KR20190015797A - The System and the method of offering the Optimized answers to legal experts utilizing a Deep learning training module and a Prioritization framework module based on Artificial intelligence and providing an Online legal dictionary utilizing a character Strings Dictionary Module that converts legal information into significant vector - Google Patents

Abstract

The present invention provides knowledge of an answer or online legal information dictionary for making a legal document. More specifically, when a user inputs a search keyword, ontology of the corresponding search keyword and surrounding words are compared to check relationship so as to form a feature, and the feature is stored in a deep learning database and a use history database. When an optimized vector value is calculated through a deep learning training module by using artificial intelligence (AI), ranking of the corresponding vector value is determined through a ranking framework module. When a specific vector value continuously and repetitively solves a problem, an objective function for the specific vector value is formed. Accordingly, such specific vector values are connected through the objective function and thus an optimized answer is able to be suggested to the user.

Description

Deep learning training module through artificial intelligence (AI) and ranking framework module are used to provide legal experts with optimal solutions, while legal information is converted into semantic vectors, stored in a database, [0001] The present invention relates to a system and method for providing an online legal information dictionary using a dictionary module, and a method for providing the online legal information dictionary using a dictionary module dictionary usable a character Strings Dictionary Module that converts legal information into significant vector}

The present invention relates to a system for providing legal experts with knowledge of an online legal information dictionary or a model answer for preparing a legal document by utilizing techniques relating to artificial intelligence (AI) and online information dictionary, ≪ / RTI >

The present invention utilizes a deep learning training module through artificial intelligence (AI) and a ranking framework module to present a model answer optimized for a legal professional, transform legal information into a semantic vector, store it in a database, A system and a method for providing an online legal information dictionary using the dictionary module for the dictionary,

The underlying technology is a context-dependent search technique using Microsoft's Deep Learning model, which is disclosed in Publication No. 10-2016-0144384. Microsoft has a training method using deep learning model, which is a method of training artificial intelligence, as a basic technology, and has applied for it internationally. However, the above is a patent on the original technology, and it is possible to register a separate technology patent that commercializes it. In related areas, such as Google's Alphago and IBM's Watson, it has grown into an industry using it.

 As a purely technical content, Microsoft is explaining deep learning methods and Ranking Frame Work as a way to utilize artificial intelligence. The field does not explain the commercialization.

In Korea, there are no patents that provide a system model or method to utilize artificial intelligence in the field of law in Korea, and furthermore, artificial intelligence is used to create legal documents such as title, answer, , A system that provides a model answer for the user

It was not conceived.

In the case of legal information retrieval by artificial intelligence, based on the deep learning training technique by the ranking framework, the most suitable document is found, and priority is given to a large number of documents to provide optimal information , And the amount and quality of legal information that is provided gradually grow over time.

 In addition, as the number of users who use the document increases, the part of the user's choice is also involved in the formation of the attribute value and is reflected in the preference, so that the documents that receive the most selection are ranked in the top page in the search engine, You can share information with bad information.

On the other hand, technologies such as an online knowledge encyclopedia such as Wiki Pedia already exist, and in this form, industrial forms such as an online encyclopedia of legal information providing legal information online are also available. According to the related art, in the case of the encyclopedia of legal information using the online, the media can be grown through user participation. It is possible to participate in the editing process directly by the users, to help reflect the legal information in real time, and as an encyclopedia of on-line legal information as a collective intelligence, a legal information system that allows users to directly grow is also possible.

The advantage of the above technique is that artificial intelligence (AI) does not replace human expertise in the future but works only as a role to provide information for human users as a legal expert. For example, lawyers such as attorneys can not afford to fight information battles with large law firms and provide enough information so that individual lawyers can do their best to do their best for their mission or public interest, But only as a legal secretary who provides the best legal information and best practice.

