In May 2023, A lent 10 million won to B."

-Extracted information:
Person: A (creditor), B (debtor)
Act: lending money
Amount: 10 million won
Date: May 2023
Legal Keywords: claim for return of rental deposit, possibility of termination of contract - Named entity recognition (NER), relationship inference
-Example models: SpaCy, Stanford NLP, DeepSeek-V3 Legal Reasoning & Case Matching Analyze the facts of the case in connection with the law
-Legal Text Matching → Linking extracted information to related legal provisions
-Case Law Retrieval → Find similar case laws and provide legal basis
-Building a legal logic tree → Organizing the legal application process through logical thinking Example: "What legal action can A take if B does not repay the money?"

-Legal analysis process
-Related legal provisions: Civil Act Article 548 (Termination of Contract), Article 750 (Compensation for Damages for Tort)
-Case search: "Supreme Court 2015da23456 - Case law recognizing damages for nonperformance of obligation"
-Legal logic tree: Money loan → Breach of contract → Debtor's responsibility for payment - Legal Knowledge Graph, RAG (Augmented Search Generation)
-Example models: Westlaw AI, LexisNexis

At this time, in the case of the legal logic tree, the legal reasoning section (397) will explain in detail.

On the other hand, there are elements of the law to satisfy this, and if these elements are not satisfied, each law will not be established. For example, in the case of Article 347 (Fraud) of the Criminal Act, it only states that [A person who deceives another person to receive property or obtain a property benefit shall be punished by imprisonment for not more than 10 years or a fine not exceeding 20 million won], but it is impossible for the general public to understand what elements of the law it contains. Anyone with legal knowledge knows that the elements of fraud (establishment elements) must exist as follows: [Deception (cheating) → obtaining a property benefit → occurrence of damage]. However, even if you have legal knowledge, it is difficult to determine which acts are actually deception, which situations are property benefits, and which are occurrences of damage if you do not actually gain experience in a law office or law firm.

Accordingly, in the structure of [instruction-input-output], a dataset of [instruction-facts-elements] can be constructed, and instruction tuning can be performed so that [instruction] outputs [What are the elements of the law?], [input] outputs [facts], and [output] outputs [elements]. Accordingly, when asked, [Is it fraud if I send bricks from Junggonara?], it provides guidance on whether the act of sending bricks corresponds to [deception → acquisition of property benefit → occurrence of damage].

Component Description Fraud must be committed through deception → acquisition of property benefit → occurrence of damage.
In a second-hand transaction, if you deceive the buyer into believing that you are selling a normal item (deception) and instead send bricks or other worthless items (causing damage), this can be recognized as an act of deceiving the buyer and obtaining money (acquiring financial benefits).

At this time, instruction tuning is a method to pre-learn so that the LLM-based generative AI can better follow the given instructions. In other words, it is not simple data learning, but a process of tuning the model to generate an appropriate response according to the model builder's request (command). In addition, even an LLM-based generative AI tuned in this way may not be able to give an accurate answer, so similar case data can be searched and then provided based on the facts and elements. At this time, when searching for similar case data similar to the facts, the Korean embedding model based on Sentence BERT, jhgan/ko-sroberta-multitask, can be used to identify the similarity between sentences. This is an effective method for extracting case data similar to the facts.

<Guide to burden of proof>

The burden of proof guidance unit (396), referring to (c) of FIG. 3b, collects case law from the national law information server (400), derives facts and laws from the case law, maps the burden of proof according to laws to the facts and laws and stores it to build a dataset, performs instruction tuning on a generative AI (Artificial Intelligence) based on a Large Language Model (LLM), and when facts are input at the inference stage, it can guide the laws and burden of proof. At this time, the burden of proof (Burden of Proof) means the responsibility to legally prove that the facts asserted in the lawsuit are true. That is, in the case of a civil lawsuit, in order for a claim to be accepted in court, it must be proven that the claim is true (Article 288 of the Civil Procedure Act), and in the case of a criminal lawsuit, the prosecutor must prove the criminal charge (Article 307 of the Criminal Procedure Act). However, there are cases where the burden of proof is reversed. In the case of medical malpractice lawsuits, it is too difficult for the patient to assert this, so the medical staff must prove that there was no medical error. In disputes related to breach of contract, the debtor must prove performance of the contract. In tort damages cases, the burden of proof must also be proven that the perpetrator was not liable.

Accordingly, when constructing a dataset of [facts-laws-burden of proof] and performing the instruction tuning described above, a dataset of [instructions-facts-burden of proof] can be constructed in the structure of [instructions-input-output], and instruction tuning can be performed so that [instructions] are [What should I prove in the facts?], the input is [facts], and the [output] is [burden of proof]. Accordingly, when the fact that [I had plastic surgery but side effects occurred] is input, the configuration elements and burden of proof can be guided as in Table 6 below.

