KR20230051652A - Systems and Methods to Provide AI Answering Service for Legal Questions - Google Patents

Abstract

A system for providing an AI digitized answering service for a legal question according to one embodiment of the present invention, which enables a user to receive an answer for legal knowledge or legal application in a gauge form including a positive or negative numerical value, comprises: a question terminal that requests legal advice on question content; and an AI legal/question answering server that collects case precedent information, in which a case similar to the question content is applied, and legal counsel information of layers from at least one external server to process the same in a pre-set form and then be constructed as a database, and analyzes a positive or negative legal answer for the question content by using an artificial intelligence machine learning algorithm based on the constructed case precedent information and legal counsel information to provide the positive or negative legal answer as a numerical value. The question terminal executes a question answering platform that displays the numerical value in a gauge form for a user.

Description

Systems and Methods to Provide AI Answering Service for Legal Questions}

The present invention relates to a system and method for providing a digitized AI answer service to legal questions.

Consultation is basically required in the legal field, but at present, most of them are limited to providing various means or forms (or forms) to connect experts and counseling and proceeding in the form of preliminary investigation.

This standardizes counseling and checks and processes various stages and conditions in the ARS method, so the service progress is very tedious and the service experience is poor, which is a very big obstacle to the automated processing of the service. However, if consultation is made possible through natural language processing, the precision of consultation will increase according to machine learning, and related industries can be revitalized without time and space limitations.

Accordingly, the present invention intends to disclose a system and method capable of providing answers to users' legal questions using an AI learning algorithm.

Publication No. 10-2019-0143583 (Title of Invention: A method for providing answers based on a hybrid instrumental dialog flow model applying machine learning to dialog management technology by a conversation understanding AI system and a computer readable recording medium )

The object of the present invention is to provide a system and method for providing a digitized AI answer service to legal questions that can solve conventional problems.

A system for providing a digitized AI answer service to legal questions according to an embodiment of the present invention to solve the above problems is an inquiry terminal requesting legal advice on the inquiry: and a case to which an event similar to the above inquiry is applied. After collecting case precedent information and legal counseling information from at least one external server, processing it in a preset form and constructing a database, affirming and An AI legal/question answering server that analyzes legal answers to injustice using artificial intelligence machine learning algorithms and provides them as numerical values of positive and negative values, and the inquiry terminal provides the numerical values to the user in the form of gauges. It is characterized in that it executes a question and answer platform displayed to.

In one embodiment, the AI legal/question response server converts the case precedent information and legal consultation information collected from the at least one external server into a question and answer format, and then sets the inclination of the answer to 'positive (positive)'. and 'negative (negative)', characterized in that the database is constructed.

In one embodiment, when the AI law/question answering server builds the questions and answers into a database, at least one or more of the raw data for the questions and answers, the category for the question, the case number, and the underlying law It is characterized by being built to include.

In one embodiment, the AI legal/question response server is characterized in recognizing the user's query content as either positive or negative.

In one embodiment, the AI legal/question answering server calculates sentence vectors of a plurality of questions and answers constructed as a database using a sentence vector similarity algorithm, and then selects 10 sentences (questions and answers) having the highest similarity between sentences. ) is selected, and the similarity felt by humans for the selected sentences is defined as '0' or '1', and then the defined sentences are learned using a natural language processing learning algorithm (BERT NLP), and the learned data It is characterized in that the learning data is derived by re-learning.

In one embodiment, the AI legal/question response server is characterized in that it extracts a semantic analysis, a negative or positive sense, and a preset category for the content of the query using a natural language processing algorithm.

In one embodiment, the AI law/question response server is characterized in that it processes the sentence content (meaning) in the query content into multi-dimensional space vector coordinates using a natural language processing learning algorithm (BERT NLP).

In one embodiment, the AI legal/question response server extracts case precedent information and legal consultation information including categories and multidimensional space vector coordinates similar to the categories of the inquiry contents and multidimensional space vector coordinates from the database. .

In one embodiment, the AI legal/question response server is characterized in that it provides an average value of positive and negative categories (numerical values) of at least one extracted case precedent information and legal consultation information as the numerical value.