SUMMARY OF THE INVENTION The present invention has been made in order to solve the above problems, and it is an object of the present invention to allow a search range to be recognized through an ontology that is a bundle of synonyms from a search word when a user inputs a search word. When the related search range is recognized, a specific feature value is formed by comparing the keywords existing in the legal information of the field with the keywords inputted by the user, and through the deep learning learning module by the AI, A deep artificial neural network is formed between the related keywords, and the most appropriate information among them is selected through the ranking framework work module to be recorded as a solution vector value. At this time, the user examines related related keywords, selects a table of contents which is most helpful for solving the problem, and the extracted exemplary legal phrases are provided as a single web document in the form of a model answer based on the string dictionary. It aims to provide the best level of legal information search ability and model answer through AI (Artificial Intelligence) to those who lack the ability to search for legal information.

  In addition, the present invention stores and manages legal information in a highly efficient manner so that a large amount of legal information can be easily found online. The present invention utilizes the dimension reduction module and the semantic vector conversion module to extract valuable legal statements from legal information and store them in a semantic vector manner with a small capacity. At this time, if the corresponding vector value is stored together with the string through the qualification value forming module, the corresponding vector value stored in the string can be called at any time in the future through the approach using the table of contents of the online dictionary. Since the vector values include legal information, it is aimed at constructing or providing an online legal information dictionary system with vast amount of information.

In order to achieve the object of the present invention described above, it is necessary to understand the programming in practice, understand the practical legal document writing method as a legal person, and operate the artificial intelligence technology, Deep Learning It is necessary to understand everything.

 In order to allow the artificial intelligence to process the legal information document by itself in the deep learning training technique utilizing the artificial intelligence (AI), the association of the keyword is first determined from legal information inputted from the input interface Extracting legal phrases automatically by recognizing the range, converting the extracted statistical phrases into a semantic vector value (Vector) that can be processed by the artificial intelligence by the dimension reduction module and the semantic vector conversion module, A step of assigning a quality value through a value forming module and storing the attribute value in a usage record database; comparing the input values with the result values through a deep learning neural network training model; And setting a priority value of each of the plurality of vector values, Setting an objective function vector which is a solution semantic vector, and thereafter arranging the objective function vectors as a table of contents linked to a string dictionary according to relevance, and when the user selects a table of contents, A step of extracting valuable legal phrases from a large amount of legal information inputted in order to construct or provide an online legal information dictionary system; , Storing the capacity in a small semantic vector manner, and storing the vector value together with the character string through the qualification value forming module.

The step of the artificial intelligence itself recognizing the range based on the relevance of the keyword and automatically extracting the legal statements includes a step of performing a synonym search according to the relation between the word and the word according to the ontology of the keyword, It may include a step of recognizing and designating the legal field, the civil law, the criminal law, the constitution, etc. as the limit of the search range in consideration of the relationship between the keyword and the thesaurus. For example, if the word creditor cancellation right is searched by a keyword, it can include the step of comparing the word piracy with the thesaurus and designating only the civil or civil litigation law as the limit of the search range.

The step of converting the extracted legal information into a semantic vector value (Vector) that can be processed by the artificial intelligence by the dimension reduction module and the semantic vector conversion module and storing the extracted legal information is a minimum character information, And storing it as a semantic vector value with a minimum capacity to impose a single symbol or number.

The step of assigning a quality value through the quality value formation module and storing the qualification value in the usage history database may include converting the extracted legal phrase to a semantic vector value of a minimum capacity through a semantic vector conversion module , And stored in the usage history database.

The step of comparing the input information with the results of the de-learning artificial neural network training model, comparing the priorities by the ranking framework, and setting the priorities of the artificial intelligence (AI) Searching for an identical word according to the relation between the word and the word according to an ontology, and concatenating the related individual vectors one by one.

The step of setting the objective function vector, which is a solution semantic vector, through the feature value calculation module from the parallel vector value input having the same priority, the connection operation is stored in the usage record database, , And the step of universalizing the related tasks, and the step of deriving it as a logic function as an objective function for the result when the neural network is continuously connected and the iterative resolution mechanism that always leads to the same conclusion is reached .

The step of rearranging the objective function vectors as a table of contents linked to a string dictionary according to relevancy comprises the steps of retrieving a string from the string DB and storing a new string of the string dictionary concatenated as the table of contents back into the string DB Step < / RTI >

Wherein when the user selects the table of contents, the step of arranging the legal statements extracted from the string dictionary to be provided as a web document in a form of one-by-one form is summarized in the up-and-down scroll method, For example, a deep learning database, a usage history database, and a character string database.