Medical malpractice lawsuit elements and burden of proof Components explanation Burden of proof ① Violation of the doctor’s duty of care (negligence) Did the medical staff not provide treatment that meets the standard of general medical care? Victim's proof (but hospital submits medical records) ② Causality (link between medical error and damage) Has the patient suffered harm as a result of the doctor's negligence? The victim must prove it (but the burden of proof may be reduced in court) ③ Damage Occurrence (whether the patient is harmed) Has the patient actually suffered physical or mental harm? The victim proved ④ Illegality (whether it is a socially acceptable medical practice) Was the medical procedure appropriate? Doctor-certified (medical staff may need to verify legitimacy)

In order to derive facts from case law, you can use the TextRank algorithm. TextRank is an extractive summarization algorithm that constructs a graph based on the similarity between sentences, calculates the importance of each sentence, and summarizes sentences in order of importance. Of course, you can extract facts using various other algorithms.

<Legal Reasoning>

The legal reasoning unit (397), referring to (d) of Fig. 3b, can generate a tree structure by arranging evidence and social conventions as lower nodes, intermediate facts as middle nodes, and the final facts that must be ultimately proven in the case as upper nodes in order to perform legal reasoning from the facts to the final facts, and provide the tree structure to the user terminal (100). Currently, most LLM-based generative AIs perform tasks such as summarizing cases, searching legal documents, or analyzing precedents, but lack a logical system for the process of deriving legal facts. In other words, the main evidence and logic of the case are not systematically organized, which may cause a lack of transparency and bias in the judgment process. Accordingly, by providing the legal reasoning process as an LLM-based generative AI and visualizing it, the LLM-based generative AI learns what logic is used to reach a conclusion (judgment), and based on this, it can predict what kind of judgment may be made on the user's legal dispute in the future. This is significant in that it is not simply the result of memorizing case law like the existing LLM generative AI, but rather it is the result of evolving the LLM-based generative AI to actually think like a human and go through a logical process to reach a conclusion.

The tree can be constructed by connecting the root node (the (top) node) of the fact that must be proven, and the intermediate facts and evidence supporting it to the lower nodes. For example, let's assume the process of concluding a murder (murder under Article 250 of the Criminal Act). For example, if we assume that it is a case where A stabbed B to death, an inference process like Table 8 below and a tree construction like Table 9 are possible.

Tree structure core elements Node location explanation evidence Subnode - Objective data (documents, testimonies, CCTV footage, etc.) that serve as the basis for legal judgment
-Located at the bottom of the tree (leaf node)
-Example: “The defendant was caught on CCTV stabbing the victim.” Social conventions Subnode Common experience or social conventions that judges apply when making legal decisions
-Example: "If a weapon is used, it is usually considered to be intent to kill." Intermediate facts Intermediate node -Interim conclusion derived by synthesizing multiple pieces of evidence
-Mixing several intermediate facts to develop a higher-level conclusion (final fact)
-Example: “It is admitted that the defendant stabbed the victim.” Final Fact Parent node (top node) -Key facts that must be ultimately proven in a case (critical facts)
-The facts that are at the core of the trial
-Example: "The defendant is guilty of murder."

(top node)
● Final facts
└── A is guilty of murder.

(middle node)
● Intermediate facts
├── A stabbed B with a weapon.
├── B died because of it.
├── A had the intent to kill.

(subnode)
● Evidence
├── CCTV footage: A is seen stabbing B with a weapon
├── Doctor's opinion: B was stabbed to death
├── Witness statement: "A swung a weapon at B."

(social convention)
● Social conventions
├── "If a weapon is used, it is usually considered to be intent to kill."
├── "If a weapon is used without a serious fight, it is likely a planned crime."

At this time, it is assumed that social conventions are pre-database. Also, when configuring each node of the tree structure, various models can be used as if the facts were identified, and if organized, they can be as shown in Table 10. However, the models are not limited to this.

AI models and techniques for tree building AI model role Large language models (LLM, GPT-4, Claude, Mistral, etc.) Natural language understanding and legal document analysis Legal expert language models (LexiLaw, CaseLawBERT, KoLegalQA, etc.) Optimizing legal document and case law search Named Entity Recognition (NER) Extracting people, dates, and legal provisions from the incident Knowledge Graph (KG) Modeling relationships between legal concepts Rule-based Reasoning (RLHF) Application of legal logic and case-based reasoning Retrieval-Augmented Generation (RAG) Reflects the latest legal data and case law searches

For example, by combining AI models such as ① a large language model (GPT-4, Claude) to analyze facts, ② named entity recognition (NER) to extract case information (people, contracts, dates, etc.), ③ a legal knowledge graph (Neo4j) to model legal relationships, and ④ augmented search generation (RAG) to reflect the latest legal information and case law, a logical legal inference tree like a human lawyer can be automatically generated. Of course, the method of combining each model is not limited to this.