In one embodiment, the AI legal/question response server has a positive or negative numerical value among the average values of positive and negative categories (numeric values) of at least one case case precedent information and legal consultation information extracted relative to a predetermined value or more. When calculated as , it is characterized in that it provides a numerical value reflecting the weight.

A method for providing a digitized AI answer service for legal questions according to an embodiment of the present invention for solving the above problems includes requesting legal advice on inquiry contents from an inquiry terminal; And in the AI legal/question response server, case precedent information and legal counseling information of lawyers to which cases similar to the above query were applied are collected from at least one external server, processed in a preset form, and built into a database. Based on case precedent information and legal consultation information, analyzing legal answers to positive and negative questions about the contents of the inquiry using an artificial intelligence machine learning algorithm, and then providing them as numerical values, wherein the inquiry terminal includes the numerical values It is characterized by executing a question answering platform that displays a value to a user in the form of a gauge.

Using the system and method for providing a digitized AI answer service to legal questions according to an embodiment of the present invention, users can easily inquire about themselves or legal knowledge or legal application of cases that have occurred to them. There is an advantage in that answers to can be provided in the form of gauges containing positive and negative values, and similar precedents and cases can be provided.

1 is a network configuration diagram of a system that provides a digitized AI answer service to legal questions according to an embodiment of the present invention.
2 is a detailed configuration diagram of the AI legal/question response server shown in FIG.
FIG. 3 is an exemplary view showing a data construction form of the database shown in FIG. 1 .
4 to 6 are exemplary diagrams for explaining a natural language processing learning algorithm (BERT NLP).
7 and 8 are exemplary diagrams for explaining an execution process of the query and similarity analysis/extraction unit shown in FIG. 2 .
FIG. 9 is an exemplary view of a gauge bar in which a numerical value provided by a result providing unit shown in FIG. 2 is reflected in a query terminal.
10 is a flowchart of a method of providing a digitized AI answer service to legal questions according to an embodiment of the present invention.
FIG. 11 is a detailed flowchart of the process S720 shown in FIG. 10 .
12 is an illustration of an example computing environment in which one or more embodiments set forth herein may be implemented.

Hereinafter, embodiments of the present invention will be described in detail so that those skilled in the art can easily practice the present invention with reference to the accompanying drawings. However, the present invention may be embodied in many different forms and is not limited to the embodiments described herein. And in order to clearly explain the present invention in the drawings, parts irrelevant to the description are omitted, and similar reference numerals are attached to similar parts 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 interposed therebetween. . In addition, when a part "includes" a certain component, this means that it may further include other components, not excluding other components, unless otherwise stated, and one or more other characteristics. However, it should be understood that it does not preclude the possibility of existence or addition of numbers, steps, operations, components, parts, or combinations thereof.

As used throughout the specification, the terms "about", "substantially", etc., are used at or approximating that value when manufacturing and material tolerances inherent in the stated meaning are given, and do not convey an understanding of the present invention. Accurate or absolute figures are used to help prevent exploitation by unscrupulous infringers of the disclosed disclosure. The term "step of (doing)" or "step of" as used throughout the specification of the present invention does not mean "step for".

In this specification, a "unit" includes a unit realized by hardware, a unit realized by software, and a unit realized using both. Further, one unit may be realized using two or more hardware, and two or more units may be realized by one hardware. On the other hand, '~ unit' is not limited to software or hardware, and '~ unit' may be configured to be in an addressable storage medium or configured to reproduce one or more processors. Thus, as an example, '~unit' refers to components such as software components, object-oriented software components, class components, and task components, processes, functions, properties, and procedures. , subroutines, segments of program code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays and variables. Functions provided within components and '~units' may be combined into smaller numbers of components and '~units' or further separated into additional components and '~units'. In addition, components and '~units' may be implemented to play one or more CPUs in a device or a secure multimedia card.