In order to construct or provide the on-line legal information dictionary system, a valuable legal phrase is extracted from a large amount of legal information to be inputted and stored in a semantic vector system having a small capacity. In this case, With a string. ≪ RTI ID = 0.0 >

The two steps may include the step of the user accessing the table of contents and retrieving the vector values stored with the string from the string database.

As mentioned above, if you use the Deep Learning Training Module (AI) and the Ranking Framework Module through Artificial Intelligence (AI) to present a model answer optimized for legal experts, There is also an advantage in that, even in the case of a lawyer with a career, he can gain a better understanding of the new field that he did not undertake and expand his / her field of specialization. And while the time and cost of legal information retrieval has been drastically reduced, the results of legal information search will be of the highest quality.

 Because the legal professional selects the most appropriate placeholder according to the solution of his / her desired solution task among the provided placeholders according to the keyword inputted by him, and according to the dictionary of strings arranged in the above table, Semantic vectors are provided as a single document. Such legal semantics as the solution semantic vectors themselves have the value of legal stability. In the process of examining the table of contents, we can find a new legal logic structure and find various approaches and legal basis for the case in the model proposed by AI.

  By using the various examples provided by this artificial intelligence to learn, you can experience the amazing experience that instantly grows your expertise in the field you want, at a minimum cost you could not have imagined.

On the other hand, if the legal information is converted into a semantic vector and stored in a database, and an online legal information dictionary using the string dictionary module is provided, only legal legal statements are extracted from the inputted legal information data, Because it can be stored as a vector, it has the advantage of handling a large amount of legal information data at once, while the vector value and the string are stored together, so you can see the latest legal information you need at any time, .

1 is a detailed block diagram showing the overall system of the present invention and the configuration of each apparatus.
FIG. 2 is a flowchart illustrating extraction of a legal phrase and semantic vectorization process for each input device according to an embodiment of the present invention.
FIG. 3 is a diagram illustrating a preferred embodiment of each string pre-module and artificial intelligence deep learning module according to an embodiment of the present invention.
FIG. 4 is a diagram illustrating an embodiment of a ranking framework module according to an embodiment of the present invention.
5 is a block diagram illustrating an embodiment of a quality value-related module according to an embodiment of the present invention.
to be.
FIG. 6 is a diagram illustrating an example of connection between functions of each interface and devices according to an embodiment of the present invention.
FIG. 7 is an exemplary diagram illustrating a function and an example of a search interface according to an embodiment of the present invention. FIG. 8 is a diagram illustrating a function and an implementation example of a detailed interface according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily carry out the present invention.

The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In order to clearly illustrate the present invention, parts not related to the description are omitted, and similar parts are denoted by like reference characters throughout the specification.

 Throughout the specification, when an element is referred to as "comprising ", it means that it can include other elements as well, without excluding other elements unless specifically stated otherwise.

Also, the terms of " part ", "... module" in the description mean units for processing at least one function or operation, which may be implemented by hardware or software or a combination of hardware and software.

And, the term " logic " described in the specification includes any physical and tangible function to perform a task. For example, each operation shown in the flowchart corresponds to a component of the logic for performing the operation. Which may be configured to perform operations using hardware, software, or the like, or any combination thereof.

 The following description may describe one or more features as being "optional", although other features may be considered optional, even if not explicitly mentioned. This is because any description of a single entity does not deny the use of a plurality of entities and the description of a plurality of entities does not preclude the use of a single entity. For example, the terms " exemplary ", " for example " refer to one implementation among a plurality of possible implementations.

 Hereinafter, by using a deep learning training module based on AI and a ranking framework module according to an embodiment of the present invention, it is possible to provide a model answer optimized for a legal professional, A system and method for converting a semantic vector into a semantic vector, storing it in a database, and providing an online legal information dictionary using the string dictionary module will be described in detail.

