Method for processing documents for compulsory execution

Abstract

According to one embodiment of the present disclosure, a method for processing enforcement documents performed by a computing device is disclosed. The method may include: receiving an input document including an enforcement title document; obtaining at least one of text data or image data from the input document; generating structured data including at least one of party information, claim amount information, or interest information included in the input document based on the obtained data; and normalizing the structured data based on a predefined schema of enforcement case data.

Description

Technology Field

[0001] The present invention relates to a technology for automating enforcement tasks, and more specifically, to a method for extracting party information, claim amount information, or interest information, etc., as structured data from an input document including an enforcement title document, and normalizing it based on a predefined schema of enforcement case data. Furthermore, the present invention may include a method for automatically generating an enforcement application document using the normalized enforcement case data and automatically inputting it into an electronic litigation portal. Background Technology

[0003] In compulsory execution operations aimed at realizing creditors' rights, a procedure is performed to draft various applications for attachment or execution based on case numbers, parties, claimed amounts, and interest information contained in execution title documents such as judgments, notarial deeds, payment orders, and decisions recommending performance, and to input the drafted content into the electronic litigation portal. However, traditionally, it was common practice for personnel to manually review execution title documents to draft applications and then re-enter the same content into the electronic litigation portal. This resulted in long processing times, susceptibility to input errors, and bottlenecks when handling multiple cases simultaneously. Furthermore, the step-by-step progress of ongoing cases often relied on separate manual management methods, which limited work standardization and mass processing capabilities.

[0004] Meanwhile, some prior art discloses technologies that automatically generate legal documents or support electronic submission using user input information or pre-structured claim files. However, these prior art technologies differ from the specific technical concept of automatically extracting information necessary for enforcement practice from unstructured or semi-structured legal documents into structured data using the enforcement title document itself as input, normalizing this into the format of enforcement case data, and reusing it for generating subsequent applications and inputting into the electronic litigation portal. In particular, a hybrid structure that processes enforcement title documents from the perspective of text and image data and combines AI model-based extraction processing with rule-based extraction processing holds technical significance that distinguishes it from prior art.

[0005] Furthermore, documents subject to compulsory execution vary in format and expression depending on the document type, and may include not only electronic documents with text layers but also documents requiring character recognition processing, such as scanned copies. In such an environment, it is difficult to reliably extract information using a single method, and a normalization process suitable for the compulsory execution domain is required to directly utilize the extracted information for drafting subsequent application documents and inputting it into the e-litigation portal. Therefore, there is a growing need for technology that can automatically structure information necessary for compulsory execution from input documents, including enforcement title documents, convert this into normalized compulsory execution case data, and consistently utilize it for generating application documents and inputting it into the e-litigation portal.

[0006] Korean registered patent No. 10-2119147 (May 29, 2020) discloses a method and system for the automatic generation of legal documents. The problem to be solved

[0008] The present disclosure aims to automate the processing of enforcement documents by automatically extracting information necessary for enforcement from an input document including an enforcement title document, generating it as structured data, and then normalizing it based on a predefined schema of enforcement case data.

[0009] As a specific example, the present invention may aim to reliably obtain information from different types of execution title documents by determining the type of input document, selectively performing direct text data extraction and image-based character recognition processing depending on whether a text layer exists or the amount of text, and generating structured data through a first extraction process based on an artificial intelligence model and a second extraction process based on rules.

[0010] In addition, the present invention aims to improve the consistency and efficiency of enforcement application tasks by automatically generating application documents corresponding to multiple enforcement types based on normalized enforcement case data, and automatically mapping and inputting the data included in the generated application documents into the input fields of the electronic litigation portal.

[0011] Meanwhile, the technical problem that the present disclosure aims to solve is not limited to the technical problem mentioned above, and various technical problems may be included within the scope obvious to a person skilled in the art from the contents described below. means of solving the problem

[0013] According to one embodiment of the present disclosure for realizing the aforementioned objectives, a method for processing enforcement documents is disclosed, which is performed by a computing device. The method may include: receiving an input document including an enforcement title document; obtaining at least one of text data or image data from the input document; generating structured data including at least one of party information, claim amount information, and interest information included in the input document based on the obtained data; and normalizing the structured data according to a predefined schema of enforcement case data.

[0014] In one embodiment, the execution title document may include at least one of a judgment, a notarial deed, a payment order, or a decision recommending performance.

[0015] In one embodiment, the input document further includes at least one of a resident registration abstract or a corporate registration abstract in addition to the execution title document, and the structured data may further include at least one of address information extracted from the resident registration abstract or corporate information extracted from the corporate registration abstract.

[0016] In one embodiment, the step of generating the structured data may include the step of applying at least one of an artificial intelligence model-based first extraction process or a rule-based second extraction process to the text data or image data.

[0017] In one embodiment, the step of acquiring at least one of the text data or image data may include: determining whether a text layer exists and the amount of text of the input document; if the text layer exists and the amount of text is greater than or equal to a predetermined threshold, directly extracting the text data from the text layer; and if the text layer does not exist or the amount of text is less than the predetermined threshold, converting the input document into a page-by-page image and then performing character recognition processing to extract the text data.

[0018] In one embodiment, the step of converting the input document into a page-by-page image may include a step of dynamically determining at least one of a resolution scale factor during image conversion, a page splitting mode for character recognition, or whether to apply binarization preprocessing based on the type of the input document.

[0019] In one embodiment, the step of generating the structured data may further include scanning a text area or an image area of ​​the input document to determine the type of the input document based on the matching result with a plurality of predefined keyword patterns, and applying an extraction schema corresponding to the determined type.

[0020] In one embodiment, the second extraction process is a process of selecting a dedicated extraction function corresponding to the type of the determined input document and applying a plurality of regular expression patterns included in the selected dedicated extraction function to the text data to extract second structured data, and the plurality of regular expression patterns may include a plurality of alternative patterns for each extraction target field and a multi-stage fallback structure that automatically switches to a second pattern when extraction by the first pattern fails.

[0021] In one embodiment, the plurality of regular expression patterns may include at least one of a hyphen normalization pattern that uniformly matches hyphen-like characters; a space flexible matching pattern corresponding to irregular spaces; a masking matching pattern that recognizes a masked identification number; a amount extraction pattern that normalizes an amount format; or a pattern that selects the latest address from an address history.

[0022] In one embodiment, the step of generating the structured data may further include the step of cross-comparing the first structured data generated by the first extraction process and the second structured data generated by the second extraction process by field.

[0023] In one embodiment, the step of cross-comparing the first structured data and the second structured data by field may include: a step of determining the corresponding field if the corresponding field value of the first structured data and the second structured data matches; and, if the corresponding field value does not match, a step of selecting one value or classifying the corresponding field as a user review target based on at least one of the success of extraction of each of the first extraction process and the second extraction process, a preset priority, or reliability.

[0024] In one embodiment, the step of normalizing the structured data based on a predefined schema of enforcement case data may include: a step of automatically distinguishing between natural persons and legal persons based on the number of digits of a personal identification number or a corporate registration number included in the party information; a step of converting an amount written in Hangul included in the claim amount information into numerical data; a step of structuring a start date, an end date, and an interest rate in an array form for each of a plurality of interest periods included in the interest information, while converting an expression of an undetermined end date into a predetermined reservation value; or at least one of the steps of individually extracting information for each of a plurality of debtors and structuring it into an array.

[0025] In one embodiment, the method may further include the step of displaying the structured data or the normalized structured data on a user interface and reflecting information modified or confirmed by the user.

[0026] In one embodiment, the method may further include the step (S500) of automatically generating an application document corresponding to at least one selected execution type among a plurality of execution types based on the normalized structured data.

[0027] In one embodiment, the step of automatically generating the application document may include: loading creditor information, debtor information, and claim amount information from the normalized structured data; specifying third-party debtor information corresponding to the selected execution type if the selected execution type involves a third-party debtor; automatically calculating an interest amount by applying period-specific interest rates to a plurality of interest periods based on interest information included in the normalized structured data, and automatically processing a transition from the interest rate at the time of filing the complaint to the interest rate after the judgment is finalized; and generating an application document file by mapping the loaded information, the third-party debtor information, and the calculated interest amount to a predetermined application form.

[0028] In one embodiment, the plurality of execution types include a type of seizure of claims including at least one of seizure of deposit claims, seizure of wage or severance pay claims, seizure of virtual assets, seizure of lease deposit claims, seizure of credit card accounts receivable, seizure of customer accounts receivable, seizure of automobiles, seizure of stocks, seizure of dividend claims, or seizure of insurance claims; and other execution types including at least one of forced auction of real estate, seizure of movable property, and registration on a list of defaulters, and when two or more of the plurality of execution types are selected simultaneously, application documents corresponding to each execution type may be generated in parallel.

[0029] In one embodiment, the method may further include the step of automatically mapping and inputting data included in the automatically generated application document into an input field of the electronic litigation portal.

[0030] In one embodiment, the step of automatically mapping and inputting into an input field of the electronic litigation portal may include: converting data included in the application document into a predetermined data format and storing it in a storage area of ​​a browser extension program; mapping each field of the stored data to a corresponding DOM (Document Object Model) element of the electronic litigation portal through a content script inserted into a web page of the electronic litigation portal by the browser extension program; displaying the mapped result as a preview to receive confirmation from a user; and inputting the mapped data into the DOM element in response to the user's confirmation.

[0031] In one embodiment, the method further includes the step of determining asset-related information of the debtor by cross-analyzing the input document and the credit report regarding the debtor, and the asset-related information may include at least one of a type of residence, employment status, vehicle registration status, business operation status, financial institution usage information, insurance subscription information, or accounts receivable-related information.

[0032] In one embodiment, the step of determining asset-related information of the debtor by cross-analyzing the input document and the credit report concerning the debtor may include the step of comparing the debtor identification information extracted from the input document with the identification information included in the credit report to link residence information, workplace information, or registered property information concerning the same debtor.

[0033] In one embodiment, the method may further include the step of automatically recommending a combination of executable execution types among a plurality of execution types based on the asset-related information.

[0034] In one embodiment, the combination of executable execution types may include at least two combinations of deposit receivables, wage receivables, severance pay receivables, lease deposit receivables, credit card accounts receivables, customer accounts receivables, automobiles, stocks, dividend receivables, or insurance receivables.

[0035] In one embodiment, the step of automatically recommending the combination of executable execution types may include: evaluating at least one of the feasibility, third-party debtor identification, or recovery feasibility of each execution type based on the asset-related information; and calculating the priority of the plurality of execution types based on the result of the evaluation.

[0036] A computing device according to one embodiment of the present disclosure is disclosed for realizing the aforementioned tasks. The device comprises at least one processor and memory, wherein the at least one processor may be configured to receive an input document including an enforcement title document, obtain at least one of text data or image data from the input document, generate structured data including at least one of party information, claim amount information, and interest information included in the input document based on the obtained data, and normalize the structured data based on a predefined schema of enforcement case data.

[0037] In one embodiment, the processor may be further configured to generate first structured data based on an artificial intelligence model based on image data of the input document; generate second structured data based on rules based on text data of the input document; and determine the structured data by comparing or integrating the first structured data and the second structured data by field.

[0038] In one embodiment, the processor may be further configured to automatically generate an application document corresponding to at least one selected execution type among a plurality of execution types based on the normalized structured data; and to perform automatic input by mapping the data included in the application document to an input field of the electronic litigation portal.

[0039] A computer program stored on a computer-readable storage medium is disclosed in accordance with one embodiment of the present disclosure for realizing the aforementioned objectives. When the computer program is executed on one or more processors, the one or more processors are configured to perform the following operations for processing an enforcement document, wherein the operations may include: receiving an input document containing an enforcement title document; obtaining at least one of text data or image data from the input document; generating structured data including at least one of party information, claim amount information, and interest information included in the input document based on the obtained data; and normalizing the structured data based on a predefined schema of enforcement case data. Effects of the invention

[0041] According to the present disclosure, at least one of text data or image data is obtained from an input document including an enforcement title document, structured data is generated based thereon, and then normalized according to a predefined schema of enforcement case data, thereby enabling the acquisition of data usable for subsequent processing from enforcement-related documents of different formats in a consistent format.

[0042] Furthermore, in one embodiment, direct extraction or character recognition processing based on the presence of a text layer and the amount of text, a first extraction processing based on an artificial intelligence model, a second extraction processing based on rules, or cross-comparison by field can be used, so that key information necessary for enforcement can be reliably extracted despite differences in the format of the title document, scan status, and variations in the written form.

[0043] In addition, according to the present disclosure, by using normalized enforcement case data, application documents corresponding to multiple types of enforcement can be automatically generated, and furthermore, the designation of a third-party debtor and the automatic calculation of interest can be performed, thereby alleviating the repetitiveness and complexity of the process of drafting enforcement application documents and promoting standardization of document drafting.

[0044] Furthermore, according to the present disclosure, data included in automatically generated application documents can be automatically mapped and entered into input fields of the electronic litigation portal. This reduces duplicate entry between the drafting of the application and the input of the electronic litigation, and reduces the possibility of input inconsistencies through the reuse of the same data. As a result, it is expected that the processing time for enforcement application tasks will be shortened, input errors will be reduced, tasks will be standardized, and the efficiency of processing large volumes of cases will be improved.

[0045] Meanwhile, the effects of the present disclosure are not limited to those mentioned above, and various effects may be included within the scope obvious to a person skilled in the art from the contents described below. Brief explanation of the drawing

[0047] FIG. 1 is a block diagram of a computing device for providing a method for processing an execution document according to one embodiment of the present disclosure. FIG. 2 illustrates an exemplary structure of an artificial intelligence-based model according to one embodiment of the present disclosure. FIG. 3 is a diagram schematically illustrating the data flow of an input document, various data, and an application document file according to one embodiment of the present disclosure. FIG. 4 is a flowchart illustrating a method for processing compulsory execution documents according to one embodiment of the present disclosure. FIG. 5 is a flowchart illustrating a method for processing enforcement documents using a first extraction process or a second extraction process according to an embodiment of the present disclosure. FIG. 6 is a flowchart for more specifically explaining a method for processing an execution document according to one embodiment of the present disclosure. FIG. 7 is a flowchart illustrating the steps for automatically generating application documents according to one embodiment of the present disclosure. FIG. 8 is a flowchart illustrating the automatic input step of an electronic litigation portal according to one embodiment of the present disclosure. FIG. 9 is a brief and general schematic diagram of an exemplary computing environment in which embodiments of the present disclosure may be implemented. Specific details for implementing the invention

[0048] Various embodiments are now described with reference to the drawings. In this specification, various descriptions are provided to provide an understanding of the present disclosure. However, it is evident that these embodiments can be practiced without such specific descriptions.

[0049] As used herein, terms such as “component,” “module,” “system,” etc. refer to computer-related entities, hardware, firmware, software, combinations of software and hardware, or executions of software. For example, a component may be, but is not limited to, a procedure executed on a processor, a processor, an object, an execution thread, a program, and / or a computer. For example, both an application executed on a computing device and the computing device itself may be a component. One or more components may reside within a processor and / or an execution thread. A component may be localized within a single computer. A component may be distributed among two or more computers. Additionally, these components may be executed from various computer-readable media having various data structures stored therein. Components may communicate through local and / or remote processes, for example, according to signals having one or more data packets (e.g., data from a component interacting with another component in a local system or distributed system, and / or data transmitted through signals to other systems and networks such as the Internet).

[0050] Furthermore, the term "or" is intended to mean an implicit "or" rather than an exclusive "or." That is, unless otherwise specified or evident from the context, "X uses A or B" is intended to mean one of the natural implicit substitutions. In other words, if X uses A; if X uses B; or if X uses both A and B, "X uses A or B" may apply to any of these cases. Additionally, the term "and / or" as used herein should be understood to refer to and include all possible combinations of one or more of the enumerated related items.