In this specification, some of the operations or functions described as being performed by a terminal, device, or device may be performed instead by a server connected to the terminal, device, or device. Likewise, some of the operations or functions described as being performed by the 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 mapping or matching with the terminal mean mapping or matching the terminal's unique number or personal identification information, which is the terminal's identifying data. can be interpreted as

Hereinafter, based on the accompanying drawings, a system and method for providing a digitized AI answer service to legal questions according to an embodiment of the present invention will be described in more detail.

1 is a network configuration diagram of a system that provides a digitized AI answer service for legal questions according to an embodiment of the present invention, and FIG. 2 is a detailed configuration diagram of the AI legal/question response server shown in FIG. 1, 3 is an exemplary diagram showing a data construction form of the database shown in FIG. 1, FIGS. 4 to 6 are exemplary diagrams for explaining a natural language processing learning algorithm (BERT NLP), and FIGS. 7 and 8 are diagrams. 2 is an exemplary diagram for explaining the execution process of the query and similarity analysis/extraction unit, and FIG. 9 is an exemplary diagram of a gauge bar reflecting numerical values provided by the result providing unit shown in FIG. 2 in a query terminal. .

First, as shown in FIG. 1, the system 100 for providing a digitized AI answer service to legal questions according to an embodiment of the present invention includes a query terminal 200 and an AI legal/question response server 300 includes

However, since the system 100 for providing a digitized AI answer service to legal questions in FIG. 1 is only an embodiment of the present invention, the present invention is not limited to interpretation through FIG. 1 .

At this time, each component of FIG. 1 is generally connected through a network. For example, as shown in FIG. 1 , at least one query terminal 200 may be connected to the AI legal/question response server 300 through a network.

Here, the network means a connection structure capable of exchanging information between each node, such as a plurality of terminals and servers, and examples of such networks include RF, 3rd Generation Partnership Project (3GPP) network, Long Term LTE Evolution) network, 5GPP (5th Generation Partnership Project) network, WIMAX (World Interoperability for Microwave Access) network, Internet, LAN (Local Area Network), Wireless LAN (Wireless Local Area Network), WAN (Wide Area Network) , PAN (Personal Area Network), Bluetooth (Bluetooth) network, NFC network, satellite broadcasting network, analog broadcasting network, DMB (Digital Multimedia Broadcasting) network, etc. are included, but are not limited thereto.

In the following, the term at least one is defined as a term including singular and plural, and even if at least one term does not exist, each component may exist in singular or plural, and may mean singular or plural. It will be self-evident. In addition, the singular or plural number of each component may be changed according to embodiments.

At least one query terminal 200 may be a terminal that inputs a search keyword and receives a search result using a web page, app page, program, or application related to a digitized AI answer service for legal questions. In addition, at least one query terminal 200 may be a terminal that receives positive and negative answers to search results as numerical information and then displays them in the form of a gauge.

Here, at least one query terminal 200 may be implemented as a computer capable of accessing a remote server or terminal through a network. Here, the computer may include, for example, a laptop, a desktop, a laptop, and the like equipped with a navigation system and a web browser. In this case, at least one user terminal 100 may be implemented as a terminal capable of accessing a remote server or terminal through a network. At least one user terminal 100 is, for example, a wireless communication device that ensures portability and mobility, 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 ) may include all kinds of handheld-based wireless communication devices such as terminals, smartphones, smart pads, tablet PCs, and the like.

Next, the AI legal/question response server 300 collects case precedent information and legal counseling information of lawyers to which cases similar to the above query were applied, from at least one external server, and then processes them in a preset form to create a database. After analyzing the positive and negative legal answers to the above inquiries using an artificial intelligence machine learning algorithm based on the case precedent information and legal consultation information, each positive and negative numerical value is provided. It may be a configuration that

The AI legal/question response server 300 may be a server that provides digitized AI answer service web pages, app pages, programs or applications for legal questions. The AI legal/question response server 300 detects the form of the query content and keywords of words or sentences included in the query content, and extracts a preset number of precedent information and legal consultation information having a high degree of correlation or similarity with the keywords. Afterwards, it may be configured to provide the extracted precedent information and the result of the answer (positive or negative) of the legal consultation information as numerical information.