 Herein, a system for providing a model answer optimized for a legal professional by utilizing a deep learning training module and a ranking framework module through artificial intelligence (AI) according to an embodiment of the present invention, The ontology of the search keyword, and the surrounding words to obtain a feature value, and stores the feature in a deep learning database and a usage record database. The AI is then used to derive the optimized vector values through the Deep Learning training module, which is then used to rank the vector values through a ranking frame work module. do. At this time, when the specific vector value repeatedly solves the problem, the objective function is formed and the specific vector values are connected to present the optimal solution to the user.

 On the other hand, a system for converting legal information into a semantic vector, storing it in a database, and providing an on-line legal information dictionary utilizing the string dictionary module is a system in which a user or an external server transmits legal information The dimension is reduced and converted into a semantic vector and stored in a character string database. After extracting specific strings through the string extracting module, the mediating server compares these strings, forms a common point through the attribute value forming module, and stores the vector address including the attribute value in the use record database. Such qualities include, for example, the case where the same act or case is included at the same time. In the future, the string dictionary module provides the vector address and the vector value including the attribute value to the document rank based output interface, and the document rank based output interface arranges them and provides the legal information dictionary in the form of an online web document to the user .

 First, FIG. 1 is a detailed block diagram showing the overall system of the present invention and the configuration of each apparatus.

Referring to FIG. 1, the overall system and each apparatus of the present invention includes an input unit 100, an application unit 200, and a calculation unit 300. Here, the object range recognition apparatus can be implemented as a computing-based apparatus. Such a computing-based device may be a personal computer, a mobile device, or a server device, but is not limited thereto.

The input unit 100 searches for keywords (110) and receives arbitrary legal information documents (120). The arbitrary document may be a paper document or a document that can be searched or read on-line, for example, a legal information original document file, a data packet for a legal information document fetched from an external server, .

The function of the input unit is to receive all the relevant information in a predetermined format so as to import new legal information or to store it in the related database in order to receive the user's query command.

The application unit 200 is composed of a wide variety of modules and is mainly composed of a Deep Learning training module 230 for forming a multi-layer artificial neural network, a ranking frame operation module 235 associated therewith, A string extracting module 280 for sorting the sentences and extracting related strings, and a string dictionary module 285 for categorizing the extracted strings and storing them in a dictionary form. A dimension reduction module 260 and a semantic vector conversion module 270 that reduce the size of the provided legal information and convert it into a semantic vector as a minimum letter symbol and store it. And a document rank based output interface unit 290 that finally arranges the ranked documents in order and provides them to the user. The deep learning database unit 240, the usage history database unit 241, and the character string database unit 242 function in conjunction with each other depending on their relevance. As a module linked to the keyword search unit 110, when a user inputs a keyword, a keyword ontology (Ontology), a keyword range recognizing the scope of search and defining a domain A qualification value formation module 220 for forming a common feature value from the synonym information that is compared with the recognition module 210 and a feature value formation module 220 for forming a deep artificial neural network or a feature value or a purpose And a feature value calculation module 250 for calculating a function.

The calculation unit 300 has a purpose of providing a legal document calculated for the user, and includes a model answer learning unit 310 for providing a vector value optimized by the artificial intelligence AI as a web document according to a table of contents based on a character string ) And an enormous string of strings An online law that includes a string dictionary function that extracts strings from a database, constructs them into categories according to their associations, and provides a string as a dictionary type, And an information document providing unit 320.

The logic and apparatuses included in the input 100, application 200, and calculator 300 of the system are not so limited but are provided by way of example, and other logic and devices may be considered as optional . It should be understood that this is not an exemplary implementation of a plurality of possible implementations, in accordance with the principle that any description of the entity does not deny the use of a plurality of entities and that the description of a plurality of entities does not preclude the use of a single entity It can be seen as including an example.

FIG. 2 is a flowchart illustrating extraction of a legal phrase and semantic vectorization process for each input device according to an embodiment of the present invention. This represents a flow in which a particular operation is described as constituting a separate block performed in a particular order. This implementation is illustrative and not limiting. Thus, the specific blocks described herein may be grouped together and may be performed in a single operation, wherein a particular block may be divided into a plurality of constituent blocks, and a particular block may be performed in a different order than that shown herein It is possible. (E. G., Performing blocks in parallel). The blocks shown in this flowchart may be implemented in any manner and by any physical and tangible means, such as by computer software, hardware, or any combination thereof have.