<Legal Advice Expert Matching>

The expert matching unit (398) classifies whether the query is for seeking advice or information using a pre-established automatic classification model when a query is made from a user terminal (100), and if the query is for seeking advice, extracts a list of legal experts whose fields of expertise are in the category of the search results corresponding to the query, and sorts the list of legal experts in order of high matching rate based on demographic data of the user terminal (100) and then provides it to the user terminal (100). Recently, there was a case where an LLM-based generative AI that provides legal advice stopped providing services because it violated the Attorney-at-Law Act.

Accordingly, in one embodiment of the present invention, the user's intention is classified as to whether he or she wants to conduct a legal search or to seek legal advice, and if he or she wants to seek legal advice, ethical and legal issues may arise, so legal experts corresponding to this are extracted, and then the legal experts are arranged in order of preference, that is, those that the user is likely to like, and provided to the user terminal (100).

At this time, if excessive personal information is entered into the LLM-based generative AI, an order can be given to save the query by pseudonymization or anonymization based on GDPR. In order to match a preferred legal professional, demographic data such as the user's age, education, residence, and occupation can be extracted, and the characteristics (features) of the legal professional preferred by the client with this demographic data can be extracted, and then the legal professional with these characteristics can be sorted and matched. Preferably, a rule-based recommendation system can be applied first, and then AI-based recommendations can be utilized as data accumulates.

Recommended method type explanation yes Rule-Based Recommendation System Establish rules for recommending lawyers based on specific demographic characteristics - Users in their 20s and 30s tend to prefer young lawyers → Recommend lawyers of similar age
-When a female victim requests counseling for a sexual crime case, priority is given to recommending a female lawyer. Content-Based Filtering Analyzes the user's past lawyer selection history (reviews, consultation records) to recommend lawyers with similar characteristics Recommendations based on the lawyer's area of expertise, consulting style, ratings, reviews, etc.
-If the user preferred [Divorce Attorney B], recommend an attorney with similar experience and style to Attorney B. Collaborative Filtering (CF) Recommend lawyers by leveraging the selection patterns of other similar users -If a middle-class man in his 40s prefers a particular lawyer, he will also recommend him to users with similar backgrounds.
- Recommend lawyers preferred by users in the same area (Gangnam, Seoul) Hybrid Recommendation System Provides optimized recommendations by combining rule-based + content-based + collaborative filtering - When a female user in her 30s requests divorce consultation → A female lawyer is recommended first (rule-based)
-Recommendation of lawyers with similar experience to those previously used (content-based)
-Recommendations for lawyers chosen by other users with similar backgrounds (collaborative filtering) Recommendation system using machine learning/AI Data collection & preprocessing - Collect user demographic information, search history, and consultation history
- Organize data on lawyers’ careers, areas of expertise, and client reviews Applying machine learning models -Decision Tree, Random Forest → Rule-based classification
- Deep Learning (Neural Network Model) → Large-scale data-based recommendation optimization
- Natural Language Processing (NLP) → Analysis of user reviews and lawyer profiles Predictions and Recommendations -When you enter the user's characteristics, AI predicts and recommends the best lawyer
-Continuously learn recommendation results to gradually improve recommendation accuracy

Hereinafter, the operation process according to the configuration of the platform service providing server of the above-described Fig. 2 will be described in detail with reference to Figs. 3 and 4 as examples. However, it will be obvious that the embodiment is only one of various embodiments of the present invention and is not limited thereto.

Referring to FIG. 3a, (a) the platform service providing server (300) learns a dataset of everyday terms and legal terms using an artificial intelligence algorithm so that when an everyday term is entered as a query, the corresponding legal term can be output, (b) by providing legal navigation so that the keyword extracted from the query can be visually identified as to which legal term the keyword is related to and which legal article the legal term is based on, (c) by providing a related keyword map so that intuitive identification is possible, and (d) by using generative AI so that an answer to the user's query can be generated and provided. Since FIG. 3b has been described with reference to FIG. 2, redundant description will be omitted.