Here, the AI law/question response server 300 may be implemented as a computer capable of accessing a remote server or terminal through a network. Here, the computer may include, for example, a laptop, a desktop, a laptop, and the like equipped with a navigation system and a web browser.

Referring to FIG. 2, the AI law/question response server 300 includes a collection unit 310, a pre-processing unit 320, a database building unit 330, a machine learning unit 340, a natural language processing unit 350, a query and similarity It includes an analysis/extraction unit 360, a positive/negative digitization unit 370, and a result value providing unit 380.

The AI law/question response server 300 according to an embodiment of the present invention or another server (not shown) operating in conjunction with at least one query terminal 200 and at least one information providing server (not shown) In the case of transmitting a digitized AI answer service application, program, app page, web page, etc. to legal questions, at least one query terminal 200 installs or opens the corresponding service application, program, app page, web page, etc. can Also, a service program may be run in at least one query terminal 200 using a script executed in a web browser.

Here, the web browser is a program that enables the use of web (WWW: world wide web) services and means a program that receives and displays hypertext described in HTML (hyper text mark-up language). For example, Netscape , Explorer, Chrome, etc. In addition, an application means an application on a terminal, and includes, for example, an app running on a mobile terminal (smart phone).

Referring to FIG. 2 , the collection unit 310 may be configured to collect precedent raw data and legal consultation (Q&A) data from at least one external server.

Here, the external server may be a server of the National Penalty Information Center ( www.law.go.kr ) of the Republic of Korea, and the precedent raw data may be precedent big data of district courts, high courts, and the Supreme Court. In addition, legal advice (Q&A) information includes legal information based on lawyers' answers and answers to precedents.

The collection unit 310 may collect precedent information and legal counseling information using the Selenium algorithm. At this time, the collection unit 310 can apply any program that can collect precedent information and legal counseling information by scraping.

The pre-processing unit 320 converts raw data and legal counseling data of case precedent information into an index, question, answer, category, case number, ground law, precedent result, and counseling answer into positive/negative categories. It can be a process of sorting/sorting/editing.

Here, the pre-processing unit 320 may convert and process case precedent data in the form of a question (Q) and an answer (A) using a natural language processing algorithm.

For example, ordinary precedent data consists of many parts such as <judgment matter>, <judgment gist>, <reference clause>, and <reference precedent>, and the contents are also complex. Therefore, in the present invention, only necessary parts of precedent data are extracted and preprocessed (categorized) so that future similarity analysis and numerical answers can be provided to users.

That is, the pre-processing unit 320 classifies the sense of the judgment sentence into 'positive' and 'negative' based on the content of the issue of <judicial matters> among the precedent data, and digitizes it.

For example, in the precedent data provided by the National Legal Information Center, there is a consistent item called <Judgement>, and issues in the form of [1] and [2] appear under that item.

For example, the number from <Judgement> to [3] means that there are three issues in the case. Each number shows what the issue is and the judge's decision on that issue, marked affirmatively (“positive”) and negatively (“negatively”).

Next, the legal database unit 330 classifies/sorts/categorizes positive/negative tagging of the preprocessed index, question, answer, question tagging category, case number, ground law, precedent result, and counseling answer preprocessed in the preprocessing unit. It can be a configuration that you edit and build into a database.

Here, the index may be a unique number (unique address) given to data registered in the database.

A question is the raw data of a natural language question. The answer may be an answer provided by an attorney to the user's question.

The tagging category is a question-related association keyword, and may be, for example, lease, retirement, or dismissal. The grounding law may be the grounding law for the lawyer's answer to the case.

Negative/positive tagging is tagging information that determines whether the user's question form ends in a positive or negative form.

The negative/positive tagging is applied to numerical values to be presented to the user. For example, if the question is 'Is it applicable to defamation?', it is a positive type, and if 'Is it not applicable to defamation?', it is a negative type. This kind of tagging is because 20% of positive questions and 20% of negative questions in numerical answers have different meanings.

Therefore, when the numerical answer of a negative question is replaced with a positive one later, (100 - numeric answer) = may be information for replacing the numerical answer.