Accordingly, referring to the embodiment of FIG. 2, A. the operation principle of the keyword search unit 110 is presented in the form of a series of processes performed according to a process of searching for a keyword. When the user inputs a keyword, determining the word relevance from the inputted search word and limiting the range of the keyword search within the meaning range by understanding the range of meaning with the ontology logic considering the legal terms;

Comparing the associated thesaurus through the ontology logic, generating a formed feature, and storing the associated vector value in the usage history database;

A step of storing a deep learning database by forming a connection between keywords using an ontology logic and a feature value calculation logic in a use record or a character string database;

Deep learnning trainning learning logic is performed by repeating the task of artificial neural network formation while performing the artificial intelligence through the artificial intelligence and concurrently forming the deeper artificial neural network connection structure step;

 Ranking the respective vector values through the ranking framework logic and using the document rank based output interface according to the determined rank to provide a user with an optimal model document.

2, when a legal document information provided by a user or an external server is received and a legal document input unit 120 performs an operation, a user or an external server If the original legal document is input to the legal document input unit 120, extracting the legal legal information from the original legal document, reducing the dimension, and converting the legal legal information into a semantic vector as a minimum character symbol;

 Storing the semantic vector in the usage record database, including the feature, based on the semantic vector;

 Forming a character string through the feature value calculation module in the usage record database and storing the formed character string in a character string database;

 Extracting a character string from the character string database through a character string extracting module, constructing a character string dictionary module by associating the related vector value with a character string, classifying the character string, and providing the classified character string in a dictionary form;

 And providing the online legal information dictionary to the user via the document rank based output interface, the online legal document dictionary providing the online web document.

FIG. 3 is a diagram illustrating a preferred embodiment of each string pre-module and artificial intelligence deep learning module according to an embodiment of the present invention.

Here, a method of selecting a legal document having a high relevance measure based on a string dictionary module is to extract a character string from a semantic vector value whose dimension of the legal statement is reduced. If a query is projected by a user or a query of artificial intelligence is projected much, Select and proceed based on the relevant vector values. When legal documents having a high degree of relevance to the keywords inputted by each user are presented in category form, the user can search for the category as a dictionary and find out the desired legal document.

The document refers to an item that contains any content for which the query is to be compared. In one case, a document is generated by a tool that generates an arbitrary document, corresponds to an arbitrary subject, and corresponds to a content item containing individual text represented in any format. For example, the document may correspond to a text document generated by a word processing program, an electronic mail message generated by an electronic mail program, an image with a text tag or annotations, a web page or other Internet accessible content item have. In other cases, the document may correspond to records of any type of data structure or any unstructured category. A document may correspond to an item in a table, and may correspond to a node such as an associated graph or the like. On the other hand, it can be connected to content that is created or consumed by the user. It is a feature of online legal information dictionary in order to provide related openness so that users can participate.

On the other hand, based on the AI learning module for deep AI learning, a method for generating a model answer through continuous deep learning converts the document information into a contextual context vector of a specific semantic space using a deep learning model. The context concept vectors continue to form a multi-layer artificial neural network among the vectors in the module through the structured semantic module work performed by the artificial intelligence. Through the construction of the artificial neural network, it is possible to select the frequently preferred vector values and the non-preferred vector values. Since the frequently preferred vector values are superior to other vector values or optimized, It is necessary to continue the operation of outputting a vector optimized for the situation so that the problem can be solved in various concept vectors.

FIG. 4 is a diagram illustrating an embodiment of a ranking framework module according to an embodiment of the present invention.

Here, the first artificial intelligence deep learning model 231 receives query information associated with a query input by a user. The first artificial intelligence deep learning model 231 can project the query information as the query concept vector value in the semantic space using the internal instance logic. The second artificial intelligence deep learning model 232 receives context information associated with the context in which the user submitted the query. The second artificial intelligence deep learning model 232 may use the inner instance to transform the context information into a context query vector. The third artificial intelligence deep-learning model 233 receives document information describing a candidate document from among the set of documents stored in the data repository. The third AI deep-learning model 233 may use its internal instance to transform its document information into its document concept vector.