[0051] Additionally, the terms “comprising” and / or “comprising” should be understood to mean that such features and / or components are present. However, the terms “comprising” and / or “comprising” should be understood not to exclude the presence or addition of one or more other features, components and / or groups thereof. Furthermore, unless otherwise specified or clearly evident from the context to indicate a singular form, the singular in this specification and claims should generally be interpreted to mean “one or more.”

[0052] And, the term "at least one of A or B" should be interpreted to mean "a case including only A," "a case including only B," or "a combination of A and B."

[0053] Those skilled in the art should recognize that the various exemplary logical blocks, configurations, modules, circuits, means, logics, and algorithmic steps described in connection with the embodiments disclosed herein may be implemented in electronic hardware, computer software, or a combination of both. To clearly exemplify the interchangeability of hardware and software, various exemplary components, blocks, configurations, means, logics, modules, circuits, and steps have been generally described above in terms of their functionality. Whether such functionality is implemented in hardware or software depends on the specific application and design constraints imposed on the overall system. Skilled technicians may implement the described functionality in various ways for each specific application. However, such decisions regarding implementation should not be construed as going beyond the scope of this disclosure.

[0055] FIG. 1 is a block diagram of a computing device for providing a method for processing an execution document according to one embodiment of the present disclosure.

[0056] The configuration of the computing device (100) illustrated in FIG. 1 is merely a simplified example. In one embodiment of the present disclosure, the computing device (100) may include other configurations for performing the computing environment of the computing device (100), and only some of the disclosed configurations may constitute the computing device (100).

[0057] The computing device (100) may include a processor (110), memory (130), and a network unit (150).

[0058] The processor (110) may be composed of one or more cores and may include processors for data analysis and deep learning, such as a central processing unit (CPU) of a computing device, a general purpose graphics processing unit (GPGPU), and a tensor processing unit (TPU). The processor (110) may read a computer program stored in memory (130) and perform data processing for machine learning according to one embodiment of the present disclosure. According to one embodiment of the present disclosure, the processor (110) may perform calculations for learning a neural network. The processor (110) may perform calculations for learning a neural network, such as processing input data for learning in deep learning (DL), extracting features from input data, calculating errors, and updating the weights of the neural network using backpropagation. At least one of the CPU, GPGPU, and TPU of the processor (110) may process the learning of a network function. For example, a CPU and a GPGPU can work together to process the learning of a network function and data classification using the network function. Additionally, in one embodiment of the present disclosure, processors of a plurality of computing devices can be used together to process the learning of a network function and data classification using the network function. Furthermore, a computer program executed on a computing device according to one embodiment of the present disclosure may be a CPU, GPGPU, or TPU executable program.

[0059] According to one embodiment of the present disclosure, the memory (130) can store any form of information generated or determined by the processor (110) and any form of information received by the network unit (150).

[0060] According to one embodiment of the present disclosure, the memory (130) may include at least one type of storage medium among a flash memory type, a hard disk type, a multimedia card micro type, a card type memory (e.g., SD or XD memory), RAM (Random Access Memory), SRAM (Static Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Read-Only Memory), magnetic memory, a magnetic disk, and an optical disk. The computing device (100) may operate in conjunction with web storage that performs the storage function of the memory (130) on the internet. The description of the memory described above is merely an example and the present disclosure is not limited thereto.

[0061] A network unit (150) according to one embodiment of the present disclosure can use various wired communication systems such as a public switched telephone network (PSTN), xDSL (x Digital Subscriber Line), RADSL (Rate Adaptive DSL), MDSL (Multi Rate DSL), VDSL (Very High Speed ​​DSL), UADSL (Universal Asymmetric DSL), HDSL (High Bit Rate DSL), and a local area network (LAN).

[0062] In addition, the network unit (150) presented in this specification may use various wireless communication systems such as CDMA (Code Division Multi Access), TDMA (Time Division Multi Access), FDMA (Frequency Division Multi Access), OFDMA (Orthogonal Frequency Division Multi Access), SC-FDMA (Single Carrier-FDMA), and other systems.

[0063] In the present disclosure, the network unit (150) can be configured regardless of the communication mode, such as wired and wireless, and can be configured as various communication networks such as a Local Area Network (LAN), a Personal Area Network (PAN), and a Wide Area Network (WAN). In addition, the network may be a known World Wide Web (WWW) and may utilize wireless transmission technology used for short-range communication, such as Infrared Data Association (IrDA) or Bluetooth.

[0064] The technologies described in this specification can be used not only in the networks mentioned above but also in other networks.

[0066] FIG. 2 illustrates an exemplary structure of an artificial intelligence-based model according to one embodiment of the present disclosure.

[0067] Throughout this specification, artificial intelligence model, artificial intelligence-based model, computational model, neural network, network function, and neural network may be used interchangeably.

[0068] A neural network can be composed of a set of interconnected computational units that may generally be referred to as nodes. These nodes may also be referred to as neurons. A neural network is composed of at least one node. The nodes (or neurons) constituting neural networks may be interconnected by one or more links.

[0069] In a neural network, one or more nodes connected via links can form relative input and output node relationships. The concepts of input and output nodes are relative; any node in an output node relationship with respect to one node may be in an input node relationship with respect to another node, and vice versa. As described above, the input node versus output node relationship can be generated based on links. One or more output nodes may be connected to a single input node via links, and vice versa.

[0070] In a relationship between an input node and an output node connected through a single link, the value of the output node's data can be determined based on the data input to the input node. Here, the link interconnecting the input node and the output node may have a weight. The weight can be variable and can be varied by the user or an algorithm to enable the neural network to perform the desired function. For example, if one or more input nodes are interconnected to a single output node by respective links, the output node's value can be determined based on the values ​​input to the input nodes connected to the output node and the weights set on the links corresponding to each input node.

[0071] As described above, a neural network consists of one or more nodes interconnected through one or more links, forming input-output node relationships within the network. The characteristics of a neural network can be determined by the number of nodes and links within the network, the relationships between the nodes and links, and the weight values ​​assigned to each link. For example, if two neural networks exist with the same number of nodes and links but different weight values ​​for the links, the two neural networks may be recognized as different from each other.

[0072] A neural network can be composed of a set of one or more nodes. A subset of nodes constituting a neural network can form a layer. Some of the nodes constituting a neural network can form a layer based on their distances from an initial input node. For example, a set of nodes with a distance of n from an initial input node can form n layers. The distance from the initial input node can be defined by the minimum number of links that must be traversed to reach that node from the initial input node. However, this definition of a layer is arbitrary for illustrative purposes, and the degree of a layer within a neural network can be defined in a way different from that described above. For example, a layer of nodes may be defined by its distance from a final output node.

[0073] In one embodiment of the present disclosure, a set of neurons or nodes may be defined by the expression a layer.

[0074] Initial input nodes may refer to one or more nodes within a neural network to which data is directly input without passing through links in their relationships with other nodes. Alternatively, in terms of link-based relationships between nodes within the neural network, they may refer to nodes that do not have other input nodes connected by links. Similarly, final output nodes may refer to one or more nodes within a neural network that do not have output nodes in their relationships with other nodes. Furthermore, hidden nodes may refer to nodes constituting the neural network that are neither initial input nodes nor final output nodes.

[0075] A neural network according to one embodiment of the present disclosure may have the number of nodes in the input layer equal to the number of nodes in the output layer, and may be a neural network in which the number of nodes decreases and then increases again as it progresses from the input layer to the hidden layer. Additionally, a neural network according to another embodiment of the present disclosure may have the number of nodes in the input layer less than the number of nodes in the output layer, and may be a neural network in which the number of nodes increases as it progresses from the input layer to the hidden layer. Additionally, a neural network according to yet another embodiment of the present disclosure may have the number of nodes in the input layer greater than the number of nodes in the output layer, and may be a neural network in which the number of nodes decreases as it progresses from the input layer to the hidden layer. A neural network according to yet another embodiment of the present disclosure may be a neural network in which the above-described neural networks are combined.

[0076] An artificial intelligence-based model according to one embodiment of the present disclosure may include a deep neural network (DNN). A deep neural network may refer to a neural network that includes a plurality of hidden layers in addition to an input layer and an output layer. By using a deep neural network, latent structures of data can be identified. That is, latent structures of photos, text, video, voice, protein sequence structures, gene sequence structures, peptide sequence structures, music (e.g., what objects are in the photo, what is the content and emotion of the text, what is the content and emotion of the voice, etc.), and / or binding affinity between peptides and MHCs can be identified. Deep neural networks may include convolutional neural networks (CNN), recurrent neural networks (RNN), autoencoders, restricted Boltzmann machines (RBM), deep belief networks (DBN), Q networks, U networks, Siamese networks, Generative Adversarial Networks (GAN), Transformers, etc. The description of deep neural networks described above is merely illustrative and the present disclosure is not limited thereto.

[0077] The artificial intelligence-based model of the present disclosure may be represented by a network structure of any structure described above, including an input layer, a hidden layer, and an output layer.

[0078] A neural network that can be used in an artificial intelligence-based model of the present disclosure may be trained in at least one of supervised learning, unsupervised learning, semi-supervised learning, transfer learning, active learning, or reinforcement learning. Training of a neural network may be a process of applying knowledge to the neural network to perform a specific operation.

[0079] Neural networks can be trained to minimize the error in their output. The training process involves repeatedly inputting training data into the network, calculating the error between the network's output and the target for the training data, and updating the weights of each node by backpropagating the error from the output layer to the input layer in a direction that reduces the error. In supervised learning, training data is used where the correct answer is labeled for each data point (i.e., labeled training data), whereas in unsupervised learning, the correct answer may not be labeled for each training data point. For instance, in the case of supervised learning for data classification, the training data may consist of data where each training point is labeled with a category. Labeled training data is input into the neural network, and the error can be calculated by comparing the network's output (category) with the labels of the training data. As another example, in the case of unsupervised learning for data classification, the error can be calculated by comparing the input training data with the neural network's output. The calculated error is backpropagated in the neural network (i.e., from the output layer to the input layer), and through backpropagation, the connection weights of each node in each layer of the neural network can be updated. The amount of change in the connection weights of each node being updated can be determined by the learning rate. The neural network's calculation of the input data and the backpropagation of the error can constitute a learning cycle (epoch). The learning rate can be applied differently depending on the number of iterations of the neural network's learning cycle. For example, a high learning rate can be used in the early stages of training to quickly achieve a certain level of performance and increase efficiency, while a low learning rate can be used in the later stages to improve accuracy.

[0080] In the training of neural networks, the training data is generally a subset of the real-world data (i.e., the data intended to be processed using the trained neural network). Consequently, a training cycle may exist where errors decrease on the training data but increase on the real-world data. Overfitting is a phenomenon where the network learns excessively on the training data, leading to increased errors on the real-world data. For example, a neural network trained on yellow cats might fail to recognize cats when seeing anything other than yellow, which can be considered a type of overfitting. Overfitting can act as a cause for increased errors in machine learning algorithms. Various optimization methods can be used to prevent this overfitting. To prevent overfitting, methods such as increasing the training data, regularization, dropout (which disables some nodes in the network during training), and the use of batch normalization layers can be applied.

[0082] The present disclosure relates to a method for processing enforcement documents performed by a computing device, a computer program for performing the same, and a computing device including the same. Specifically, the present disclosure provides a method for obtaining data usable for subsequent processing from various enforcement-related documents in a consistent format by performing structuring and normalization operations based on an input document including an enforcement title document.

[0083] A computing device according to one embodiment of the present invention can provide an enforcement document processing service by utilizing a Large Language Model (LLM).

[0084] A method for processing enforcement documents according to one embodiment of the present invention may be configured to include a service providing server (hereinafter, a computing device) and an artificial intelligence server that provides a large-scale language model (LLM).

[0086] In the present disclosure, the term “document of execution” may refer to a document serving as the basis for compulsory execution. For example, a document of execution may include, but is not limited to, judgments, notarial deeds, payment orders, decisions recommending performance, etc., and may include other documents serving as a prerequisite for compulsory execution.

[0087] Additionally, the “input document (200)” may refer to a set of documents of a higher concept that includes the aforementioned execution title document (201), and may additionally include auxiliary public documents or related documents such as a resident registration abstract, a corporate registration certificate, documents related to delivery, documents related to credit investigation, and address verification documents. Accordingly, the input document (200) may be a single document or a bundle of multiple documents, and may be provided as an electronic document, a scanned document, an image document, or a combination thereof.

[0088] Additionally, in the present disclosure, “text data (210)” may refer to all character information that is directly extracted from an input document (200) or obtained through character recognition processing. For example, text data (210) may include not only simple strings, but also paragraph order, page number, line order, coordinate information, block section information, or location information of specific phrases.

[0089] Additionally, “image data (220)” may refer to data that represents all or part of the page of an input document (200) in the form of an image, and may include scanned images, grayscale images, binarized images, or partial images with only specific areas cut out as a result of rendering the original PDF (Portable Document Format).

[0090] Additionally, “structured data (230)” may refer to intermediate data arranged in a machine-processable form of unstructured or semi-structured information recorded in an input document (200). Structured data (230) may be expressed, for example, in a Key-Value format, an Object format, a JSON (JavaScript Object Notation) format, a Table format, a Record format, or a Graph format. Structured data (230) may include at least one of party information (231), claim amount information (232), or interest information (233), and may additionally include, if necessary, a case number, court name, date of service, date of finalization, whether a provisional execution order was issued, address information, corporate information, information related to the execution order, document type information, extraction reliability information, etc.

[0091] Additionally, the “predefined enforcement case data (240)” may be understood as a standard data model or standard schema for aligning, integrating, and reusing structured data (230) in a format suitable for enforcement practice, rather than as the final result data for individual cases itself. For example, the predefined enforcement case data (240) may be a standard structure that includes at least some of the elements, such as a set of fields to represent enforcement cases, reference relationships between fields, data types, whether they are mandatory, a repeatable array structure, rules for representing dates and amounts, rules for processing identification numbers, or methods for representing interest periods.

[0092] Additionally, the “normalized enforcement case data (250)” is the result data in which the structured data (230) is converted to fit the above standard structure, and can be reused in subsequent steps for generating an application, processing branching by enforcement type, and inputting into the electronic litigation portal. In the present disclosure, the normalized enforcement case data (250) may be used with the same meaning as the normalized structured data (250).

[0093] Furthermore, in this specification, “normalization” does not mean merely the operation of matching data formats, but can be used to encompass a series of processes including integrating information extracted from multiple documents, replacing different expressions with the same expression system, rearranging incomplete expressions into a structure capable of subsequent processing, and mapping them into a field structure suitable for processing enforcement cases. For example, processes such as integrating “creditor” and “plaintiff” into the same field when they have the same meaning, converting amounts written in Korean into numeric amounts, converting expressions with an undetermined deadline such as “until full repayment” into reserved values, or separating information about multiple debtors into an array structure may all be included in normalization.

[0094] Furthermore, in this disclosure, “execution” or “type of execution” may be understood as a higher-level concept encompassing individual means of execution, forms of application, or methods of execution that may be selected or performed in compulsory execution proceedings. On the other hand, “seizure” may be understood as a lower-level concept included in the aforementioned execution or type of execution, as a type or stage of compulsory execution premised on restricting the disposition of a debtor’s property or property rights. For example, seizure of deposit claims, seizure of wage claims, seizure of severance pay claims, seizure of lease deposit claims, seizure of credit card accounts receivable, seizure of accounts receivable from business partners, seizure of automobiles, seizure of stocks, seizure of dividend claims, or seizure of insurance claims may be examples of types of execution corresponding to seizure. Conversely, procedures or forms of application that can be utilized in compulsory execution practice but are not limited solely to the same legal nature as seizure, such as compulsory auction of real estate, execution of movable property, or registration on the list of defaulters, may also be included in the aforementioned type of execution. Accordingly, in this disclosure, “type of execution” may be a broader concept than “type of seizure” and, unless specifically limited otherwise in the context, may be used to encompass all methods of execution related to compulsory execution, including but not limited to seizure.