Next, the machine learning unit 340 calculates sentence vectors of a plurality of questions and answers constructed as a database using a sentence vector similarity algorithm, and then selects 10 sentences (questions and answers) having the highest similarity between sentences. After defining the sensory similarity felt by humans as '0' or '1' for the selected sentences, learning the defined sentences using a natural language processing learning algorithm (BERT NLP) and relearning the learned data It performs the function of deriving learning data.

That is, the machine learning unit 340 may obtain a degree of similarity between embedded sentence vectors based on cosine similarity and Jaccard similarity. In addition, 10 sentences having the highest similarity between sentences may be selected, but the present invention is not limited thereto.

In addition, data for learning BERT (Bidirectional Encoder Representations from Transformers) NLP (Natural Language Processing) by manually tagging 0 (similar) and 1 (dissimilar) for the similarity felt by humans for the selected 10 sentences After generating, it operates to learn BERT NLP.

For reference, BERT is designed to resolve ambiguity by making the same word have different embeddings depending on the sentence form and position. For example, banks of “account” and “of the river” have the same vectors in Word2Vec or GloVe. However, this word should have a completely different meaning depending on the context, and in fact it has a completely different vector in ELMo or BERT. Examples shown in the drawing are 1) I love you 2) I loved you 3) Love Who Said. When the cosine distance is checked using the morphological analysis results for these three sentences as input values, the word embedding value for the position of 'love' is identically taken and the cosine distance is output. Among them, the 'love' vectors of No. 1 and No. 2, which have similar sentences, appear at a very close distance. On the other hand, in the case of number 3, which has a completely different sentence, the distance is marked as very far even though it is the same 'love'.

As another example, BERT may provide similar meaning as a vector distance through cosine similarity between sentence or word vectors.

For example, when measuring the distance from the first 'love' vector with the simple sum of the second and third 'love' vectors, the distance between the two vectors sum is about 0.17. Among them, look at the distance with the Weighted Sum, which weights the vector 2, which has a short distance. The distance has gotten much closer. Conversely, try to assign weights to vector 3, which is far away. You can see it goes farther away. Since cosine distance measures an angle, the distance remains the same even if the vector is increased. More detailed information can be found in the meaning of cosine similarity.

Next, the natural language processing unit 350 may be a component that analyzes the meaning of the query content, extracts a negative or positive sense, and a preset category using a natural language processing algorithm.

The natural language processing unit 350 performs a meaning analysis vectorization process, a negative/positive tagging process, and a category tagging process of the question in analyzing the content of the question.

Here, the semantic analysis vectorization process may be a process of processing the corresponding query content with vector coordinates in a multidimensional space identified based on the meaning of the question. At this time, natural language processing algorithms such as Universal Sentence Encoder and BERT are used. Through this, questions with similar meanings can have unique coordinates with a close distance.

Next, the negative/positive tagging process is a process of tagging after determining whether the user's question ends in a positive or negative form.

For example, if the question is 'Is it applicable to defamation?', it is a positive type, and if 'Is it not applicable to defamation?', it is a negative type. This kind of tagging is because 20% of positive questions and 20% of negative questions in numerical answers have different meanings. Therefore, if the numerical answer of a negative question is replaced with a positive one in the future, (100 - numerical answer) = the replaced numerical answer.

Next, the category tagging process is a process of tagging keywords related to the question content, and may be, for example, a process of tagging field keywords related to the question content or key keywords such as leasing, retirement, and dismissal.

Next, the query and similarity analysis/extraction unit 360 may be a component that compares and analyzes categories and multidimensional space vector coordinates similar to the categories and multidimensional space vector coordinates of the query content with information in the database.

That is, the query and similarity analysis/extraction unit 360 first extracts precedent information and legal consultation information having a category similar to that of the query content from the database, and then, after first extracting the query content, has a similar multidimensional spatial vector coordinate (the coordinate distance is the largest). It may be a configuration of secondarily extracting information having multidimensional space vector coordinates from firstly extracted information.

Meanwhile, the query and similarity analysis/extraction unit 360 may extract legal cases related to the query contents through the above-described process.

The positive/negative digitization unit 370 calculates an average value of positive and negative categories (numerical values) for each of the at least one case precedent information and legal consultation information extracted from the query and similarity analysis/extraction unit 360. can be config.