One embodiment of the present invention provides integrated provision of legal information as shown in FIG. 4a, and also provides legal navigation that provides the same visually. As shown in FIG. 4b, even if a search is made using everyday terms, a corresponding search result can be received by using a legal knowledge base, and as shown in FIG. 4c, it displays which legal provisions are related by associating them with everyday (life) terms and legal (legal) terms, and as shown in FIG. 4d, it performs everyday term and sentence-type meaning-based search, derives candidate keywords based on the frequency of occurrence of search keywords, and provides an intelligent search that analyzes similarity, thereby providing legal navigation of [everyday terms-legal terms-legal provisions-legislation], and can also provide related search terms related thereto. Currently, a project as shown in FIG. 4e is in progress, and a user can input a query and receive a response at a national law information center as shown in FIG. 4f. Figures 4g to 4k are drawings added as basic concepts for indexing, and Figures 4k and 4l illustrate a part of an open API provided by the Ministry of Government Legislation.

Matters not described in the method for providing an intelligent law information service platform service using the law navigation of FIGS. 2 to 4 are identical to or can be easily inferred from the contents described in the method for providing an intelligent law information service platform service using the law navigation of FIG. 1 above, and therefore, description thereof will be omitted below.

FIG. 5 is a diagram showing a process of transmitting and receiving data between each component included in the intelligent law information service platform providing system using the law navigation of FIG. 1 according to one embodiment of the present invention. Hereinafter, an example of a process of transmitting and receiving data between each component will be described through FIG. 5, but the present invention is not limited to this embodiment, and it is obvious to those skilled in the art that the process of transmitting and receiving data shown in FIG. 5 may be changed according to various embodiments described above.

Referring to FIG. 5, the platform service providing server constructs a legal knowledge base by mapping a legal term DB and a legal term DB to define the relationship between legal terms and legal terms (S5100), and when a query is entered in a user terminal using legal terms, it extracts search words from the query and analyzes the morphemes of the search words (S5200).

Then, the platform service providing server transmits the analysis result of morpheme analysis to the law knowledge base (S5300), and if a law term corresponding to a search word is extracted from the law knowledge base, it searches for index information from the previously constructed law information index data (S5400).

Finally, the platform service providing server returns search results including index information to the user terminal (S5500).

The order of the above-described steps (S5100 to S5500) is only an example and is not limited thereto. That is, the order of the above-described steps (S5100 to S5500) may be mutually changed, and some of the steps may be executed simultaneously or deleted.

Matters not described in the method for providing an intelligent law information service platform service using the law navigation of FIG. 5 are identical to or can be easily inferred from the contents described in the method for providing an intelligent law information service platform service using the law navigation through FIGS. 1 to 4 above, and therefore, description thereof will be omitted below.

The method for providing an intelligent law information service platform service using law navigation according to one embodiment described through FIG. 5 can also be implemented in the form of a recording medium including computer-executable commands, such as an application or program module executed by a computer. The computer-readable medium can be any available medium that can be accessed by a computer, and includes both volatile and nonvolatile media, removable and non-removable media. In addition, the computer-readable medium can include all computer storage media. The computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented by any method or technology for storing information, such as computer-readable commands, data structures, program modules, or other data.

The method for providing an intelligent law information service platform service using law navigation according to one embodiment of the present invention described above can be executed by an application that is installed by default in a terminal (which may include a program included in a platform or operating system that is installed by default in the terminal), and can also be executed by an application (i.e., a program) that a user directly installs in a master terminal through an application providing server such as an application store server, an application, or a web server related to the corresponding service. In this sense, the method for providing an intelligent law information service platform service using law navigation according to one embodiment of the present invention described above can be implemented by an application (i.e., a program) that is installed by default in a terminal or directly installed by a user, and can be recorded on a computer-readable recording medium such as a terminal.

The above description of the present invention 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 characteristics of the present invention. Therefore, it should be understood that the embodiments described above are exemplary in all respects and not restrictive. For example, each component described as a single component may be implemented in a distributed manner, and likewise, components described as distributed may 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 modifications derived from the meaning and scope of the claims and their equivalent concepts should be interpreted as being included in the scope of the present invention.