[0096] Hereinafter, exemplary embodiments of the present disclosure will be described in more detail with reference to the attached drawings. However, the present disclosure is not limited to the embodiments described herein and may be implemented in various modified forms. The terms used herein are intended to describe specific embodiments and are not intended to limit the scope of the present disclosure. Furthermore, throughout the present specification, the same reference numerals are assigned to identical or similar components, and redundant descriptions thereof may be omitted.

[0098] FIG. 3 is a diagram schematically illustrating the data flow of an input document, various data, and an application document file according to one embodiment of the present disclosure.

[0099] Referring to FIG. 3, a method for processing enforcement documents according to one embodiment of the present disclosure may receive an input document (200), obtain text data (210) or image data (220) from the input document (200), generate structured data (230) using the obtained data, and convert it into normalized enforcement case data (250) based on predefined enforcement case data (240). Furthermore, it may be configured as an overall flow that generates an application document file (260) and links with an electronic litigation portal (320) through a browser extension program (310). Such a flow has technical significance in that it combines document understanding, data normalization, document generation, and external portal linkage into a single continuous processing system.

[0100] Specifically, the input document (200) may include the aforementioned execution title document (201). The execution title document (201) may include documents serving as the basis for compulsory execution, such as a judgment, a notarial deed, a payment order, or a decision recommending performance, and, if necessary, auxiliary documents such as a resident registration abstract or a corporate registration abstract may be received together. In this case, the input document (200) may be composed of a set of multiple files rather than a single file, and multiple documents regarding the same case may be grouped and processed as a single case unit. For example, when a judgment and a resident registration abstract are received together, the parties and the amount claimed may be extracted from the judgment, and the latest address may be extracted from the resident registration abstract and reflected in the structured data (230) of the same case.

[0101] Text data (210) and image data (220) obtained from an input document (200) can be used as direct inputs for subsequent structuring. For example, in the case of an electronic PDF containing a text layer, the text data (210) can serve as a relatively direct and precise extraction input, while in the case of a scanned PDF or image file, the image data (220) can serve as the primary processing target. However, the present disclosure does not limit this dichotomously. For example, it includes a method of simultaneously obtaining text data (210) and image data (220) for a single input document (200) and cross-referencing extraction results of different methods.

[0102] Structured data (230) is an intermediate result derived from the text data (210) or image data (220) and can represent the facts related to compulsory execution included in the input document (200) in a machine-readable form. Structured data (230) may include at least one of party information (231), claim amount information (232), or interest information (233). However, it is not limited thereto, and structured data (230) may further include case number, court name, date of establishment of execution title, date of finalization, date of service, address information, corporate registration information, number of interest periods, document type identifier, extraction method identifier, reliability of extraction results, or whether review is required. At this time, the structured data (230) may not yet be in a state fully suitable for subsequent business processing, for example, heterogeneous expressions having the same meaning may be mixed, different field configurations may exist depending on the document type, or information regarding the same case may be scattered across multiple documents.

[0103] Accordingly, structured data (230) can be converted into normalized execution case data (250) based on predefined execution case data (240). The predefined execution case data (240) can function as a standard structure for representing execution cases, and the normalized execution case data (250) can be result data rearranged according to that standard structure. For example, the normalized execution case data (250) can be organized by utilizing at least some of the creditor field, debtor field, claimed principal field, array of multiple interest periods, third-party debtor candidate field, address field, or document type tag field so that it can be directly referenced when automatically generating an application. If the normalized execution case data (250) is maintained separately in this way, the same data can be reused in multiple subsequent processes, such as drafting an execution application, managing cases, or inputting into an electronic litigation portal.

[0104] The normalized enforcement case data (250) can be used to generate an application document file (260). The application document file (260) may be, for example, a PDF document, a printable document, an electronic document for storage, or a document for linking with an external system. Additionally, the application document file (260) or the corresponding data may be transmitted to a browser extension (310), and the browser extension (310) may interact with a web page of the electronic litigation portal (320) to input each item of the data into the corresponding field. That is, FIG. 3 indicates that one embodiment of the present disclosure is not a simple document extraction technology, but provides a series of processing systems that start with document input and continue through normalization, application generation, and input into an external portal.

[0106] FIG. 4 is a flowchart illustrating a method for processing compulsory execution documents according to one embodiment of the present disclosure.

[0107] Referring to FIG. 4, a method for processing an execution document according to one embodiment of the present disclosure may include the steps of receiving an input document (200) including an execution title document (S100), obtaining at least one of text data (210) or image data (220) from the input document (S200), generating structured data (230) based on the obtained data (S300), and normalizing the structured data (230) based on a schema of predefined execution case data (240) (S400). This basic embodiment represents the overall technical flow of the present disclosure at the highest level, and various detailed embodiments described below may be understood as further specifying or subdividing the steps.

[0108] According to one embodiment of the present disclosure, a computing device (100) may receive an input document (200) including an enforcement title document (S100). The input document (200) may be received by uploading via a user interface, transmission from an external system, network-based reception, linkage with a scanning device, calling from an electronic document repository, or a combination thereof. In one embodiment, the input document (200) may be a single PDF file corresponding to a single event. In another embodiment, the input document (200) may be a bundle of multiple documents corresponding to a single event. For example, a judgment, a resident registration abstract, and a corporate registration certificate may be received together in connection with a single event. In yet another embodiment, the input document (200) may include an auxiliary document to improve the accuracy of creating an application document file (260) to be automatically generated in a subsequent step, while centering on the enforcement title document (201). By broadly defining the input document (200) in this way, the present disclosure can flexibly accommodate various combinations of documents actually used in enforcement practice.

[0109] According to one embodiment of the present disclosure, a computing device (100) may obtain at least one of text data (210) or image data (220) from an input document (200) (S200). This step may be understood as a preprocessing step for subsequent structured data generation. In one embodiment, the computing device (100) may obtain text data (210) by using a text layer embedded in the input document (200). In this case, the text data (210) may have the form of page-by-page text, line-by-line text, block-by-block text, or text fragments associated with specific coordinates. In another embodiment, the computing device (100) may obtain image data (220) by converting the input document (200) into a page-by-page image or a region-by-region image. In this case, the image data (220) may be used as an input for character recognition processing or as an input for interpreting the meaning of an unstructured document layout, seal, tabular format, or image-based document. In another embodiment, the computing device (100) may acquire both text data (210) and image data (220) for the same input document (200) and use them complementarily. For example, text data (210) acquired directly from a text layer may be used as the primary processing input, and additional verification using image data (220) may be performed for a specific page or specific area.

[0110] According to one embodiment of the present disclosure, a computing device (100) may generate structured data (230) including at least one of party information (231), claim amount information (232), or interest information (233) included in an input document (200) based on acquired data (S300). Here, the structured data (230) may not be data that is simply copied from the original text, but may be data having a certain field structure so that it can be utilized for subsequent business processing. For example, the party information (231) may be expressed as a creditor name, debtor name, whether there are multiple debtors, address, whether it is a corporation, personal identification number, or corporate registration number. In addition, the claim amount information (232) may include principal, late payment damages, costs, Korean notation of the amount, numerical notation of the amount, or information on the separation of multiple amount items. In addition, the interest information (233) may include an interest rate, a calculation date, an end date, the number of interest periods, a classification by period, or a classification of interest rates before and after the final judgment. In another embodiment, the structured data (230) may include additional data in addition to the above information, such as document type, case number, court name, notarization date, date of finalization, whether the payment order has been finalized, the latest address, or corporate information. The present disclosure is not limited by the number of items or the method of representation of the structured data (230), and various structures may be adopted as long as the semantic information required for the subsequent processing of the enforcement case is expressed in a machine-processable form.

[0111] According to one embodiment of the present disclosure, a computing device (100) can normalize structured data (230) based on a schema of predefined enforcement case data (240) (S400). In one embodiment, the predefined enforcement case data (240) may be a standard data model that defines field structures commonly required in an enforcement domain. For example, field structures such as creditor, debtor, claim amount, multiple interest periods, type of enforcement title, information related to the writ of execution, address, and corporate information may be predefined. The computing device (100) can rearrange the information contained in the structured data (230) according to the standard structure, integrate different document representations into the same field, or convert omitted or ambiguous representations into values ​​that can be processed in a subsequent step. In one embodiment, the normalized enforcement case data (250) may be directly referenced for subsequent application generation or input into an electronic litigation portal. In another embodiment, the normalized enforcement case data (250) can be stored in a database and utilized for case management, intermediate review, or change history management. In this way, the normalization step can improve the reusability and consistency of the same data by converting the extraction results into a common data format suitable for enforcement practice.

[0113] FIG. 5 is a flowchart illustrating a method for processing enforcement documents using a first extraction process or a second extraction process according to an embodiment of the present disclosure.

[0114] Referring to FIG. 5, in one embodiment of the present disclosure, at least one of an artificial intelligence model-based first extraction process (S310) and a rule-based second extraction process (S320) may be applied to generate the structured data (230). The embodiment corresponds to a technical concept of increasing the generation stability of the structured data (230) by applying extraction methods having different characteristics in parallel, sequentially, or selectively. That is, the first extraction process is suitable for understanding unstructured context and flexible semantic interpretation, whereas the second extraction process may be suitable for precise capture of specific phrases, patterns, number formats, date formats, or identification number formats. Accordingly, the present disclosure can secure practical reliability in the processing of execution title documents by combining multiple extraction methods instead of relying on a single extraction method.

[0115] According to one embodiment of the present disclosure, a computing device (100) may receive an input document (200) including an enforcement title document (S100). The step may be performed in the same or similar manner as the step of the basic embodiment described above, and this chapter focuses more on the generation of structured data. The input document (200) may include an enforcement title document (201), and, if necessary, multiple documents may be received in a state linked to a single event.

[0116] According to one embodiment of the present disclosure, a computing device (100) may obtain at least one of text data (210) or image data (220) from an input document (200) (S200). The text data (210) and image data (220) may be used independently or simultaneously. For example, if there is a sufficient text layer, extraction processing may be performed with a focus on the text data (210), and if there is insufficient text layer, such as in a scanned document, processing may be performed with a focus on the image data (220). In another embodiment, the computing device (100) may combine a plurality of data sources in such a way that a primary processing result is obtained from the text data (210) and a secondary verification result is obtained from the image data (220).

[0117] According to one embodiment of the present disclosure, a computing device (100) may perform a first extraction process based on an artificial intelligence model (S310). The first extraction process may use text data (210), image data (220), or a combination thereof as input. In one embodiment, the first extraction process may interpret the semantic structure of an input document (200) using a multimodal language model or a document understanding model and extract fields such as party information (231), claim amount information (232), and interest information (233). For example, the artificial intelligence model may recognize different terms such as “plaintiff,” “creditor,” and “applicant” as parties of the same category based on the context within the document, and may distinguish or process substantially similar amount items such as “principal,” “claim amount,” and “borrowed amount” in an integrated manner. In one embodiment, the artificial intelligence model may refer to different extraction schemas depending on the document type, prioritizing the reading of the order and interest rate in a judgment, prioritizing the reading of the loan amount and late payment damages rate in a notarial document, and focusing on extracting the date of confirmation or the content of the decision in a payment order or a decision recommending performance. Such a first extraction process may have advantages in contextual flexibility and understanding semantic relationships between fields.

[0118] According to one embodiment of the present disclosure, a computing device (100) may perform a rule-based second extraction process (S320). The second extraction process may be a process that derives structured information by regularly searching for specific expressions included in text data (210). In one embodiment, the second extraction process may select a dedicated extraction function corresponding to a document type and apply a plurality of regular expression patterns, keyword patterns, string search rules, number format rules, date format rules, or context window rules corresponding to each extraction target field. For example, the format of a case number, the format of a resident registration number or corporate registration number, the format of a number and Hangul combination of an amount, the format of an interest rate entry, expressions related to interest ranges such as “from,” “until,” and “until the day of full repayment,” and role indicator phrases such as “creditor,” “debtor,” “plaintiff,” and “defendant” may be captured as a regular expression pattern or a rule-based pattern. In addition, the second extraction process may not use only a single pattern for a single field, but may have a multi-stage fallback structure that switches to a second or third pattern if the first pattern fails. Such a configuration may be advantageous for responding to issues such as document format variations, OCR errors, space insertion, hyphen variations, and masking notation.

[0119] According to one embodiment of the present disclosure, a computing device (100) may generate structured data (230) including at least one of party information (231), billing amount information (232), or interest information (233) included in an input document (200), based on acquired data or the result of each extraction process (S300). In one embodiment, the structured data (230) may be generated through only a first extraction process. In another embodiment, the structured data (230) may be generated through only a second extraction process. In yet another embodiment, the structured data (230) may be generated by combining a first extraction result generated by a first extraction process and a second extraction result generated by a second extraction process. The combination may be a simple merging, or it may include field-unit priority selection, reliability comparison, verification of match, classification of user review targets, etc. For example, if an artificial intelligence model extracts a creditor name with high reliability and a rule-based extraction fails to extract the corresponding field, the first extraction result may be adopted. Conversely, in cases where formal rigor is important, such as with an amount field, a rule-based second extraction result may be adopted first. In another embodiment, if two extraction results present different values ​​for the same field, the computing device (100) may mark the field for user review and pass it on to a subsequent verification procedure. In this way, the structured data (230) is not fixed to a single extraction engine result but can be understood as intermediate data combining multiple results.

[0120] According to one embodiment of the present disclosure, a computing device (100) may normalize structured data (230) based on a schema of predefined enforcement case data (240) (S400). The normalization may be performed regardless of whether the structured data (230) was generated by a first extraction process, a second extraction process, or a combination thereof. In one embodiment, the computing device (100) may perform processing to classify party information included in the structured data (230) as a natural person or a legal entity, unify monetary expressions into numerical forms, organize multiple interest periods into an array structure, or integrate information derived from multiple debtors or multiple documents into a single case unit data. The processing may be performed partially, in combination, or in whole. In addition, the above normalization process can be performed considering the generation of subsequent applications, input into the electronic litigation portal, and management of case storage and modification history, and thereby the normalized enforcement case data (250) can function as a common data hub for the automation of enforcement tasks beyond simple extraction results.

[0121] In summary, the parallel extraction embodiment described in this disclosure provides a technical framework for flexibly combining multiple extraction paths based on text data (210) and image data (220), integrating each extraction result into structured data (230), and then converting it into normalized enforcement case data (250). Such a structure can enable consistent conversion into data reusable in enforcement practice despite variations in input document format, differences in document quality, and differences in recording practices.

[0123] FIG. 6 is a flowchart for more specifically explaining a method for processing an execution document according to one embodiment of the present disclosure.

[0124] FIG. 6 is a detailed flowchart illustrating the determination of the type of an input document (200), the acquisition of text data (210) or image data (220), a first extraction process based on an artificial intelligence model (S310), a second extraction process based on rules (S320), a cross-comparison of extraction results (S330), the generation of structured data (230) (S300), and normalization (S400) based on a schema of predefined enforcement case data (240) according to an embodiment of the present disclosure. The technical core of the present disclosure is not merely to read the input document (200) as text, but to generate normalized enforcement case data (250) at a level that can be directly used in enforcement practice by comprehensively considering the document type, the presence or absence of a text layer, image characteristics, extraction characteristics by field, and the data structure required in subsequent enforcement tasks. In this respect, the present disclosure can be distinguished from simple OCR (Optical Character Recognition) or simple document classification technology.