The positive/negative digitization unit 370 is configured when the positive or negative numerical value among the average values of the positive and negative categories (numeric values) of the extracted at least one case case precedent information and legal counseling information is calculated to be relatively higher than a preset value. , provides a numerical value reflecting the weight.

The result value providing unit 380 provides the positive and negative numerical values calculated by the positive/negative digitizing unit 370 as a graph in the form of a gauge.

The result value providing unit 380 may provide it through a non-linear model by using the numerical values of several cases as input values, similar to the method of using stacking in the final decision-making in the ensemble technique.

10 is a flowchart of a method for providing a digitized AI answer service for legal questions according to an embodiment of the present invention, and FIG. 11 is a detailed flowchart of the S720 process shown in FIG.

The method of providing a digitized AI answer service to legal questions shown in FIG. 10 can be implemented through the system shown in FIG.

10 and 11, in the method of providing a digitized AI answer service to legal questions according to an embodiment of the present invention (S700), first, the inquiry terminal 200 digitized AI answers to legal questions Using the service-related web page, app page, program, or application, request legal advice or legal answers to the AI legal/question-answering server (S710).

Here, the web page, app page, program, or application related to the digitized AI answer service to the legal question executed in the query terminal 200 uses the numerical value of the legal answer provided from the AI legal/question response server described later in the form of a gauge. can be displayed to the user.

When the inquiry contents are received, the AI legal/question response server 300 collects case precedent information and legal consultation information of lawyers to which a similar legal case to the above inquiry contents is applied, from at least one external server, and then converts them into a preset form. It is processed and built into a database, and based on the built case case precedent information and legal consultation information, legal answers to positive and negative questions are analyzed using artificial intelligence machine learning algorithms, and then provided as numerical values ( S720).

More specifically, in the S720 process, after converting the case precedent information and legal consultation information collected from the at least one external server into a question and answer form, the inclination of the answer is 'positive (active)' and 'negative (negative)' )', and building the database into the database.

In addition, when building the database with the questions and answers, a process of constructing the database to include at least one of raw data for the questions and answers, categories for the questions, case numbers, and supporting laws may be included.

In addition, a process of determining one of positive and negative feelings about the query contents of the user may be included.

In addition, after calculating sentence vectors of a plurality of questions and answers constructed as a database using a sentence vector similarity algorithm, after selecting 10 sentences (questions and answers) with the highest similarity between sentences, After defining the sensory similarity felt by humans as '0' or '1', learning the defined sentences using the natural language processing learning algorithm (BERT NLP), and relearning the learned data to derive the learning data. can do.

In addition, a process of analyzing the meaning of the query content, extracting a negative or positive sense, and a preset category using a natural language processing learning algorithm (BERT NLP) may be included.

In addition, a process of processing the sentence content (meaning) in the query content into multi-dimensional space vector coordinates using the natural language processing learning algorithm (BERT NLP) may be included.

In addition, a process of extracting case case precedent information and legal consultation information including the categories of the query contents and categories similar to the multidimensional space vector coordinates and multidimensional space vector coordinates may be included from the database.

In addition, a process of providing an average value of positive and negative categories (numerical values) of at least one or more extracted case precedent information and legal counseling information as the numerical value may be included.

In addition, if the positive or negative numerical value among the average values of the positive and negative categories (numeric values) of the extracted at least one case case precedent information and legal consultation information is relatively calculated to be more than a preset value, a numerical value reflecting the weight is provided. It may further include the process of

More specifically, the S720 process is a process executed in the AI legal/question response server 300, and the collection unit 310 collects precedent raw data and legal consultation (Q & A) data from at least one external server in Korea Collected from an external server such as the server of the National Penalty Information Center ( www.law.go.kr ). In this case, the precedent raw data may be precedent big data of the district court, the high court, and the Supreme Court. In addition, legal advice (Q&A) information includes legal information based on lawyers' answers and answers to precedents.