Claims (7)

A user terminal that inputs a query in everyday language and receives and outputs a response to the query; and
A construction unit that constructs a legal knowledge base by mapping and storing an everyday term database (DataBase) and a legal term database to define the relationship between everyday terms and legal terms, an analysis unit that extracts a search term from a query entered in everyday terms from the user terminal and analyzes the morpheme of the search term, a transmission unit that transmits the analysis result of the morpheme analysis to the legal knowledge base, a query unit that searches for index information from the established legal information index data when a legal term corresponding to the search term is extracted from the legal knowledge base, a provision unit that returns the search result including the index information to the user terminal, after collecting case law from a national law information provision server, extracts the upper law of the article referenced more than a preset number of times among the articles that appear as the basis law in the case law, and constructs a dataset with pairs of questions and answers in a triplet structure based on the law of the upper law, while using GPT-4 to configure the question as a case-based question based on the law so that an actual case is entered within the prompt. A case-based question composition unit that provides examples of base questions in Few-Shot format; When inputting facts in everyday terms on the user terminal, when performing instruction tuning on an LLM (Large Language Model)-based generative AI (Artificial Intelligence), a dataset of instructions, facts, and elements is constructed, and instruction tuning is performed on the LLM-based generative AI; and in the inference step of the LLM-based generative AI, when providing an inference result, a component extraction unit that searches for similar case data including the facts and elements using RAG (Retrieval-Augmented Generation) and provides the inference result together with the similar case data; After collecting case laws from a national law information provision server, deriving facts and laws from the case laws, and mapping the burden of proof according to the laws to the facts and laws and storing them, thereby creating a dataset. A proof burden guidance unit that provides guidance on laws and burden of proof when the above facts are input at the inference stage, a legal reasoning unit that generates a tree structure by arranging evidence and social conventions as lower nodes, intermediate facts as intermediate nodes, and final facts that must be ultimately proven in a case as upper nodes to perform law reasoning from the above facts to the final facts, and provides it to the user terminal, a diagramming unit that provides a keyword map that visualizes the relationship between everyday terms and legal terms using nodes and edges based on the relationship between everyday terms and legal terms that has already been built, a statutory term that is mapped to at least one statutory provision and stored, at least one statutory provision is mapped to at least one other statutory provision and stored, and at least one relationship between statutes is mapped and stored, and the relationship between statutory terms and statutory provisions, the relationship between at least one statutory provision, and at least A platform service providing server including a legal system visualization unit that provides legal navigation that visualizes the legal system including the relationship between everyday terms and legal terms and between legal provisions and laws by reflecting the relationship between one law on the related keyword map; a ranking presentation unit that guides the legal system according to the ranking between the general law and the special law when at least one legal provision and at least one other legal provision are in a relationship between general laws and special laws, and gives priority to outputting and guiding the legal provisions according to the special law that is applied preferentially; an expert matching unit that classifies, when a query is made from the user terminal, whether the query is for seeking advice or information using a pre-established automatic classification model, and, when the query is for seeking advice, extracts a list of legal experts whose fields of expertise are in the category of search results corresponding to the query, and sorts the list of legal experts in order of high matching rates based on demographic data of the user terminal and then provides the list to the user terminal;
Including,
The above component extraction part is,
In order for the general public who cannot understand the elements of the constitution that each law has, or those who have legal knowledge but have not gained experience in an actual law office or law firm and thus have difficulty judging the elements of each law to be able to understand the elements of the law, in the structure of instructions and input and output, the instructions are tuned so that the elements of the law are set as the instructions, the input is the facts, and the output is the elements of the law.
When asked whether sending bricks from Junggonara constitutes fraud, the company provides information on the components of the transaction, including information on whether the act of sending bricks constitutes obtaining property benefits through deception and causing damage.
As for the above component description information, in the case of fraud, the elements of the crime must be established as deception, acquisition of property benefits, and occurrence of damage. In a second-hand transaction, deceiving someone by saying that they are selling a normal product constitutes deception, sending bricks or worthless products instead constitutes damage, and an act of deceiving a buyer and taking money can be recognized as acquisition of property benefits.
Considering that the LLM-based generative AI, which has instruction tuning to generate responses according to the model builder's request, may not be able to provide accurate answers, the similar case data is searched and provided based on facts and elements, and when searching for similar case data similar to facts, the Korean embedding model jhgan/ko-sroberta-multitask based on Sentence BERT is used to identify the similarity between sentences.
In the above-mentioned related keyword map provided in the above-mentioned diagram section, search terms are displayed as nodes in yellow circles, legal terms related to the search terms are displayed as nodes in blue square boxes, and everyday terms related to legal terms are displayed as nodes in white circles.
The above legal system visualization department is,
An intelligent legal information service platform provision system using legal navigation, wherein when visualizing the legal system, in the above-mentioned related keyword map, three laws corresponding to the legal terms corresponding to housing lease are listed: the Housing Lease Protection Act, the Special Act on Private Rental Housing, and the Civil Act. However, in terms of the order of the relationship between laws, the Housing Lease Protection Act is listed first among the three laws, followed by the Special Act on Private Rental Housing, and lastly, the Civil Act. delete delete delete delete delete delete

Images