[0125] In the present disclosure, “document type” is information indicating what legal or administrative nature an input document (200) has, and may refer to classification results such as, for example, a judgment, a notarized document, a payment order, a decision recommending performance, a resident registration abstract, or a corporate registration certificate. However, the document type is not limited to such explicit types, and may be expanded and defined into more detailed sub-types within the same document type, such as a civil judgment, a document granting an execution order, a notarized loan agreement, an annexed document related to a seizure of claims, etc.

[0126] Additionally, in the present disclosure, “field” may refer to an information unit constituting structured data (230) or normalized enforcement case data (250), and may include individual items such as, for example, creditor name, debtor name, claim amount, principal, interest rate, calculation date, fixed date, case number, document number, address, corporate name, corporate registration number, etc.

[0127] According to one embodiment of the present disclosure, a computing device (100) can determine the type of an input document (200) (S110). The above step may function as a higher step that provides criteria for a subsequent data acquisition method, extraction schema, character recognition preprocessing conditions, a set of regular expression patterns, and normalization logic. In one embodiment, the computing device (100) scans a text area or an image area of ​​the input document (200) and determines the document type based on the matching result with a plurality of predefined keyword patterns. For example, if phrases such as “judgment,” “order,” “plaintiff,” “defendant,” or “late payment damages” are detected in a certain combination, it can be determined as a judgment document, and if phrases such as “notarial deed,” “certificate number,” “notary authorization,” or “loan” are detected, it can be determined as a notarial deed. Additionally, depending on a combination of phrases such as “payment order,” “recommendation for performance,” “service,” or “confirmation,” it can be determined as a payment order or a decision recommending performance. In another embodiment, the document type determination may be performed by considering, in addition to keyword patterns, the layout structure of text blocks, specific table titles, the location of seals or stamps, the number of pages, the format of the heading at the top of the document, the repeatability of the formatting, or the results of an image-based classification model. That is, the document type determination may be implemented in a rule-based, AI-based, or combined form.

[0128] According to one embodiment of the present disclosure, a computing device (100) can determine whether a text layer exists and the amount of text (S200, S210). Here, “text layer” may refer to a layer of character information included within a PDF or electronic document that can be directly extracted. Additionally, “amount of text” may be defined as the number of characters, the number of valid tokens, the number of text blocks recognizable in a specific page or the entire document, or the number of semantic unit texts of a certain length or longer. In one embodiment, if a text layer exists and the amount of text is greater than or equal to a predetermined threshold, the computing device (100) may determine that the document is a machine-readable electronic document. On the other hand, if a text layer does not exist, or if a text layer exists but its amount is excessively small and unsuitable for substantial structured extraction, the computing device (100) may determine that the document is a scanned copy or an image-centric document. The predetermined threshold may be set based on, for example, the number of valid characters per page, the number of characters based on the entire document, the ratio of characters excluding spaces, or the ratio of valid characters in combinations of numbers / Korean / English. However, this is merely an example, and settings may vary depending on the document type.

[0129] According to one embodiment of the present disclosure, a computing device (100) can directly extract text data when a text layer exists and is greater than or equal to a threshold value (S220). In one embodiment, the text data (210) may be extracted as a string of lines, text in paragraphs, text fragments with coordinate information added, or location-based text corresponding to a specific block. In another embodiment, the text data (210) may be extracted while preserving a table or item structure. For example, if a case number is located in the upper right area or party information is listed under a specific title, the location information may be preserved together to improve the accuracy of subsequent extraction. Additionally, since the directly extracted text data (210) often does not entail OCR errors, it may be advantageous for relatively fast and accurate field capture.

[0130] According to one embodiment of the present disclosure, a computing device (100) may convert an input document (200) into a page-by-page image and perform character recognition processing when there is no text layer or when the threshold is lower (S230). At this time, “conversion into a page-by-page image” may be a process of rendering each page of the input document (200) into independent image data (220). The image data (220) may be generated at the original resolution or at an enlarged resolution to increase character recognition accuracy. Additionally, preprocessing such as grayscale conversion, contrast adjustment, noise removal, tilt correction, binarization preprocessing, region segmentation, paragraph boundary reinforcement, and seal or background removal may be applied to the image data (220). In one embodiment, if the document type is a notarized document, binarization preprocessing may be applied more actively to increase type contrast. In another embodiment, for document types where specific forms are repeated, such as a resident registration abstract or a corporate registration certificate, region-based segmentation may be applied to prioritize the recognition of the area where the main fields are located. Furthermore, the Page Segmentation Mode (PSM), language settings, layout recognition strategies, or post-processing correction rules of the character recognition processing can be configured differently depending on the document type. In other words, the above step can be understood as an adaptive preprocessing and adaptive OCR pipeline designed to secure text data of suitable quality for subsequent structuring, going beyond the level of simply converting images into text.

[0131] According to one embodiment of the present disclosure, a computing device (100) may perform a first extraction process (S310) based on an artificial intelligence model. In the present disclosure, the “first extraction process” may be defined as a process that extracts enforcement-related information by considering the semantic structure, context, correlation between items, or visual arrangement of an input document (200). In one embodiment, the first extraction process may be performed using a multimodal language model. For example, the computing device (100) may convert the input document (200) into a page-by-page image and input an extraction schema corresponding to the converted image and the identified document type into a multimodal language model to obtain structured data (230) or a result corresponding to the first structured data in a predetermined format. At this time, the “extraction schema” may include the name of the field to be extracted, the data type of the field, the dependency relationship between fields, whether it is a required item, a repeatable array structure, exception rules by document type, etc. For example, for the type of judgment, the case number, court name, creditor, debtor, order, principal, interest rate, and confirmation status can be set as priority extraction targets, and for the type of notarial document, the certificate number, notarization date, loan amount, interest rate, and late payment damages rate can be set as priority extraction targets. In addition, for payment orders or decisions recommending performance, the date of confirmation, date of service, claimed amount, and names of parties can be set as core fields. The above first extraction process may be constrained so as not to supplement information not specified in the document through inference, masked personal information may be extracted while maintaining the masked state or displayed separately, and for amount notations, the original expression and a normalizable expression may be extracted in parallel.

[0132] According to one embodiment of the present disclosure, a computing device (100) may perform a rule-based second extraction process (S320). In the present disclosure, the “second extraction process” may be defined as a process for detecting a specific field from text data (210) using a predefined dedicated extraction function or a set of regular expression patterns. In one embodiment, the second extraction process may select a dedicated extraction function corresponding to a determined document type and apply a plurality of regular expression patterns included in the dedicated extraction function to the text data (210) to generate second structured data. For example, the event number may be extracted using a specific year-event code-serial number format, and the amount may be extracted using a combined numerical and Korean amount expression, an amount expression including a comma, a “won” unit notation, etc. In addition, interest rates can be extracted from expressions such as “annual ○%,” “annual interest rate ○ percent,” and “late payment penalty ○%,” and dates can be extracted by combining with contexts such as “YYYY.MM.DD,” “YYYY year M month D day,” “delivery date,” and “fixed date.” Furthermore, the second extraction process of the present disclosure may have a multi-stage fallback structure. That is, if the first pattern fails for a single extraction target field, it may attempt extraction by automatically switching to the second pattern or the third pattern. This allows for the response to variations in document format, OCR errors, irregular spaces, hyphen-like characters, masking symbols, variations in amount format, and address history expressions. For example, the hyphen normalization pattern can treat various hyphen-like characters equally, the space flexible matching pattern can handle irregular OCR spaces, and the masking response pattern can recognize the corresponding field even when part of a resident registration number or corporate registration number is obscured by '*', '○', '-', etc. Additionally, the address history pattern can be implemented as a rule for selecting the latest address among multiple address histories listed in a resident registration abstract.

[0133] Furthermore, according to one embodiment of the present disclosure, a single character recognition processing alone may not be sufficient to reliably obtain information at a level usable for subsequent processing from an input document (200) related to compulsory execution. This is because the documents handled in compulsory execution practice are not limited to electronic documents of the same format. For example, while there are digital documents containing a text layer, such as judgments, payment orders, or decisions recommending performance, there may also be documents, such as notarial certificates, where character recognition accuracy may be degraded due to the influence of scan status, resolution, background noise, type contrast, or seal overlapping. Additionally, since documents such as resident registration abstracts and corporate registration certificates may include table formats, repeating items, multi-column compositions, or multi-row structures, it may be difficult to secure stable extraction results for all document types using only the same character recognition settings. Therefore, the present disclosure may adopt an adaptive data acquisition structure that directly utilizes text data when there is a sufficient text layer, and switches to page-by-page image conversion and character recognition processing when there is no or insufficient text layer. In addition, when processing character recognition, at least one of the resolution scale factor, page splitting mode, whether binarization preprocessing is performed, or other image preprocessing conditions may be applied differently depending on the document type. As such, the present disclosure can mitigate problems such as digit misrecognition, hyphen variations, space separation, date format errors, or partial omission of identification numbers that are prone to occur in single OCR-based methods through a preprocessing branching structure corresponding to document characteristics.

[0134] Furthermore, according to one embodiment of the present disclosure, it may be difficult to reliably obtain all fields necessary for enforcement practice using only a single rule-based extraction method. This is because even information having the same legal meaning may be expressed in different forms depending on the document type, author, court form, writing conventions, or contextual arrangement. For example, even the same party may be referred to as “plaintiff,” “creditor,” “applicant,” or other expressions, and even the same claim amount may be recorded in different formats within the order section, the claim statement section, the main text explanation section, or within a notarial deed clause. Additionally, interest information may be expressed in various ways, such as “annual ○%,” “annual interest rate ○ percent,” “late payment penalty ○%,” or “annual ○% until full repayment,” and interest rate conversions over multiple periods may also have different descriptive styles depending on the document. In such an environment, while rule-based extraction may be advantageous in terms of formal accuracy, its flexibility may be limited for fields requiring contextual interpretation, fields with strong layout dependency, or fields whose expression style has been changed by new forms. Accordingly, the present disclosure does not exclusively use a rule-based second extraction process, but can be combined with an AI model-based first extraction process that has strengths in contextual understanding and visual layout interpretation. For example, the identification of party roles with strong contextuality, the interpretation of amounts in order areas, and the identification of contextual connectivity between multiple interest ranges can be performed by the AI ​​model-based first extraction process, while the capture of event numbers, identification numbers, amount formats, date formats, or specific label-based fields can be supplemented by the rule-based second extraction process. In this way, contextual understanding and responsiveness to format variations, which may be weak with a single rule-based method, can be supplemented through combination with the first extraction process.

[0135] Additionally, according to one embodiment of the present disclosure, the first extraction process and the second extraction process may not be limited to a relationship where one is replaced by the other if the other fails. In one embodiment, the two extraction processes may be operated interchangeably. For example, if a response error, parsing failure, or omission of a specific field occurs in the first extraction process based on an artificial intelligence model, it may automatically switch to the second extraction process based on a rule. However, in another embodiment, the two extraction processes may be performed on the same input document (200), and the results of each process may be compared field by field to determine the final structured data (230), operating as a combined hybrid structure. In this case, fields where both results match may be confirmed with relatively high confidence, fields where only one result exists may have the corresponding extraction result adopted, and fields where both results do not match may be processed according to a preset priority, confidence, format completeness, or subsequent user review procedure. Compared to a simple fallback structure, this combined structure may provide the possibility of automatic error detection, the possibility of assigning risk flags at the field level, and the possibility of integrating mutually complementary information. For example, if a first extraction process based on an artificial intelligence model accurately captures the court name through context understanding and a second extraction process based on rules accurately captures the interest rate and case number, the computing device (100) can combine the two results to form more complete structured data (230). Conversely, if different amounts are extracted for the same field, they can be classified as subjects for user review after automatic risk detection. As such, the present disclosure may include a technical structure that goes beyond simple alternative fallback and improves extraction reliability by using multiple extraction engines in combination and mutually verifying the results.

[0136] In addition, according to one embodiment of the present disclosure, the step of normalizing structured data (230) based on a schema of predefined enforcement case data (240) may not be merely an incidental step for matching the output format, but may be a core step that enables the generation of subsequent application documents and automatic input into the electronic litigation portal to be reused based on the same data. Generally, structured data (230) immediately after document extraction may have different field configurations depending on the document type, and even if the fields are the same, the perspective of expression may differ. For example, "plaintiff" in a judgment and "creditor" in a notarial deed may have the same role in subsequent processing, but may be expressed with different names or different structures in the extraction results for each document. In addition, interest information may be expressed as a combination of multiple periods and interest rates rather than a single value, and expressions with an undetermined end date, such as "until full repayment," need to be converted into a form that can be directly used by the subsequent calculation module. In addition, address information, corporate information, case number, court name, etc. derived from multiple debtors or multiple documents need to be integrated into a single case unit data. If such normalization is not performed, the application generation module must repeatedly have separate interpretation logic for each document type, and the electronic litigation portal automatic input module must also perform field processing at the original document level again, making it easy for data inconsistencies and redundant logic to occur between modules. On the other hand, when structured data (230) is converted into normalized enforcement case data (250) according to the standard schema, the application generation automatic generation step can generate forms for each type based on the same field structure, and the electronic litigation portal automatic input step can also perform field mapping based on the same data structure. In other words, the normalized enforcement case data (250) can function as a data hub that is commonly referenced by multiple downstream processing modules, rather than being an intermediate object that simply stores the results of document extraction.This structure enables the parallel generation of multiple execution type applications for the same case, automatic calculation of interest, designation of third-party debtors, and automatic input into the electronic litigation portal to be connected as a single data axis.

[0137] In addition, according to one embodiment of the present disclosure, the normalization may be implemented by a domain-specific post-processing engine. For example, it may include processing to convert amounts written in Korean into numerical data, processing to decompose court names or case numbers into constituent elements, processing to distinguish between natural persons and legal persons based on the format difference between personal identification numbers and corporate registration numbers, processing to separate multiple debtors into an array form, or processing to convert expressions of interest with an undetermined end date into reserved values ​​or null values. Such post-processing can be understood as a step that ensures the usability of enforcement domain-specific data, which is difficult to immediately secure solely through the general output of a general document understanding model. That is, the core of the present disclosure may lie not merely in reading information from legal documents, but in rearranging the read information into a structure that can be directly used for drafting enforcement applications and inputting into the electronic litigation portal.

[0138] According to one embodiment of the present disclosure, a computing device (100) can cross-compare a first extraction result and a second extraction result by field (S330). In the present disclosure, “cross-compare” may refer to a process of comparing field values ​​derived from different extraction sources to determine whether a field is confirmed, selected, withheld, or subject to user review. In one embodiment, the computing device (100) can confirm a field if the corresponding field values ​​match. For example, if both an artificial intelligence model and a rule-based engine extract the same case number, the same creditor name, or the same amount, the field may be determined to have high reliability. On the other hand, if the corresponding field values ​​do not match, the computing device (100) may select one of the values ​​based on at least one of the success of the first extraction process and the second extraction process, a preset priority, or reliability, or classify the field as subject to user review (S332). In this case, “reliability” may be a score calculated by the model itself, an empirical score set based on the combination of document type and field type, or a comprehensive indicator reflecting the completeness of the number format, the consistency of the date format, and logical consistency with other fields. Additionally, “pre-set priority” may be configured such that, for example, fields with strong formal regularity, such as case numbers, amounts, and interest rates, prioritize rule-based results, while fields with strong contextuality, such as party role distinctions or order interpretation, prioritize AI model-based results. In another embodiment, the computing device (100) may adopt the value when only one extraction result exists and record the state where the other side is missing as separate metadata.