Then, in the pre-processing unit 320, raw data and legal counseling data of case precedent information are indexed, questions, answers, categories, case number, ground law, precedent results, and counseling answers are categorized as positive/negative. The process of classifying/sorting/editing and converting case precedent data into the form of questions (Q) and answers (A) using natural language processing algorithms is performed.

When the operation of the pre-processing unit 320 is completed, the machine learning unit 340 calculates sentence vectors of a plurality of questions and answers constructed as a database using a sentence vector similarity algorithm, and then selects 10 sentence vectors having the highest similarity between sentences. Sentences (questions and answers) are selected, and the similarity felt by humans for the selected sentences is defined as '0' or '1', and then the defined sentences are learned using a natural language processing learning algorithm (BERT NLP), , It performs the function of deriving learning data by re-learning the learned data.

That is, the machine learning unit 340 may obtain a degree of similarity between embedded sentence vectors based on cosine similarity and Jaccard similarity. In addition, 10 sentences having the highest similarity between sentences may be selected, but the present invention is not limited thereto.

In addition, data for learning BERT (Bidirectional Encoder Representations from Transformers) NLP (Natural Language Processing) by manually tagging 0 (similar) and 1 (dissimilar) for the similarity felt by humans for the selected 10 sentences After generating, it operates to learn BERT NLP.

Next, when the natural language processing unit 350 analyzes the meaning of the query content, extracts a negative or positive sense, and analyzes the question content using a natural language processing algorithm, a process of analyzing the meaning of the question and vectorizing the question content , negative/positive tagging process and category tagging process are performed.

Here, the semantic analysis vectorization process may be a process of processing the corresponding query content with vector coordinates in a multidimensional space identified based on the meaning of the question. At this time, natural language processing algorithms such as Universal Sentence Encoder and BERT are used. Through this, questions with similar meanings can have unique coordinates with a close distance.

Next, the negative/positive tagging process is a process of tagging after determining whether the user's question ends in a positive or negative form.

For example, if the question is 'Is it applicable to defamation?', it is a positive type, and if 'Is it not applicable to defamation?', it is a negative type. This kind of tagging is because 20% of positive questions and 20% of negative questions in numerical answers have different meanings. Therefore, if the numerical answer of a negative question is replaced with a positive one in the future, (100 - numerical answer) = the replaced numerical answer.

Next, the category tagging process is a process of tagging keywords related to the question content, and may be, for example, a process of tagging field keywords related to the question content or key keywords such as leasing, retirement, and dismissal.

When the role of the natural language processing unit 350 is completed, the query and similarity analysis/extraction unit 360 performs a process of comparing and analyzing the categories and multidimensional space vector coordinates similar to those of the query content with information in the database. do.

Here, the inquiry and similarity analysis/extraction unit 360 first extracts precedent information and legal consultation information having a category similar to that of the inquiry contents from the database, and then, after first extracting from the database, the query and similarity analysis/extraction unit 360 is similar to the multidimensional space vector coordinates of the inquiry contents (the coordinate distance is the most It may be a configuration of secondarily extracting information having multidimensional space vector coordinates from firstly extracted information.

In addition, the query and similarity analysis/extraction unit 360 performs the process of extracting legal cases related to the query contents through the above-described process.

Thereafter, the positive/negative quantification unit 370 calculates the average value of positive and negative categories (numerical values) for each of the at least one case precedent information and legal consultation information extracted from the query and similarity analysis/extraction unit 360. If the positive or negative numerical value of the average value of the positive and negative categories (numerical value) of the calculation process and at least one extracted case case precedent information and legal consultation information is calculated to be relatively higher than the predetermined value, the numerical value reflecting the weight provides

Finally, the result value providing unit 380 provides the query terminal 200 with a gauge-type graph of the positive and negative numerical values calculated by the positive/negative digitization unit 370 . Here, the result value providing unit 380 may use the numerical values of several cases as input values and provide them through a nonlinear model, similar to the method of using stacking in the final decision-making in the ensemble technique. .

Therefore, by using the system and method for providing a digitized AI answer service to legal questions according to an embodiment of the present invention, users can easily find out what they are curious about or legal knowledge or legal knowledge about events that have happened to them. It has the advantage of being able to receive an answer on whether or not to apply in the form of a gauge that includes positive and negative values, and to be provided with similar precedents and cases.