 The comparison module 236 may form a plurality of relevance metrics for each document based on each semantic concept vector described above. For example, the first AI run model 231 may generate an relevance measure that reflects the conceptual relevance of the query concept vector to the document concept vector. Second artificial intelligence deep learning model 232 may generate an relevance metric that reflects the conceptual relevance of the document concept vector to the contextual concept vector. Although not shown in FIG. 4 above, alternatively or additionally, the comparison module may further form an association metric based on all three concept vectors and any joint analysis. Ranking module 237 may rank a plurality of candidate documents based on a plurality of features, e.g., relevance metrics, for each document supplied to it by comparison module 236. [ Ranking module 237 may additionally or alternatively add a new concept vector to progress Ranking according to the progress of AI training.

  As the artificial intelligence continues the individual ranking work for each ranking module steadily, it is possible to finally derive the vector value 239 as the optimized result.

FIG. 5 is a diagram illustrating an embodiment of a quality value-related module according to an embodiment of the present invention.

There are various methods for generating the feature value. For example, a common feature among the synonyms generated in the learning process of the deep learning module 230 may be extracted to generate specific qualities. For example, if there is a common law, or if there is a common law case, that law or case can be recognized as a feature. It extracts common features of the legal statements in which each vector value is embedded, and is also used as a qualifier to form a situation concept vector as a measure of relevance. When the deep learning learning module 230 continuously performs a task of changing a semantic vector into a situation vector, the qualification value formation module compares the parallel resources and forms a common feature as a qualification, And stored in the database 241.

If the feature value calculation module 250 assigns a weight to a specific feature value (251), one of the specific vector values A and B having the same ranking is selected (252). When a high weighting value is applied to a specific feature value, when the related vectors are used to obtain a high problem solving result, it is recognized that the iterative solving mechanism 253 is found, and the objective function 254 ). In this objective function 254, the objective functions having the highest relevance to the results are collected, and the resultant objective function logic 259 is formed again to classify the vector values optimized for the problem solution into categories classified by specific qualities . This is the source technology that forms the model answer.

FIG. 6 is a diagram illustrating an example of connection between functions of each interface and devices according to an embodiment of the present invention.

Referring to FIG. 6, the apparatuses constituting each interface include a three-level system device including a user input interface 100, an artificial intelligent AI deep learning training module 230, a document rank based output interface 290, The characterization module 220, the characterization module 250, the character string dictionary module 285, the ranking framework module 235, the document-based output interface 290, the legal information dictionary section 310, and a model receiving unit 320, and a computing device related to the method.

The computing devices corresponding to the interface may include a computing device such as a personal computer, a portable terminal, a server, and the like.

FIG. 7 is a diagram illustrating functions and an example of a search interface according to an embodiment of the present invention.

 The search interface is configured as a multifaceted frame so that all information can be viewed on one screen. This is based on a web document which can be scrolled up and down, and up and down scroll buttons are determined for each page frame of the bottom sorting method, and visibility is good and scrolling is possible.

The structure of the frame will be described.

 On the left is a category frame that provides categories of judgments, statutes, forms, and other materials (literature, administration, corporate law, etc.) and related legal documents;

 In the center, a frame for each of the frames for the purpose of providing each legal document is presented in an upper and lower frame form;

On the right is a document formed by artificial intelligence, AI frame for the purpose of providing an online dictionary and a template; And

  In order that the user can easily find the desired information by providing legal documents ranked in order of the document information provided in each frame in order of relevance, .

8 is a diagram illustrating an example of a function and an example of a detailed interface according to an embodiment of the present invention. The detailed interface is configured as a multi-frame frame so that all information can be viewed on one screen. This is based on a web document which can be scrolled up and down, and up and down scroll buttons are determined for each page frame of the bottom sorting method, and visibility is good and scrolling is possible.