[0139] According to one embodiment of the present disclosure, a computing device (100) may generate structured data (230) including at least one of party information (231), billing amount information (232), or interest information (233) included in an input document (200) (S300). The structured data (230) may be individual extraction results prior to cross-comparison, or data aggregating field values ​​that have been determined or selected after cross-comparison. That is, the structured data (230) may be understood as a common intermediate representation for using the extraction processing results as input for subsequent normalization. In one embodiment, the structured data (230) may be stored in JSON format, object list format, key-value mapping format, or table format, and each field may have not only a value but also information such as extraction source, extraction method, reliability, whether review is required, or original document location information attached. Such meta-information may be utilized for subsequent user review, modification, audit log recording, or retraining data generation.

[0140] According to one embodiment of the present disclosure, a computing device (100) can normalize structured data (230) based on a schema of predefined enforcement case data (240) (S400). At this stage, the “predefined enforcement case data (240)” can function as a reference structure specialized for the enforcement domain. For example, the predefined enforcement case data (240) may include a case identification field, a creditor array, a debtor array, an address field, a corporation field, a claim amount field, a plurality of interest period arrays, an enforcement title type field, a third-party debtor candidate field, a document source field, etc. The computing device (100) can generate normalized enforcement case data (250) by mapping the structured data (230) to the reference structure. In one embodiment, automatic distinction between natural persons and corporations may be performed based on the number of digits, patterns, masking formats, etc. of a personal identification number or a corporate registration number. Additionally, the amount written in Hangul included in the claim amount information (232) can be converted into numeric data, multiple interest periods can be structured into an array of start dates, end dates, and interest rates, and expressions with an undetermined end date, such as “until full repayment,” can be converted into a predetermined reserved value or a Null value. Furthermore, information regarding multiple debtors can be separated into an array structure of individual objects rather than being connected as a single string. In other embodiments, processing may also be included to decompose court names or case numbers by components, unify legal terms of the same meaning into pre-set terms, or merge fields extracted from multiple documents for the same case to integrate them into a single normalized enforcement case data (250). Thus, the normalization step can function not merely as a simple format conversion, but as a technical step for creating a common data hub for the subsequent automation of enforcement tasks.

[0141] Additionally, according to one embodiment of the present disclosure, a computing device (100) may display structured data (230) or normalized enforcement case data (250) on a user interface and reflect information modified or confirmed by the user. In this case, the user interface may display the original text by field, extraction results, extraction method, fields requiring review, comparison results, modification history, etc. For example, for fields where the first extraction result and the second extraction result are different, the two values ​​may be presented side by side, and the user may be allowed to select which value to adopt or directly modify the value. In another embodiment, the original document image and the extracted field may be displayed in conjunction so that the user can immediately verify the basis of the original text. Such user review steps do not compromise the automation structure of the present disclosure but can be understood as a correction loop for high-reliability case processing and exception document response. The method of the present disclosure does not necessarily involve the user's thought process for review itself, and the above correction loop may be implemented by data or input existing as a result of the user review.

[0142] Additionally, according to one embodiment of the present disclosure, the first extraction process based on the artificial intelligence model may be performed on the server side. For example, a computing device (100) transmits an input document (200) in the form of encoded data to a server-side extraction module, and the server-side extraction module converts each page of the input document (200) into an image, and then inputs the image and a dedicated prompt for each document type into a multimodal language model to obtain a structured response. At this time, the structured response may include a block in JSON format, and the computing device (100) may extract and parse the JSON block from the response. In one embodiment, if a failure to parse the JSON block is detected, the computing device (100) may re-call the multimodal language model by reinforcing the prompt or redirecting the response format. Additionally, the parsed result may be normalized to match a preset data type and field structure and returned as structured data (230) or first structured data. According to this server-side processing structure, page-by-page image conversion of the input document (200), document type detection, structured response verification, retries, and subsequent processing compatibility can be configured as a single processing flow.

[0143] In summary, the embodiment illustrated in FIG. 6 provides a multi-layered document understanding and case data generation structure that includes determining the type of an input document (200), acquiring text / image data, AI-based extraction, rule-based extraction, cross-comparison of extraction results, generation of structured data (230), schema-based normalization of predefined enforcement case data (240), and a user confirmation procedure. This structure enables the stable acquisition of normalized enforcement case data (250) suitable for enforcement practice, even in an environment where the format of enforcement title documents is diverse and document quality is uneven.

[0144] Additionally, according to one embodiment of the present disclosure, at least some steps of the execution document processing may be performed by an artificial intelligence agent (AI agent). Here, the term “artificial intelligence agent” does not refer solely to a single model itself, but may refer to a software execution entity that determines the next execution action based on at least one of a target state, processing rules, reliability, document status, user input, or external system response, and sequentially calls or controls a plurality of processing modules. For example, the artificial intelligence agent may autonomously orchestrate at least some steps of determining the type of an input document (200), acquiring text data (210) or image data (220), generating structured data (230), generating normalized execution case data (250), generating an application document file (260), or preparing for input into an electronic litigation portal (320).

[0145] According to one embodiment of the present disclosure, the method of the present disclosure may further include the step of an artificial intelligence agent autonomously performing at least some of the following: determining the type of the input document, generating structured data, normalization, generating an application document, or inputting into an electronic litigation portal, and receiving user confirmation or approval at each pre-set checkpoint. In one embodiment, the artificial intelligence agent may operate in a Human-in-the-Loop structure in which it forms checkpoints between a plurality of processing steps and receives user confirmation or approval at each checkpoint. Here, a “checkpoint” may be understood as a logical control point configured to receive a review or approval input before determining whether to execute a subsequent step. For example, the checkpoint may be set after the generation of structured data (230), before the finalization of normalized enforcement case data (250), after the presentation of the execution type recommendation result, after the designation of a third-party debtor, before or after the generation of the application document file (260), or before the execution of automatic input to the electronic litigation portal (320). In one embodiment, the checkpoint may be dynamically generated in response to at least one of a decrease in reliability, field inconsistency, competition among multiple recommendations, inclusion of sensitive information, planned execution of an irreversible external operation, or the need for authorization verification.

[0146] According to one embodiment of the present disclosure, an artificial intelligence agent may comprehensively evaluate the reliability of the results produced at each step, document completeness, field matching, asset recommendation results, or the readiness for portal input, and determine whether to automatically proceed to a subsequent step, request user review, or rollback to a specific step. For example, if the results of the first extraction process and the second extraction process both match, the artificial intelligence agent may automatically proceed to a subsequent normalization step; if the principal field or interest rate field does not match, it may classify it as a subject for user review and delay the subsequent application generation step until an approval input is received. Additionally, if there are multiple recommendations for execution types or seizure types, or if the priorities among the recommended execution types are close, the agent may present multiple recommendations to the user and proceed to the next step after receiving a selection or approval input.

[0147] Additionally, according to one embodiment of the present disclosure, in the above-described human intervention structure, user intervention does not refer to the user's thought process itself, but may be received as an input in a form that the computing device (100) can process, such as approval, rejection, modification, selection, priority adjustment, or exception handling instructions. For example, the user may confirm a specific field included in the structured data (230), adopt only some of the recommended seizure types, modify third-party debtor information, or approve the execution of automatic input to the electronic litigation portal (320). The result of the approval or modification may be reflected again as an input value or control condition in a subsequent step, and the artificial intelligence agent may continue to perform the remaining steps based on the result.

[0148] In addition, according to one embodiment of the present disclosure, the artificial intelligence agent can enhance security, accountability, and practical suitability through the human intervention structure. For example, steps that are relatively reversible and internal processing-oriented, such as document extraction and normalization, may be performed autonomously, while steps that have significant practical externalities or are difficult to reverse, such as determining a third-party debtor, selecting at least one attachment or execution type, determining the creation of an application document file (260), or executing automatic input to the electronic litigation portal (320), may be configured to be performed only after user approval. According to such a structure, the benefits of automation of the artificial intelligence agent and controllability through user review can be secured simultaneously.

[0149] Additionally, according to one embodiment of the present disclosure, the artificial intelligence agent may be implemented as a single agent or distributed among multiple detailed agents, such as a document identification agent, an extraction agent, a normalization agent, a recommendation agent, a document generation agent, and a portal linkage agent. In this case as well, the intermediate output results of each detailed agent may be linked through a common data structure, for example, structured data (230) or normalized enforcement case data (250). Accordingly, the artificial intelligence agent in the present disclosure is not limited to a specific implementation form and can be understood as a concept encompassing various software execution structures that autonomously control the enforcement document processing flow while accommodating human intervention-based review at checkpoints.

[0151] FIG. 7 is a flowchart illustrating the steps for automatically generating application documents according to one embodiment of the present disclosure.

[0152] The application automatic generation technology of the present disclosure can be understood as a processing flow that allows the previously generated normalized enforcement case data (250) to be directly converted into application document files (260) corresponding to multiple types of enforcement, rather than leaving it as simple storage data. That is, the normalized enforcement case data (250) can function as a result of the document extraction engine and as common input data for the application creation system. Such a structure provides technical and practical efficiency in enforcement practice where it is necessary to repeatedly create different types of enforcement applications for the same case.

[0153] In the present disclosure, “execution type” may refer to a type of application document that can be generated based on normalized compulsory execution case data (250). For example, the execution type may include types of seizure of claims such as seizure of deposit claims, seizure of wage claims, seizure of severance pay claims, seizure of virtual assets, seizure of lease deposit claims, seizure of credit card accounts receivable, seizure of accounts receivable from business partners, seizure of automobiles, seizure of stocks, seizure of dividend claims, seizure of insurance claims, etc. Additionally, the execution type may include other types of execution such as forced auction of real estate, seizure of movable property, and registration on a list of defaulters. However, the above enumeration is merely illustrative, and new types of compulsory execution applications used in practice may be added in the future, detailed types may be subdivided, or modified types tailored to the forms of specific institutions or courts may be further included.

[0154] In addition, according to one embodiment of the present disclosure, a computing device (100) can determine asset-related information of a debtor by cross-analyzing an input document (200) including an enforcement title document and a credit report. Furthermore, based on the determination result, it can automatically recommend a combination of an executable type of seizure or an execution type. At this time, the input document (200) may refer to an input document excluding the credit report. The asset-related information of a debtor may include at least one of a type of residence, employment status, vehicle registration status, business operation status, financial institution usage information, insurance subscription information, or accounts receivable-related information. That is, in one embodiment, the method of the present disclosure may further include the step of determining at least one of the debtor's type of residence, employment status, vehicle registration status, business operation status, financial institution usage information, insurance subscription information, or accounts receivable-related information by cross-analyzing an enforcement title document and a credit report, and automatically recommending a combination of executable execution types among a plurality of execution types based on the determination result. Here, “credit investigation report” may refer to an investigation document containing information related to the debtor’s assets or living and employment status, and may include, for example, at least one of the following: type of residence, employment status, business registration status, vehicle registration status, traces of financial institution usage, traces of insurance subscription, information related to card sales, or information related to business partners. However, the name or format of the credit investigation report is not limited thereto and may include asset search results provided by an external investigation agency, verification data related to finance, employment, vehicles, or addresses, or equivalent data sets.

[0155] According to one embodiment of the present disclosure, a computing device (100) may link residential information, workplace information, or registered property information regarding the same debtor by matching debtor identification information extracted from an input document with identification information included in a credit investigation report for cross-analysis. For example, for cross-analysis, the computing device (100) may link party information (231), claim amount information (232), interest information (233), case number, debtor identification information, or address information extracted from an enforcement title document with residential type, employment status, vehicle registration status, business operation status, or other asset-related clues obtained from a credit investigation report. Here, “cross-analysis” does not mean simple parallel viewing, but can be understood as an analysis process including at least one of verifying the consistency of identification information regarding the same debtor, determining whether facts match between multiple data sources, deriving a correspondence relationship between asset clues and objects subject to seizure, or calculating the possibility of parallel execution of multiple execution types. For example, if a specific debtor is identified as a natural person in an enforcement title document and the debtor's first few digits of their resident registration number, current address, name of the company where they are employed, or vehicle ownership information is confirmed in a credit investigation report, the computing device (100) can integrate the information into asset clues for the same debtor.

[0156] According to one embodiment of the present disclosure, a computing device (100) can determine asset-related information, such as the debtor's type of residence, employment status, and vehicle registration status, based on the cross-analysis results. Here, the “type of residence” may be classified, for example, into self-occupied residence, rented residence, Jeonse, monthly rent, commercial lease, residence under a family member’s name, or unknown status. Additionally, the “employment status” may be classified, for example, into a state such as currently employed, retired, self-employed, corporate representative, daily worker, business unknown, or existence of multiple workplaces. Additionally, the “vehicle registration status” may be determined as information including at least one of the presence or absence of vehicle registration, vehicle type, registered name, or ownership of multiple vehicles. In another embodiment, the computing device (100) may also determine additional asset indicators, such as whether insurance is in effect, whether card sales exist, whether business partners exist, dividend potential, and circumstances related to stocks or virtual assets. The items listed above are merely exemplary items of information and are described to include content with a similar meaning to each item; the names of each item may be modified as necessary. Ultimately, the present disclosure does not present the raw information recorded in the credit report as is, but may include a step of reinterpreting it into meaningful asset status information from an enforcement perspective.

[0157] According to one embodiment of the present disclosure, a computing device (100) can automatically recommend a combination of executable execution types among a plurality of execution types based on the determination result of the asset-related information. Here, the “combination of executable execution types” may refer to a set of one or more execution types derived by considering at least one of the debtor’s asset status, the possibility of identifying a third-party debtor, the possibility of the existence of a target for seizure, the possibility of legal or practical execution (execution), the possibility of expected recovery, or the possibility of parallel execution of multiple types. In one embodiment, the combination of execution types may include a combination of executable seizure types, and may include at least two combinations of seizure of deposit claims, seizure of wage claims, seizure of severance pay claims, seizure of lease deposit claims, seizure of automobiles, seizure of credit card accounts receivable, seizure of business partner accounts receivable, or seizure of insurance claims. For example, if the debtor’s residence type is determined to be a leased residence and landlord information is identifiable, the possibility of seizure of lease deposit claims may be derived and recommended. In addition, if employment status is confirmed and workplace information can be identified, seizure of wage receivables or severance pay receivables may be recommended. Furthermore, if vehicle registration is confirmed, seizure of automobiles may be recommended. Additionally, if traces of card sales or business operation information exist, seizure of credit card accounts receivables or accounts receivables from business partners may be recommended. In another embodiment, if traces of financial institution usage exist, seizure of deposit receivables may be recommended, and if traces of insurance subscription exist, seizure of insurance claim receivables may be recommended.

[0158] In addition, according to one embodiment of the present disclosure, the computing device (100) may not recommend only a single type of seizure, but may produce a recommendation plan that combines multiple types of seizure. For example, if employment status and rental residence information are confirmed simultaneously, a combination of seizure of wage claims and seizure of rental deposit claims may be recommended. In addition, if vehicle registration status and deposit transaction circumstances are confirmed together, a combination of seizure of automobiles and seizure of deposit claims may be recommended. In one embodiment, the recommended combination may be sorted according to at least one of priority, expected recovery potential, difficulty of preparing for execution, ease of identifying third-party debtors, expected processing period, or cost-effectiveness. Through this, the user may receive a proposal for multiple execution strategies that can be carried out in parallel or sequentially in the same case, rather than just a single seizure target.

[0159] Furthermore, according to one embodiment of the present disclosure, a computing device (100) may evaluate at least one of the feasibility of each execution type, the possibility of identifying a third debtor, or the possibility of recovery based on asset-related information, and may calculate a priority for the plurality of execution types based on the evaluation result. Such priority may be provided to a user.