12 is a diagram illustrating an example computing environment in which one or more embodiments set forth herein may be implemented, an illustration of a system 1000 that includes a computing device 1100 configured to implement one or more embodiments described above. shows For example, computing device 1100 may be a personal computer, server computer, handheld or laptop device, mobile device (mobile phone, personal digital assistant, media player, etc.), multiprocessor system, consumer electronics, mini computer, mainframe computer, distributed computing environments that include any of the foregoing systems or devices; and the like.

Computing device 1100 may include at least one processing unit 1110 and memory 1120 . Here, the processing unit 1110 may include, for example, a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor, an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Arrays (FPGA), and the like. and may have a plurality of cores. The memory 1120 may be volatile memory (eg, RAM, etc.), non-volatile memory (eg, ROM, flash memory, etc.), or a combination thereof. Additionally, computing device 1100 may include additional storage 1130 . Storage 1130 includes, but is not limited to, magnetic storage, optical storage, and the like. The storage 1130 may store computer readable instructions for implementing one or more embodiments disclosed herein, and may also store other computer readable instructions for implementing an operating system, application programs, and the like. Computer readable instructions stored in storage 1130 may be loaded into memory 1120 for execution by processing unit 1110 . Computing device 1100 can also include input device(s) 1140 and output device(s) 1150 .

Here, input device(s) 1140 may include, for example, a keyboard, mouse, pen, voice input device, touch input device, infrared camera, video input device, or any other input device. Output device(s) 1150 may also include, for example, one or more displays, speakers, printers, or any other output devices, or the like. Additionally, computing device 1100 may use an input device or output device included in another computing device as input device(s) 1140 or output device(s) 1150 .

Computing device 1100 may also include communication connection(s) 1160 that allow computing device 1100 to communicate with other devices (eg, computing device 1300).

Here, communication connection(s) 1160 may be a modem, network interface card (NIC), integrated network interface, radio frequency transmitter/receiver, infrared port, USB connection, or other device for connecting computing device 1100 to other computing devices. May contain interfaces. Further, communication connection(s) 1160 may include a wired connection or a wireless connection. Each component of the aforementioned computing device 1100 may be connected by various interconnections such as a bus (eg, peripheral component interconnection (PCI), USB, firmware (IEEE 1394), optical bus structure, etc.) and may be interconnected by the network 1200. Terms such as "component" and "system" as used herein generally refer to a computer-related entity that is hardware, a combination of hardware and software, software, or software in execution.

For example, a component may be, but is not limited to, a process running on a processor, a processor, an object, an executable, a thread of execution, a program, and/or a computer. For example, both the application running on the controller and the controller may be components. One or more components can reside within a process and/or thread of execution and a component can be localized on one computer or distributed between two or more computers.

In the above, the present invention has been described in detail with reference to the examples, but those skilled in the art to which the present invention belongs will be able to make various substitutions, additions, and modifications without departing from the technical idea described above. Of course, it should be understood that such modified embodiments also belong to the protection scope of the present invention, which is defined by the appended claims below.

100: A system that provides digitized AI answering services to legal questions
200: query terminal
300: AI Legal/Q&A server
310: collection unit
320: pre-processing unit
330: database conversion unit
340: machine learning unit
350: natural language processing unit
360: query and similarity analysis/extraction unit
370: positive/negative quantification unit
380: result value provider

Claims (1)

Inquiry terminal requesting legal advice on the contents of inquiry: and
Case precedent information and legal consultation information of lawyers to which cases similar to the above query were applied are collected from at least one or more external servers, processed into a preset form, and built into a database, and the constructed case precedent information and legal consultation information Based on this, it includes an AI legal/question response server that analyzes legal answers to positive and negative questions about the contents of the inquiry using an artificial intelligence machine learning algorithm and provides them as numerical values of positive and negative values,
The query terminal provides a digitized AI answer service for legal questions, characterized in that for executing a question answering platform that displays the numerical value to the user in the form of a gauge.

Images