In the structure of the frame, a search frame capable of keyword search at the upper part, and a type of an interface part provided by the calculation unit are based on a web document which can be scrolled up and down and composed of a multi-frame frame so that all information can be viewed on one screen A search frame in which a vertical scroll button is determined for each page frame of the bottom alignment method, visibility is good, scroll is possible, and keyword search is possible at the top;

 A table of contents frame, on the left side, which provides a table of contents composed of strings;

 In the center, a frame for each of the frames for the purpose of providing each legal document is presented in an upper and lower frame form;

On the right is a document formed by artificial intelligence, AI frame for the purpose of providing an online dictionary and a template; And

  In order that the legal documents in which the document information provided in each frame are arranged in the order of relevance so that the user can easily find the desired information, the entire detailed information And an information output interface screen.

100: input unit, 200: application unit, 300: calculation unit
110: a keyword search unit, 120: a legal information provider
210: keyword range recognition module, 220: feature value formation module
230: Artificial Intelligence AI Deep Learning Module,
231: Artificial Intelligence Deep Learning Learning Model 1, 232: Artificial Intelligence Deep Learning Learning Model 2,
233: Artificial Intelligence Deep Learning Learning Model 3, 235: Ranking Framework Work Module
236: comparison module, 237: ranking module
239: Ranked Result Derived Logic
240: Deep learning database unit, 241: Use log database unit
242: string database part,
250: qualification value calculation module, 251: weight comparison logic for various qualification values
252: weighting logic for the selected feature value,
253: Repetition resolution mechanism logic, 254: Object function generation logic
259: Resulting objective function storage logic, 260: Dimension reduction extraction module
270: Semantic vector conversion module, 280: String extraction module
285: string dictionary module, 290: document position based output interface part
310: Anonymous, 320: Legal Information Dictionary Department

Claims (8)

The system structure diagram is a system for providing legal information for a legal professional, the system comprising: an input unit allowing a legal expert to search for desired legal information through a keyword or input a desired legal information document;

A keyword range recognition module for recognizing an associated search range from an ontology that is a bundle of synonyms from the search keyword and restricting the search from other ranges;

A feature value formation module for enabling a specific feature to be formed through comparison between similar keywords existing in a search range of the field and keywords entered by the user;

For an input keyword with the qualification value set. Artificial intelligence (AI) deep learning module for artificial intelligence (AI) to input similar keywords at the same time and derive self-relevant vector values through each deep learning training technique;

A ranking framework task module for comparing optimized vector values calculated by said artificial intelligence and determining respective rankings;

A dimension reduction extraction module for selecting each legal statement in the corresponding legal document for reducing the dimension and extracting the provided legal document;

A semantic vector conversion module for converting the legal statement into a semantic vector having a small capacity in the form of a minimum letter symbol and storing the converted semantic vector;

A feature value calculation module for determining a new priority from a parallel vector value input having the same priority, through comparison of mutual feature values;

A string dictionary module for organizing the table of contents constructed by the string extracting module into a string dictionary;

A document rank based output interface unit which compares priorities of the respective legal document information and arranges and arranges high priority legal information to be displayed first;

A deep learning database unit for storing task information and optimized vector values and ranking performed by the artificial intelligence;

A usage record database unit for storing a record of a task performed by each of the modules, for example, a change processing record such as a semantic vector including a feature value and a feature value;

A string database unit for storing the semantic vector value, the character string, and a number of pieces of legal information inputted as provided legal information;

(AI) deep learning learning module, a ranking framework work module, a deep learning database part, a dimension reduction extraction module, a semantic vector conversion module, a feature value calculation module, a character dictionary Module, a document rank based output interface unit, a deep learning database unit, a use log database unit, a character string database unit, and the like;

A calculation unit for providing the provision document provided by the document rank based output interface unit in the form of a model document or an online web document;