[0160] Here, the “feasibility” of each type of enforcement may refer to whether or to what extent a specific type of enforcement can proceed to actual compulsory enforcement proceedings. For example, feasibility may be evaluated based on at least one of the following criteria: the likelihood of the existence of specific assets or property rights of the debtor; whether essential information required for the relevant type of enforcement has been secured; whether basic requirements necessary for drafting an application have been met; or the feasibility of practical preparations for initiating enforcement proceedings.

[0161] Furthermore, according to one embodiment of the present disclosure, “possibility of identifying a third-party debtor” may be a concept that evaluates the existence and degree of specificity of a third-party debtor in a specific type of execution. For example, in the case of attachment of wage claims, the possibility of identifying the debtor’s company name, business information, workplace information, or the entity paying the wages may be considered; in the case of attachment of lease deposit claims, the name or corporate name of the lessor, the address of the leased property, or the availability of information that can presume the lease relationship may be considered. Additionally, in the case of attachment of deposit claims, the circumstances of transactions with financial institutions; in the case of attachment of insurance claim claims, information on the insurance company or insurance contract; and in the case of attachment of accounts receivable, the possibility of identifying business partners may each be used as evaluation factors for the possibility of identifying a third-party debtor. That is, the possibility of identifying a third-party debtor can be understood as an indicator that evaluates whether the debtor is identifiable to a level necessary for electronic litigation input or application document drafting, going beyond merely presuming the existence of the third-party debtor.

[0162] Additionally, according to one embodiment of the present disclosure, “recoverability” may refer to whether there exists an amount that can actually be recovered or a possibility of successful recovery when a specific type of execution is performed. For example, recoverability may be evaluated based on factors such as the possibility of asset existence, the scale of asset value, the relationship with the claimed amount, the risk of extinction or disposal, or recovery performance in past similar cases. In one embodiment, if the debtor’s employment status is stably verified and the entity paying the salary is clear, the recoverability of a wage garnishment may be evaluated as relatively high. Additionally, if tenancy is verified and the existence of a lease deposit is probabilistically acknowledged, the recoverability of a lease deposit garnishment may be evaluated as high. On the other hand, even if vehicle registration is verified, the recoverability of a vehicle garnishment may be evaluated as relatively low or at an intermediate level depending on the possibility of the vehicle being used as collateral, uncertainty of its actual value, or ease of disposal.

[0163] According to one embodiment of the present disclosure, the evaluations may be performed using a rule-based method, a scoring method, a statistical method, a machine learning-based method, or a combination thereof. For example, a computing device (100) may assign a pre-set weight to at least one of residential type, employment status, vehicle registration status, financial institution usage information, insurance subscription information, or accounts receivable-related information, and calculate an execution feasibility score, a third-party debtor identification feasibility score, and a recovery feasibility score for each execution type. In another embodiment, the computing device (100) may apply a threshold value to each evaluation item to classify it into grades such as “possible,” “average,” or “difficult,” and then synthesize them to calculate a priority for each execution type. In yet another embodiment, the evaluation results may be expressed in the form of a single numeric score, a multiple indicator vector, ranking information, a recommendation grade, or a recommendation / non-recommendation flag.

[0164] In addition, according to one embodiment of the present disclosure, a computing device (100) can calculate a priority by comparing evaluation results for each of a plurality of execution types. For example, a high priority may be assigned to a wage claim attachment if it is evaluated that both the feasibility of execution and the possibility of identifying a third-party debtor are high, and the possibility of recovery is also high; a medium priority may be assigned to a lease deposit claim attachment if the landlord can be identified but the size of the deposit is uncertain; and a low priority may be assigned to a vehicle attachment if vehicle registration is confirmed but the possibility of recovery is low. In addition, if it is advantageous to perform a plurality of execution types simultaneously, the computing device (100) can calculate a combination priority by considering not only the score of each individual execution type but also the complementary relationship or parallel possibility when performing the combination. For example, if employment status and rental residence information are confirmed simultaneously, a combination of wage claim attachment and lease deposit claim attachment may be recommended as a high priority, and if financial institution usage information and insurance subscription information exist together, a combination of deposit claim attachment and insurance claim attachment may be recommended.

[0165] In addition, according to one embodiment of the present disclosure, the priority calculation result is not merely used as an internal judgment value, but can be provided to the user along with the basis for recommendation. For example, the computing device (100) may present explanatory information for a specific execution type or combination of execution types, such as “high possibility of wage claim seizure due to confirmation of employment company,” “possibility of security deposit claim seizure due to confirmation of lease circumstances,” or “possibility of vehicle seizure due to confirmation of vehicle registration, but possibility of recovery is relatively low.” By providing the basis for recommendation in this manner, the user can utilize the recommendation result not merely as reference information, but as information to assist in actual decision-making.

[0166] Additionally, according to one embodiment of the present disclosure, the priority of multiple execution types may be adjusted according to user settings or policy information. For example, the user may select at least one of a policy that prioritizes the possibility of rapid recovery, a policy that prioritizes execution types that are easy to identify third-party debtors, a policy that prioritizes multiple parallel executions, or a policy that prioritizes recovery efficiency relative to application costs. In this case, the computing device (100) may calculate different priorities according to the policy even for the same asset-related information. Therefore, the priority calculation of the present disclosure is not limited to producing a fixed result, but can be understood as a recommendation logic that can be flexibly adjusted according to the purpose and strategy of enforcement practice. According to one embodiment of the present disclosure, the recommendation of execution types is not automatically determined but may be subject to user review or approval. For example, the computing device (100) may present the grounds for recommendation to the user along with the recommended combination of seizure types. The grounds for the above recommendation may be rule-based descriptions, for example, such as “Verification of rental residence → Security deposit can be seized,” “Identification of current company → Salary can be seized,” or “Vehicle registration verification → Vehicle can be seized.” Furthermore, the user may select only some of the recommended combinations, adjust priorities, or exclude specific types. In one embodiment, such a structure may be combined with the Human-in-the-Loop structure described above, and a subsequent application generation step may be initiated when an AI agent generates recommendations and the user approves them. In one embodiment, the method of the present disclosure may further include the step of receiving a user selection or approval for a recommended combination of enforcement types and automatically generating an application document based on the user-selected or approved combination of enforcement types.

[0167] Additionally, according to one embodiment of the present disclosure, the result of the execution type recommendation may be included in normalized enforcement case data (250) or stored as a separate recommendation data structure. For example, a list of recommended seizure types, recommendation scores, grounds for recommendation, third-party debtor candidates, execution status, approval status, or rejection status may be included as additional fields in the normalized enforcement case data (250). In this case, the subsequent application automatic generation step may generate each application document file (260) in parallel by directly referencing the combination of recommended execution types approved by the user. Therefore, the recommendation step may function not merely as a simple guidance function, but as a preliminary decision support step that induces subsequent document generation and portal input.

[0168] In short, according to the present embodiment, by cross-analyzing legal rights information extracted from an enforcement title document with asset status information obtained from a credit investigation report, it is possible to go beyond simple automatic document generation and automatically recommend enforcement types or combinations of enforcement types with a high probability of actual execution. This structure has technical significance in that it assists or automates asset assessment and enforcement type selection—which previously relied on the experience of personnel in compulsory execution practice—based on data.

[0169] According to one embodiment of the present disclosure, a computing device (100) can automatically generate an application document corresponding to at least one selected execution type among a plurality of execution types based on normalized execution case data (250) (S500). This step can be applied not only when generating an application for a single type, but also when multiple execution types are simultaneously selected for the same case. For example, if the seizure of a deposit claim and the seizure of a wage claim are to be carried out simultaneously in a single case, two types of application documents can be generated in parallel from the same normalized execution case data (250). In this case, common data is shared, but additional fields or third-party debtor information required for each type can be branched out and processed separately. Such a structure reduces redundancy in the creation of application documents and enables consistent management of multiple procedures for the same case.

[0170] According to one embodiment of the present disclosure, a computing device (100) can load creditor information, debtor information, and claim amount information from normalized enforcement case data (250) (S510). Here, “load” may mean an operation of bringing field values ​​of normalized enforcement case data (250) already stored in memory or storage into a processing context that can be used by subsequent application generation logic. For example, creditor information may include a creditor name, address, identification number, or whether it is a legal entity, and debtor information may include a debtor name, arrangement of multiple debtors, address, identification number, or whether it is a legal entity. Additionally, claim amount information may include principal, interest, costs, amount reflecting partial repayment, execution costs, or results of balance calculation. In another embodiment, the load step may further include, in addition to simple field reading, a process of selecting required fields according to the selected execution type, excluding fields not required for the form, and processing the display format of specific fields to be suitable for document output.

[0171] According to one embodiment of the present disclosure, if the selected execution type involves a third debtor, the computing device (100) may specify third debtor information corresponding to the selected execution type (S520). Here, “third debtor” may mean an institution, business entity, individual, or legal entity that can be identified as the counterparty to legal execution in relation to the debtor’s property or claim. For example, in the attachment of deposit claims, a financial institution may be the third debtor; in the attachment of wage claims, an employer or company may be the third debtor; and in the attachment of lease deposit claims, a landlord may be the third debtor. Additionally, in the attachment of credit card sales receivables, a card company or settlement business entity may be the third debtor; in the attachment of business partner sales receivables, a business partner company may be the third debtor; and in the attachment of insurance claim receivables, an insurance company may be the third debtor. In one embodiment, third debtor information may be directly entered by the user. In another embodiment, the computing device (100) may present a candidate for a third debtor based on a pre-established database, user history, customer profile, case-specific stored information, or external inquiry results, and may designate one based on a user selection. The user selection may refer to the result of the user's cognitive choice and not to the user's thought process itself. In yet another embodiment, multiple third debtors may be designated simultaneously depending on the type of execution. For example, when garnishing deposit claims against multiple financial institutions in parallel, multiple banks may be designated as separate third debtors to generate multiple application documents.

[0172] According to one embodiment of the present disclosure, a computing device (100) can automatically calculate an interest amount by applying period-specific interest rates to multiple interest periods based on interest information included in normalized enforcement case data (250), and can automatically process a transition from the interest rate at the time of filing the complaint to the interest rate after the judgment is finalized (S530). In the present disclosure, “interest information” does not mean only a single interest rate, but may encompass multiple interest periods, the start and end dates of each period, the interest rate applied per period, whether the end date is undetermined, and interest rate transition rules based on the finalized date. In one embodiment, the computing device (100) can calculate an interest amount per period based on the number of days or months corresponding to each period by using multiple interest period information that has already been organized into an array structure in the normalization step. For example, it can automatically transition to applying the agreed interest rate or the claimed interest rate from the time of filing the complaint until the pronouncement of the judgment, and to applying the statutory delay damages interest rate or the post-judgment interest rate after the judgment is finalized. In another embodiment, the interest amount may be recalculated by reflecting partial payment, interim payment, the interest calculation end date, the document creation date, the execution application date, or a user-specified date. In yet another embodiment, the result of the interest calculation may be used as part of the total claim amount directly stated in the application document, or it may be generated together in the form of a separate supplementary document of the calculation basis. As such, the automatic interest calculation step can be understood as a key step for automating a practically frequent and error-prone part of drafting an application for compulsory execution.

[0173] According to one embodiment of the present disclosure, a computing device (100) may generate an application document file (260) by mapping the loaded information, the third debtor information, and the calculated interest amount to a predetermined application form (S540). Here, the “application form” may refer to a document template that includes field configurations, paragraph structures, item arrangements, title formats, attachment display formats, etc., required for a specific execution type. In one embodiment, the application form may be implemented as at least one of an HTML (HyperText Markup Language) based template, a document object model, a word processor template, a PDF generation template, or a dedicated rendering structure. Additionally, the computing device (100) may generate an application document file (260) by inserting field values ​​into the form and adjusting line breaks, amount display formats, name masking, address splitting formats, or table titles. The application document file (260) may be stored in a PDF format, a printable document format, or an intermediate data format for electronic litigation input. In another embodiment, when multiple application documents are generated for the same event, each application document file (260) may be generated as an individual file or may be packaged into a single document bundle.

[0174] In addition, according to one embodiment of the present disclosure, when two or more of a plurality of execution types are selected simultaneously, the computing device (100) can generate application documents corresponding to each execution type in parallel. Here, “parallel” is not limited to the meaning of being performed simultaneously in terms of computing resources, but can comprehensively refer to a structure in which branching logic for each type is applied independently while sharing common data, and result documents are generated sequentially or in parallel. For example, when preparing to simultaneously seize deposit claims, accounts receivable from business partners, and insurance claims against a single debtor, multiple application documents can be generated by reusing common creditor information, common debtor information, and common interest calculation results, while applying only the third-party debtor and form parts differently according to the execution type. This structure may be particularly useful in processing large volumes of cases or performing multi-type executions.

[0175] Ultimately, the embodiment illustrated in FIG. 7 has technical significance in that it data-continues the document extraction step and the document creation step by providing a structure that automatically generates type-specific application document files (260) by directly reusing normalized enforcement case data (250). This structure reduces the repetitiveness of manual document creation and the complexity of interest calculation, and enables multiple types of enforcement applications for the same case to be executed based on consistent data.

[0177] FIG. 8 is a flowchart illustrating the automatic input step of an electronic litigation portal according to one embodiment of the present disclosure.

[0178] The automatic input technology of the present disclosure can be understood as a subsequent step of linking the generated application document file (260) or normalized enforcement case data (250) with the input fields of the electronic litigation portal (320). That is, the present disclosure does not end with document extraction and application generation, but technically mitigates double input between application creation and portal input by directly connecting already generated data to the external portal input stage. In particular, it can be distinguished from a general document generation system in that it directly interacts with the web page structure of the electronic litigation portal (320) using a browser extension program (310).

[0179] In the present disclosure, “browser extension program (310)” may refer to a client-side software module that runs in a user’s web browser environment and can perform data exchange and field control with an external web page. For example, the browser extension program (310) may include a Manifest-based extension package, a Service Worker, a Content Script, a local storage module, a message delivery module, an encryption module, a field mapping module, or a user interface module. Additionally, “electronic litigation portal (320)” may refer to a web-based system for court submission procedures and may include web input fields for receiving specific case information, party information, claim amount, purpose of application, attachment information, etc. In the present disclosure, “DOM (Document Object Model) element” may be used as a concept including input boxes, selection boxes, text areas, hidden fields, buttons, or other interactive elements that constitute the web page of the electronic litigation portal (320).

[0180] According to one embodiment of the present disclosure, a computing device (100) can automatically map and input data included in an automatically generated application document into an input field of an electronic litigation portal (320) (S600). In one embodiment, the step may be performed based on data included in an application document file (260). In another embodiment, instead of reinterpreting the application document file (260), the step may be performed based on normalized enforcement case data (250) used to generate the application document or intermediate data for mapping that is substantially identical thereto. That is, the direct input source for automatic input may be a document file or structured data corresponding to the document generation. This configuration may be flexibly selected depending on the implementation environment. For example, JSON data for input having the same content may be generated simultaneously with the generation of the document file, and a browser extension (310) may input the data into the portal field using the JSON data for input.

[0181] According to one embodiment of the present disclosure, a computing device (100) may convert data included in an application document into a predetermined data format and store it in a storage area of ​​a browser extension (310) (S610). Here, “a predetermined data format” may refer to a structured data format suitable for use by the browser extension (310) for subsequent field mapping. For example, the data format may be a JSON format, a key-value pair format, a dictionary format, a field list format, or an internal serialization format. In one embodiment, each field, such as case name, resident registration number, creditor name, debtor name, claim amount, address, case number, interest rate, etc., may be represented as a single key-value pair. Additionally, repeatable information, such as multiple debtors or multiple interest periods, may be represented in an array structure. In another embodiment, the storage area may be a local storage, session storage, encryption storage, message queue, or temporary memory of the browser extension (310). Additionally, in the storage step, the data may be encrypted or masked to protect personal information or sensitive case information. This storage structure can serve to reliably configure the data transfer path between the application creation system and the browser extension (310).