The method of claim 1, further comprising the steps of: providing a legal expert with knowledge of an online dictionary or an online legal information dictionary for preparing a legal document using the artificial intelligence (AI) Systems and methods to provide legal information In order to allow AI to process legal information documents on its own within the Deep Learning training module using AI, legal information entered from the input interface can be used to define a range And extracting legal statements automatically;
Transforming it into a significant vector which can be processed by the artificial intelligence by a dimension reduction module and a semantic vector conversion module, and storing the semantic vector conversion module;
Adding the attribute value to the attribute value creation module and storing the attribute value in the usage record database;
Comparing the input information with the results of the de-learning artificial neural network training model, comparing the priorities by a ranking framework, and setting priority levels;
Setting an objective function vector as a solution semantic vector through a feature value calculation module from a parallel vector value input having the same priority;
Arranging the objective function vectors as a table of contents linked to a string dictionary according to relevance;
In order to construct or provide an on-line legal information dictionary system, it is necessary to extract valuable legal statements from a large amount of legal information to be inputted and to store them in a small semantic vector system;
Storing the corresponding vector value with the corresponding character string through the feature value forming module;
When the user selects a table of contents, the legal statements extracted from the string dictionary may be arranged in a scroll-up manner so that the legal statements can be provided as a web document in a form of an exemplary answer. 3. The method of claim 2, further comprising: receiving search and associated search information from a user computing device, e.g., a personal computer or a user terminal, via an internet communication network;
Receiving the search word and additional detailed information to receive search information;
4. The method of claim 3, further comprising the steps of: converting, using a deep learning model, into a legal syntax concept vector in a specific semantic space through a feature value formation module;
Receiving thesaurus related information associated with an ontology of a keyword, converting the string information to a concept vector in the semantic space using the deep learning model, or retrieving a previously generated thesaurus concept vector;
Comparing the legal phrase concept vector with the concept word vector to generate a relevance measure indicating a degree of a semantic relation mutually defined with the legal statement;
Determining a ranking score for the document based at least on the relevance metric;
And providing search results to the user computing device over the computer network based on the ranking score,
The deep learning model is a multi-layer artificial neural network
How to search The method of claim 3, wherein
Converting the deep learning model operation into a feature value concept vector in a specific semantic space through a feature value formation module using a deep learning model;
Related information associated with a string and converting the string information into a string concept vector in the semantic space using the deep learning model or retrieving a previously generated string concept vector;
Generating a plurality of ranking scores for the document, and ranking the documents based on the ranking score
How to search A computer-readable storage medium having stored thereon a computer readable instruction that, when executed by one or more processing devices, provides a ranking framework implemented by a search engine,
Logic configured to receive search and associated search information from the user computing device over the Internet network;
Logic configured to identify detailed information associated with the search, including details of the search;
Logic configured to evaluate a plurality of documents based on a context concept vector including a plurality of contexts and a plurality of document concept vectors;
Wherein the context concept vector corresponds to projection of document information associated with a specific document in the semantic space using the deep learning model,
The deep learning model is a multi-layer artificial neural network
Computer readable storage medium The method according to claim 4, further comprising: generating string classification learning data using an on-line legal information dictionary;
Generating a string locale classification model through the deep learning training learning on the generated character string classification learning data;
Classifying a character string for a lemma in the online dictionary using the generated string table classification model; And
Adding a sorted string to a string dictionary;
How to build a string dictionary 2. The image processing apparatus according to claim 1, wherein the interface unit provided by the input unit is based on a web document capable of scrolling up and down and composed of a multi-sided frame so that all information can be viewed on one screen, A search frame in which a scroll button is determined, a visibility is good, a scroll is possible, and a keyword search is possible at the top;
On the left is a category frame that provides categories of judgments, statutes, forms, and other materials (literature, administration, corporate law, etc.) and related legal documents;
In the center, a frame for each of the frames for the purpose of providing each legal document is presented in an upper and lower frame form;
On the right is a document formed by artificial intelligence, AI frame for the purpose of providing an online dictionary and a template; And
The legal documents in which the document information provided in each frame is ranked are arranged in the order of relevance so that the user can easily find the desired information,
Search interface 2. The image processing apparatus according to claim 1, wherein the form of the interface unit provided by the calculation unit is based on a web document capable of scrolling up and down and composed of a multifaceted frame so that all information can be viewed on one screen, A search frame in which a scroll button is determined, a visibility is good, a scroll is possible, and a keyword search is possible at the top;
A table of contents frame, on the left side, which provides a table of contents composed of strings;
In the center, a frame for each of the frames for the purpose of providing each legal document is presented in an upper and lower frame form;
On the right is a document formed by artificial intelligence, AI frame for the purpose of providing an online dictionary and a template; And
The legal documents in which the document information provided in each frame is ranked are arranged in the order of relevance so that the user can easily find the desired information,
Detailed information output interface

Images