[0182] According to one embodiment of the present disclosure, a browser extension program (310) can map each field of the stored data to a corresponding DOM (Document Object Model) element of the electronic litigation portal (320) through a content script inserted into a web page of the electronic litigation portal (320) (S620). In the present disclosure, “mapping” may refer to a process that establishes a correspondence relationship between the stored data item and the actual input structure of the electronic litigation portal (320), going beyond the level of simple string copying. In one embodiment, the browser extension program (310) can map the case name field to the case name input box, the resident registration number field to the resident registration number input box, and the claim amount field to the amount input box. In another embodiment, when a single data field needs to be divided and entered into multiple DOM elements, for example, when a resident registration number is separated into front and back digit input fields, the browser extension program (310) can divide the necessary substrings from a single value and correspond them to each of the multiple DOM elements. Additionally, corresponding selection values ​​or events can be generated and mapped for field types that are not simple text inputs, such as dropdown selection fields, radio buttons, checkboxes, or date selection fields. In another embodiment, mapping with DOM elements can be performed based on field IDs, name attributes, CSS (Cascading Style Sheets) selectors, XPath (XML Path Language), label text, surrounding context, or page structure templates. Accordingly, the above step may be implemented in a static table manner tailored to a specific portal page structure, or in a dynamic navigation manner that responds to changes in page structure.

[0183] According to one embodiment of the present disclosure, a computing device (100) may receive confirmation from a user by displaying the mapped result as a preview (S630). Step S630 can be understood as a step that allows the user to perform a final review before automatic input, and this should also be understood as receiving the result of the user's choice rather than the user's thought process itself. In one embodiment, a browser extension (310) may display the stored field and the corresponding DOM element side by side and preview the value to be entered in the form of a table, list, highlight, overlay, or pop-up window. Additionally, if the field name does not match or the value is empty, the corresponding item may be displayed in a warning state. In another embodiment, the user may selectively enter only specific fields during the preview step, exclude them from automatic input, or modify them directly. In yet another embodiment, the preview may be rendered as a temporary display over the actual input field of the electronic litigation portal (320) so that the user can review it immediately on the page. Such a user confirmation step may function as a technical supplementary means to simultaneously ensure the stability of automation and user controllability.

[0184] According to one embodiment of the present disclosure, a computing device (100) may input mapped data into the DOM element in response to user confirmation (S640). In one embodiment, a browser extension (310) may insert a value into a text field and generate a Change event, an Input event, a Focus event, or a Blur event so that the web page recognizes the input correctly. In another embodiment, for elements such as dropdowns or checkboxes, setting a selected value and a click event may be performed together. In yet another embodiment, where the input order is important, for example in a portal structure where a lower field is dynamically created after the upper field is entered, sequential input and waiting logic by field group may be applied. Thus, the present disclosure may be configured to perform automatic input in a manner suitable for the operational context of the web application, rather than simple DOM manipulation.

[0185] In addition, according to one embodiment of the present disclosure, a browser extension (310) may include error handling logic that detects errors occurring during the input process and retries a predetermined number of times when an error occurs. For example, if a corresponding field cannot be found, if page loading is not completed, if the input page is changed due to session expiration, if the field structure is temporarily delayed in creation by a dynamic script, or if a validation error message is displayed after input, the browser extension (310) may perform re-search, re-input, retry after waiting for a certain period of time, display a user warning, or save an exception log. In addition, data stored in the storage area may be encrypted and stored, and may be automatically deleted after input is completed or expire after a certain period of time. Such error handling and data protection configurations can contribute to increasing the practical usability of browser-based portal integration.

[0186] Ultimately, the embodiment illustrated in FIG. 8 provides a structure that links an application document file (260) or normalized enforcement case data (250) to an actual input field of an electronic litigation portal (320) through a browser extension (310). This structure data-links the document generation system and the web portal submission procedure, thereby reducing the re-entry of the same data and enabling the automation of the portal submission stage.

[0188] The steps mentioned in the foregoing description may be further subdivided into additional steps or combined into fewer steps, depending on the embodiment of the present disclosure. Additionally, some steps may be omitted as necessary, and the order of the steps may be changed.

[0190] Meanwhile, a computer-readable medium storing a data structure is disclosed according to an embodiment of the present disclosure.

[0191] The aforementioned data structure may be stored in memory in the present disclosure, executed by a processor, and transmitted and received by a network unit.

[0192] A data structure can refer to the organization, management, and storage of data that enables efficient access and modification of data. A data structure can refer to the organization of data for solving specific problems (e.g., data analysis, data retrieval, data storage, data modification). A data structure may also be defined by physical or logical relationships between data elements designed to support specific data processing functions. Logical relationships between data elements may include connections between user-defined data elements. Physical relationships between data elements may include actual relationships between data elements physically stored on a computer-readable storage medium (e.g., a permanent storage device). Specifically, a data structure may include sets of data, relationships between data, and functions or instructions applicable to the data. Through an effectively designed data structure, a computing device can perform operations while minimizing the use of the device's resources. Specifically, through an effectively designed data structure, a computing device can increase the efficiency of operations, reading, insertion, deletion, comparison, exchange, and retrieval.

[0193] Data structures can be classified into linear and non-linear data structures based on their form. A linear data structure is one where only one piece of data is connected to the next. Linear data structures can include lists, stacks, queues, and deques. A list can refer to a set of data that maintains an internal order. Lists can include linked lists. A linked list is a data structure where data is connected in a line, with each piece of data possessing a pointer. In a linked list, the pointer can contain information regarding the connection to the next or previous data. Depending on its form, a linked list can be represented as a singly linked list, a doubly linked list, or a circular linked list. A stack is a data arrangement structure that allows for restricted access to data. A stack can be a linear data structure where data can be processed (e.g., insertion or deletion) only at one end of the structure. Data stored in a stack can be a Last-In, First-Out (LIFO) data structure, meaning that the later an item is entered, the sooner it is retrieved. A queue is a data arrangement structure that allows for limited access to data; unlike a stack, it can be a FIFO (First in First Out) data structure where data stored later is retrieved later. A deque is a data structure that can process data at both ends.

[0194] Non-linear data structures can be structures where multiple data are connected after a single piece of data. Non-linear data structures may include graph data structures. A graph data structure can be defined by vertices and edges, and an edge may include a line connecting two different vertices. Graph data structures may include tree data structures. A tree data structure may be a data structure where there is only one path connecting two different vertices among the multiple vertices included in the tree. In other words, it may be a data structure that does not form a loop in a graph data structure.

[0195] Throughout this specification, the terms artificial intelligence-based model, computational model, neural network, network function, and neural network may be used interchangeably. Hereinafter, they will be described uniformly as neural network. A data structure may include a neural network. Furthermore, a data structure including a neural network may be stored on a computer-readable medium. A data structure including a neural network may also include data preprocessed for processing by the neural network, data input to the neural network, weights of the neural network, hyperparameters of the neural network, data obtained from the neural network, activation functions associated with each node or layer of the neural network, loss functions for learning the neural network, etc. A data structure including a neural network may include any of the components disclosed above. That is, a data structure including a neural network may be configured to include all or any combination thereof, such as data preprocessed for processing by the neural network, data input to the neural network, weights of the neural network, hyperparameters of the neural network, data obtained from the neural network, activation functions associated with each node or layer of the neural network, and loss functions for learning the neural network. In addition to the configurations described above, a data structure including a neural network may include any other information that determines the characteristics of the neural network. Furthermore, the data structure may include any form of data used or generated during the computational process of the neural network, and is not limited to the foregoing. A computer-readable medium may include a computer-readable recording medium and / or a computer-readable transmission medium. A neural network may be composed of a set of interconnected computational units that may generally be referred to as nodes. These nodes may also be referred to as neurons. A neural network is composed of at least one node.

[0196] A data structure may include data input to a neural network. A data structure including data input to a neural network may be stored on a computer-readable medium. Data input to a neural network may include training data input during the neural network learning process and / or input data input to a neural network after training is complete. Data input to a neural network may include pre-processed data and / or data subject to pre-processing. Pre-processing may include a data processing process for inputting data into a neural network. Accordingly, a data structure may include data subject to pre-processing and data generated by pre-processing. The aforementioned data structure is merely an example, and the present disclosure is not limited thereto.

[0197] The data structure may include weights of the neural network. (In this specification, weights and parameters may be used interchangeably.) The data structure including the weights of the neural network may be stored on a computer-readable medium. The neural network may include multiple weights. The weights may be variable and may be varied by a user or an algorithm to enable the neural network to perform a desired function. For example, if one or more input nodes are interconnected to a single output node by respective links, the output node may determine the data value output from the output node based on values ​​input to the input nodes connected to the output node and weights set on the links corresponding to each input node. The aforementioned data structure is merely an example and the present disclosure is not limited thereto.

[0198] As an example rather than a limitation, weights may include weights that vary during the neural network learning process and / or weights for which neural network learning is completed. Weights that vary during the neural network learning process may include weights at the start of the learning cycle and / or weights that vary during the learning cycle. Weights for which neural network learning is completed may include weights for which the learning cycle is completed. Accordingly, a data structure containing the weights of a neural network may include a data structure containing weights that vary during the neural network learning process and / or weights for which neural network learning is completed. Therefore, the weights and / or combinations of each weight described above are included in the data structure containing the weights of a neural network. The aforementioned data structure is merely an example and the present disclosure is not limited thereto.

[0199] Data structures containing the weights of a neural network may be stored on a computer-readable storage medium (e.g., memory, hard disk) after undergoing a serialization process. Serialization may be a process of converting a data structure into a form that can be stored on the same or different computing devices and later reconstructed for use. A computing device may serialize the data structure to transmit and receive data over a network. A serialized data structure containing the weights of a neural network may be reconstructed on the same or different computing devices through deserialization. Data structures containing the weights of a neural network are not limited to serialization. Furthermore, data structures containing the weights of a neural network may include data structures designed to increase computational efficiency while minimizing the use of computing device resources (e.g., B-Tree, R-Tree, Trie, m-way search tree, AVL tree, Red-Black Tree in non-linear data structures). The foregoing is merely an example and the present disclosure is not limited thereto.

[0200] The data structure may include hyperparameters of the neural network. The data structure including the neural network hyperparameters may be stored on a computer-readable medium. The hyperparameters may be variables that are varied by the user. The hyperparameters may include, for example, a learning rate, a cost function, the number of learning cycle iterations, weight initialization (e.g., setting the range of weight values ​​subject to weight initialization), and the number of hidden units (e.g., the number of hidden layers, the number of nodes in the hidden layers). The aforementioned data structure is merely an example, and the present disclosure is not limited thereto.

[0202] FIG. 9 is a brief and general schematic diagram of an exemplary computing environment in which embodiments of the present disclosure may be implemented.

[0203] Although the present disclosure has been described as generally being implementable by a computing device, a person skilled in the art will be well aware that the present disclosure may be implemented in combination with computer-executable instructions and / or other program modules that can be executed on one or more computers and / or as a combination of hardware and software.

[0204] Generally, a program module includes routines, programs, components, data structures, etc., that perform a specific task or implement a specific abstract data type. Furthermore, a person skilled in the art will be well aware that the method of the present disclosure can be implemented in other computer system configurations, including single-processor or multi-processor computer systems, minicomputers, mainframe computers, as well as personal computers, handheld computing devices, microprocessor-based or programmable consumer electronics, etc. (each of which may be connected to and operated with one or more associated devices).

[0205] The embodiments described in this disclosure may also be implemented in a distributed computing environment in which tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules may be located in both local and remote memory storage devices.

[0206] Computers typically include various computer-readable media. Any medium accessible by a computer may be a computer-readable medium, and such computer-readable media include volatile and non-volatile media, transitory and non-transitory media, and removable and non-removable media. By example, but not limiting, computer-readable media may include computer-readable storage media and computer-readable transmission media. Computer-readable storage media include volatile and non-volatile media, transitory and non-transitory media, and removable and non-removable media implemented by any method or technique for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer-readable storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, DVD (digital video disk) or other optical disk storage devices, magnetic cassettes, magnetic tapes, magnetic disk storage devices or other magnetic storage devices, or any other media that can be accessed by a computer and used to store desired information.

[0207] Computer-readable transmission media typically include all information transmission media that implement computer-readable instructions, data structures, program modules, or other data, etc., on a modulated data signal, such as a carrier wave or other transport mechanism. The term modulated data signal means a signal in which one or more of the characteristics of the signal are set or modified to encode information within the signal. By example, not limiting, computer-readable transmission media include wired media, such as wired networks or direct-wired connections, and wireless media, such as acoustic, RF, infrared, and other wireless media. Any combination of the media described above is also considered to be within the scope of computer-readable transmission media.

[0208] An exemplary environment for implementing various aspects of the present disclosure, including a computer (1102), is shown, wherein the computer (1102) includes a processing unit (1104), system memory (1106), and a system bus (1108). The system bus (1108) connects system components, including system memory (1106) (but not limited thereto), to the processing unit (1104). The processing unit (1104) may be any processor among various commercial processors. Dual processors and other multiprocessor architectures may also be used as the processing unit (1104).

[0209] The system bus (1108) may be any of several types of bus structures that can be additionally interconnected to a local bus using any of the memory bus, peripheral bus, and various commercial bus architectures. System memory (1106) includes read-only memory (ROM) (1110) and random access memory (RAM) (1112). The basic input / output system (BIOS) is stored in non-volatile memory (1110), such as ROM, EPROM, EEPROM, etc., and this BIOS includes basic routines that help transfer information between components within the computer (1102) at times such as during startup. The RAM (1112) may also include high-speed RAM, such as static RAM, for caching data.

[0210] The computer (1102) also includes an internal hard disk drive (HDD) (1114) (e.g., EIDE, SATA)—this internal hard disk drive (1114) may also be configured for external use within a suitable chassis (not shown)—a magnetic floppy disk drive (FDD) (1116) (e.g., for reading from or writing to a removable diskette (1118)), and an optical disk drive (1120) (e.g., for reading from a CD-ROM disk (1122) or reading from or writing to other high-capacity optical media such as a DVD). The hard disk drive (1114), the magnetic disk drive (1116), and the optical disk drive (1120) may each be connected to the system bus (1108) by a hard disk drive interface (1124), a magnetic disk drive interface (1126), and an optical drive interface (1128). The interface (1124) for implementing an external drive includes at least one or both of USB (Universal Serial Bus) and IEEE 1394 interface technologies.

[0211] These drives and associated computer-readable media provide non-volatile storage of data, data structures, computer-executable instructions, etc. In the case of a computer (1102), the drives and media correspond to storing any data in a suitable digital format. Although the description of computer-readable media above refers to HDDs, removable magnetic disks, and removable optical media such as CDs or DVDs, a person skilled in the art will know that other types of computer-readable media, such as zip drives, magnetic cassettes, flash memory cards, cartridges, etc., may also be used in exemplary operating environments and that any of these media may contain computer-executable instructions for performing the methods of the present disclosure.

[0212] A number of program modules, including an operating system (1130), one or more application programs (1132), other program modules (1134), and program data (1136), may be stored in the drive and RAM (1112). All or part of the operating system, application, module and / or data may also be cached in RAM (1112). It will be well known that the present disclosure may be implemented in various commercially available operating systems or combinations of operating systems.

[0213] The user can input commands and information into the computer (1102) through one or more wired / wireless input devices, such as a pointing device like a keyboard (1138) and a mouse (1140). Other input devices (not shown) may include a microphone, an IR remote control, a joystick, a game pad, a stylus pen, a touch screen, etc. These and other input devices are often connected to the processing unit (1104) via an input device interface (1142) connected to the system bus (1108), but may also be connected via other interfaces such as a parallel port, an IEEE 1394 serial port, a game port, a USB port, an IR interface, etc.

[0214] A monitor (1144) or other type of display device is also connected to the system bus (1108) via an interface such as a video adapter (1146). In addition to the monitor (1144), the computer generally includes other peripheral output devices (not shown), such as speakers, a printer, and so on.

[0215] The computer (1102) may operate in a networked environment using a logical connection to one or more remote computers, such as remote computer(s) (1148), via wired and / or wireless communication. The remote computer(s) (1148) may be a workstation, a computing device computer, a router, a personal computer, a portable computer, a microprocessor-based entertainment device, a peer device, or other conventional network node, and generally include many or all of the components described for the computer (1102), but for brevity, only the memory storage device (1150) is illustrated. The illustrated logical connection includes a wired / wireless connection to a local area network (LAN) (1152) and / or a larger network, e.g., a wide area network (WAN) (1154). Such LAN and WAN networking environments are common in offices and companies and facilitate enterprise-wide computer networks, such as intranets, all of which can be connected to a global computer network, e.g., the Internet.

[0216] When used in a LAN networking environment, the computer (1102) is connected to a local network (1152) via a wired and / or wireless communication network interface or adapter (1156). The adapter (1156) may facilitate wired or wireless communication to the LAN (1152), and the LAN (1152) may also include a wireless access point installed therein to communicate with the wireless adapter (1156). When used in a WAN networking environment, the computer (1102) may include a modem (1158), be connected to a communication computing device on the WAN (1154), or have other means to establish communication through the WAN (1154), such as through the Internet. The modem (1158), which may be an internal or external and a wired or wireless device, is connected to the system bus (1108) via a serial port interface (1142). In a networked environment, the program modules described for the computer (1102) or parts thereof may be stored in a remote memory / storage device (1150). It will be well known that the illustrated network connection is exemplary and that other means of establishing a communication link between computers may be used.

[0217] The computer (1102) operates to communicate with any wireless device or object that is deployed and operated via wireless communication, for example, a printer, scanner, desktop and / or portable computer, PDA (portable data assistant), communication satellite, any equipment or place associated with a wireless detectable tag, and a telephone. This includes at least Wi-Fi and Bluetooth wireless technologies. Accordingly, the communication may be a predefined structure as in a conventional network, or simply ad hoc communication between at least two devices.

[0218] Wi-Fi (Wireless Fidelity) enables connectivity to the Internet and other sources without wires. Wi-Fi is a wireless technology, similar to a cell phone, that allows devices, such as computers, to transmit and receive data indoors and outdoors—that is, anywhere within the coverage area of ​​a base station. Wi-Fi networks use a wireless technology called IEEE 802.11 (a, b, g, etc.) to provide secure, reliable, and high-speed wireless connections. Wi-Fi can be used to connect computers to each other, to the Internet, and to wired networks (using IEEE 802.3 or Ethernet). Wi-Fi networks can operate in unlicensed 2.4 and 5 GHz wireless bands, for example, at data rates of 11 Mbps (802.11a) or 54 Mbps (802.11b), or in products that include both bands (dual band).

[0219] Those skilled in the art of the present disclosure will understand that information and signals may be represented using any various different technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols, and chips that may be referenced in the above description may be represented by voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields or particles, or any combination thereof.

[0220] Those skilled in the art will understand that the various exemplary logical blocks, configurations, modules, logics, processors, means, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be implemented by electronic hardware, various forms of computer programs or design code (referred to herein as software for convenience), or a combination of all such. To clearly illustrate this interoperability between hardware and software, the various exemplary logical blocks, configurations, modules, logics, processors, means, circuits, and steps have been generally described above in relation to their functions. Whether such functions are implemented in hardware or software depends on the design constraints imposed on the specific application and the overall system. Those skilled in the art may implement the functions described in this disclosure in various ways for each specific application, but such implementation decisions should not be interpreted as being outside the scope of this disclosure.

[0221] The various embodiments presented herein may be implemented as methods, devices, or articles manufactured using standard programming and / or engineering techniques. The term "article manufactured" includes a computer program, a carrier, or a medium accessible from any computer-readable storage device. For example, computer-readable storage media include, but are not limited to, magnetic storage devices (e.g., hard disks, floppy disks, magnetic strips, etc.), optical discs (e.g., CDs, DVDs, etc.), smart cards, and flash memory devices (e.g., EEPROMs, cards, sticks, key drives, etc.). Additionally, the various storage media presented herein include one or more devices and / or other machine-readable media for storing information.

[0222] It should be understood that the specific order or hierarchy of steps in the presented processes is an example of exemplary approaches. It should be understood that the specific order or hierarchy of steps in the processes may be rearranged within the scope of this disclosure based on design priorities. The appended method claims provide elements of various steps in a sample order, but do not imply being limited to the specific order or hierarchy presented.

[0223] Description of the presented embodiments is provided so that a person skilled in the art may use or practice the present disclosure. Various modifications to these embodiments will be apparent to a person skilled in the art, and the general principles defined herein may be applied to other embodiments without departing from the scope of the present disclosure. Thus, the present disclosure is not limited to the embodiments presented herein, but should be interpreted in the broadest possible scope consistent with the principles and novel features presented herein.

Claims

Claim 1 A method for processing enforcement documents performed by a computing device, comprising: receiving an input document including an enforcement title document; obtaining at least one of text data or image data from the input document; generating structured data including at least one of party information, claim amount information, or interest information included in the input document based on the obtained data; and normalizing the structured data based on a predefined schema of enforcement case data, wherein the step of generating the structured data includes applying at least one of an artificial intelligence model-based first extraction process or a rule-based second extraction process to the text data or image data. The method comprises the step of scanning a text area or an image area of ​​the input document to determine the type of the input document based on the matching result with a plurality of predefined keyword patterns, and applying an extraction schema corresponding to the determined type; the second extraction process is a process of selecting a dedicated extraction function corresponding to the determined type of the input document and applying a plurality of regular expression patterns included in the selected dedicated extraction function to the text data to extract second structured data; the plurality of regular expression patterns include a plurality of alternative patterns for each extraction target field and include a multi-stage fallback structure that automatically switches to a second pattern if extraction by the first pattern fails. Claim 2 In claim 1, the method wherein the execution title document comprises at least one of a judgment, a notarial deed, a payment order, or a decision recommending performance. Claim 3 A method according to claim 1, wherein the input document further comprises at least one of a resident registration abstract or a corporate registration certificate in addition to the execution title document, and the structured data further comprises at least one of address information extracted from the resident registration abstract or corporate information extracted from the corporate registration certificate. Claim 4 delete Claim 5 A method according to claim 1, wherein the step of acquiring at least one of the text data or image data comprises: a step of determining whether a text layer exists and the amount of text of the input document; a step of directly extracting the text data from the text layer if the text layer exists and the amount of text is greater than or equal to a predetermined threshold; and a step of extracting the text data by converting the input document into a page-by-page image and then performing character recognition processing if the text layer does not exist or the amount of text is less than the predetermined threshold. Claim 6 In claim 5, the step of converting the input document into a page-by-page image comprises the step of dynamically determining at least one of a resolution scale factor during image conversion, a page splitting mode for character recognition, or whether to apply binarization preprocessing based on the type of the input document. Claim 7 delete Claim 8 delete Claim 9 A method according to claim 1, wherein the plurality of regular expression patterns comprises at least one of: a hyphen normalization pattern that uniformly matches hyphen-like characters; a space flexible matching pattern that corresponds to irregular spaces; a masking matching pattern that recognizes a masked identification number; a amount extraction pattern that normalizes an amount format; or a pattern that selects the latest address from an address history. Claim 10 A method according to claim 1, wherein the step of generating the structured data further comprises the step of cross-comparing the first structured data generated by the first extraction process and the second structured data generated by the second extraction process by field. Claim 11 In claim 10, the step of cross-comparing the first structured data and the second structured data by field comprises: a step of determining the corresponding field if the corresponding field value of the first structured data and the second structured data matches; and, if the corresponding field value does not match, a step of selecting one value or classifying the corresponding field as a user review target based on at least one of the success of extraction of each of the first extraction process and the second extraction process, a preset priority, or a reliability. Claim 12 A method for processing enforcement documents performed by a computing device, comprising: receiving an input document including an enforcement title document; obtaining at least one of text data or image data from the input document; generating structured data including at least one of party information, claim amount information, or interest information included in the input document based on the obtained data; and normalizing the structured data based on a predefined schema of enforcement case data, wherein the step of normalizing the structured data based on a predefined schema of enforcement case data includes: automatically distinguishing between natural persons and legal persons based on the number of digits of a personal identification number or corporate registration number included in the party information; converting an amount written in Hangul included in the claim amount information into numerical data; structuring a start date, end date, and interest rate in an array form for each of a plurality of interest periods included in the interest information, while converting an expression of an undetermined end date into a predetermined reserved value; or at least one step of individually extracting information for each of a plurality of debtors and structuring it into an array. Claim 13 The method of claim 1 further comprises the step of displaying the structured data or the normalized structured data on a user interface and reflecting information modified or confirmed by the user. Claim 14 The method of claim 1 further comprises the step of automatically generating an application document corresponding to at least one execution type selected among a plurality of execution types based on the normalized structured data. Claim 15 In claim 14, the step of automatically generating the application document comprises: loading creditor information, debtor information, and claim amount information from the normalized structured data; specifying third-party debtor information corresponding to the selected execution type if the selected execution type involves a third-party debtor; automatically calculating an interest amount by applying period-specific interest rates to multiple interest periods based on interest information included in the normalized structured data, and automatically processing a transition from the interest rate at the time of filing the complaint to the interest rate after the judgment becomes final; and generating an application document file by mapping the loaded information, the third-party debtor information, and the calculated interest amount to a predetermined application form. Claim 16 In claim 15, the plurality of execution types include a type of seizure of claims comprising at least one of seizure of deposit claims, seizure of wage or severance pay claims, seizure of virtual assets, seizure of lease deposit claims, seizure of credit card accounts receivable, seizure of accounts receivable from business partners, seizure of automobiles, seizure of stocks, seizure of dividend claims, or seizure of insurance claims; and other types of execution comprising at least one of forced auction of real estate, seizure of movable property, and registration on a list of defaulters, wherein if two or more of the plurality of execution types are selected simultaneously, application documents corresponding to each execution type are generated in parallel. Claim 17 In claim 14, the method further comprises the step of automatically mapping and inputting data included in the automatically generated application document into an input field of the electronic litigation portal. Claim 18 A method for processing enforcement documents performed by a computing device, comprising: receiving an input document including an enforcement title document; obtaining at least one of text data or image data from the input document; generating structured data including at least one of party information, claim amount information, or interest information included in the input document based on the obtained data; normalizing the structured data based on a predefined schema of enforcement case data; automatically generating an application document corresponding to at least one selected enforcement type among a plurality of enforcement types based on the normalized structured data; and automatically mapping and inputting data included in the automatically generated application document into an input field of an electronic litigation portal, wherein the step of automatically mapping and inputting data included in the application document into an input field of the electronic litigation portal comprises: converting the data included in the application document into a predetermined data format and storing it in a storage area of ​​a browser extension program; mapping each field of the stored data to a corresponding DOM (Document Object Model) element of the electronic litigation portal through a content script inserted into a web page of the electronic litigation portal by the browser extension program; and receiving confirmation from a user by displaying the mapped result as a preview. A method comprising the step of inputting the mapped data into the DOM element in response to the user's confirmation. Claim 19 In claim 1, the method further comprises the step of determining asset-related information of the debtor by cross-analyzing the input document and the credit investigation report regarding the debtor, wherein the asset-related information includes at least one of a type of residence, employment status, vehicle registration status, business operation status, financial institution usage information, insurance subscription information, or accounts receivable-related information. Claim 20 In claim 19, the step of determining asset-related information of the debtor by cross-analyzing the input document and the credit investigation sheet concerning the debtor includes the step of linking residence information, workplace information, or registered property information concerning the same debtor by comparing the debtor identification information extracted from the input document with the identification information included in the credit investigation sheet. Claim 21 In claim 19, the method further comprises the step of automatically recommending a combination of executable execution types among a plurality of execution types based on the asset-related information. Claim 22 In claim 21, the combination of executable execution types comprises at least two combinations of the following: attachment of deposit claims, attachment of wage claims, attachment of severance pay claims, attachment of lease deposit claims, attachment of credit card accounts receivable, attachment of customer accounts receivable, attachment of automobiles, attachment of stocks, attachment of dividend claims, or attachment of insurance claims. Claim 23 In claim 21, the step of automatically recommending the combination of executable execution types comprises: evaluating at least one of the executableness, third-party debtor identification possibility, or recoverability of each execution type based on the asset-related information; and calculating the priority of the plurality of execution types based on the result of the evaluation. Claim 24 A computing device comprising at least one processor and memory; wherein the processor receives an input document including an enforcement title document, obtains at least one of text data or image data from the input document, generates structured data including at least one of party information, claim amount information, and interest information included in the input document based on the obtained data, and is configured to normalize the structured data based on a predefined schema of enforcement case data, and generating the structured data involves applying at least one of an artificial intelligence model-based first extraction process or a rule-based second extraction process to the text data or image data; A computing device comprising scanning a text area or image area of ​​the input document to determine the type of the input document based on a matching result with a plurality of predefined keyword patterns, and applying an extraction schema corresponding to the determined type, wherein the second extraction process is a process of selecting a dedicated extraction function corresponding to the determined type of the input document and applying a plurality of regular expression patterns included in the selected dedicated extraction function to the text data to extract second structured data, wherein the plurality of regular expression patterns include a plurality of alternative patterns for each extraction target field and include a multi-stage fallback structure that automatically switches to a second pattern if extraction by the first pattern fails. Claim 25 A computing device according to claim 24, wherein the processor generates first structured data based on an artificial intelligence model based on image data of the input document; generates second structured data based on rules based on text data of the input document; and further configured to determine the structured data by comparing or integrating the first structured data and the second structured data by field. Claim 26 A computing device according to claim 24, wherein the processor automatically generates an application document corresponding to at least one selected execution type among a plurality of execution types based on the normalized structured data; and is further configured to perform automatic input by mapping the data included in the application document to an input field of an electronic litigation portal. Claim 27 A computer program stored on a computer-readable recording medium, wherein the program causes a processor including a computing device to perform operations for processing an enforcement document, said operations comprising: receiving an input document including an enforcement title document; obtaining at least one of text data or image data from said input document; generating structured data including at least one of party information, claim amount information, and interest information included in said input document based on said obtained data; normalizing said structured data based on a predefined schema of enforcement case data; and automatically generating an application document corresponding to at least one selected enforcement type among a plurality of enforcement types based on said normalized structured data. A computer program stored on a computer-readable recording medium, comprising the operation of automatically mapping and inputting data included in the automatically generated application document into an input field of an electronic litigation portal, wherein the operation of automatically mapping and inputting data included in the application document into an input field of the electronic litigation portal comprises: the operation of converting the data included in the application document into a predetermined data format and storing it in a storage area of ​​a browser extension program; the operation of mapping each field of the stored data to a corresponding DOM (Document Object Model) element of the electronic litigation portal through a content script inserted into a web page of the electronic litigation portal by the browser extension program; the operation of displaying the mapped result as a preview and receiving confirmation from a user; and the operation of inputting the mapped data into the DOM element in response to the user's confirmation.

Images