Integrated System for Consistency Validation and Autonomous Correction in Standardized Mapping of Unstructured Data

Abstract

The present invention relates to the field of data processing technology for converting and linking unstructured data into standard data suitable for Enterprise Resource Planning (ERP) systems. More specifically, it relates to an integrated system for unstructured data standardization mapping consistency verification and autonomous correction that ensures stable linkage with ERP systems by converting unstructured data, such as documents, messages, and transaction records collected from various sources, into a standard column set suitable for each ERP module through preprocessing, intent classification, and context-based mapping processes, and by performing consistency verification and error correction procedures on the results. The technical gist of the present invention is that the unstructured data standardization mapping consistency verification and autonomous correction integrated system, which performs mapping from unstructured data to ERP standard data to achieve the above-mentioned purpose, comprises: a preprocessing module in which unstructured data is collected and normalized; an intent classification module in which the preprocessed data is classified according to business intents such as purchasing, sales, and production; a mapping module in which the classified data is mapped to a set of standard columns for each ERP module, and a ranking of candidate items is calculated based on context-based weights to determine the optimal item code; a consistency verification module in which the consistency of the results of the mapping module is verified and the data is excluded from ERP input if an error is detected; an autonomous correction module in which user correction or automatic remapping is performed by providing a recommended correction value when an error is detected; and an ERP linkage module in which only data that has passed the consistency verification module and the autonomous correction module is input into the ERP system, and traceability is ensured by applying idempotent keys and duplicate blocking logic during the input process and recording all processing logs.

Description

Technology Field

[0001] The present invention relates to the field of data processing technology for converting and linking unstructured data into standard data suitable for Enterprise Resource Planning (ERP) systems. More specifically, it relates to an integrated system for unstructured data standardization mapping consistency verification and autonomous correction that ensures stable linkage with ERP systems by converting unstructured data, such as documents, messages, and transaction records collected from various sources, into a standard column set suitable for each ERP module through preprocessing, intent classification, and context-based mapping processes, and by performing consistency verification and error correction procedures on the results. Background Technology

[0002] In the modern business environment, data generated across various business areas such as purchasing, sales, production, and logistics is rapidly increasing, and this data often exists in unstructured formats such as emails, electronic documents, messenger conversations, and spreadsheets.

[0003] However, because existing Enterprise Resource Planning (ERP) systems are designed based on standardized data structures, it is difficult to directly input unstructured data into the ERP, making it essential for humans to manually organize it or reprocess it through intermediate systems.

[0004] Such manual work not only requires significant time and cost but also causes problems such as input errors, missing data, and inconsistencies, becoming a major cause of reduced reliability and operational efficiency in ERP systems.

[0005] Conventionally, simple data conversion tools or keyword matching-based mapping engines were used, but these failed to sufficiently reflect the contextual meaning of the data or business intent.

[0006] In particular, when the same term is used with different meanings by department or project, or when notation methods differ, cases frequently occurred where it could not be mapped to the appropriate standard item code, resulting in incorrect values ​​being reflected in the ERP system.

[0007] In addition, the lack of features to automatically verify or correct errors during the mapping process caused inconvenience, as users had to identify and fix problems retrospectively.

[0008] In particular, as ERP systems serve as a core infrastructure for corporate management where all transactions and resource management are integrated, the accuracy and consistency of input data have a significant impact on overall business operations.

[0009] Therefore, technology that minimizes errors during the data standardization and mapping stages, prevents duplicate entries or inconsistencies during ERP integration, and ensures traceability by transparently recording processing history is essential.

[0010] Recently, active research is being conducted to combine AI-based text analysis technology, natural language processing (NLP), and rule-based validation logic to grasp the meaning of unstructured data and automatically map it to ERP standard structures.

[0011] However, many existing systems are specialized for specific domains or limited to single functions (e.g., typo correction, unit conversion, etc.), and as a result, they are unable to comprehensively support the entire process from preprocessing and intent classification to mapping, consistency verification, correction, and ERP integration.

[0012] Furthermore, errors are limited to simple warnings, and autonomous correction functions, such as suggesting correction values ​​or automatic remapping, are lacking, which imposes limitations on practical application.

[0013] Due to the limitations of the aforementioned conventional technology, there is an increasing need for an intelligent data processing system that can automatically convert unstructured data into ERP standard data, guarantee its consistency, and autonomously correct errors when they occur. Prior art literature

[0014] (Patent Document 0001) KR 10-2321707 B1(Patent Document 0002) KR 10-1631032 B1 The problem to be solved

[0015] The present invention was developed to improve upon the aforementioned problems, and aims to provide an integrated system for unstructured data standardization mapping consistency verification and autonomous correction that simultaneously solves the problems of inefficiency and error caused by reliance on manual work during the process of converting unstructured data existing in various forms into standard data suitable for an ERP system, the limitations of existing mapping techniques that fail to reflect contextual meaning and business intent, the absence of a system capable of automatically verifying the consistency of mapping results and autonomously correcting errors when they occur, and the problem of reduced reliability caused by duplicate input, inconsistency, and the lack of processing history during the ERP integration process, thereby automating data processing, ensuring the accuracy and reliability of ERP input, increasing the efficiency of business execution, and allowing the processing process to be easily traced.

[0016] The objectives of the present invention are not limited to those mentioned above, and other unmentioned objectives will be clearly understood from the description below. means of solving the problem

[0017] The technical features of the present invention include an integrated system for unstructured data standardization mapping consistency verification and autonomous correction that performs mapping from unstructured data to ERP standard data to achieve the above-mentioned purpose, wherein the integrated system comprises: a preprocessing module in which unstructured data is collected and normalized; an intent classification module in which the preprocessed data is classified according to business intents such as purchasing, sales, and production; a mapping module in which the classified data is mapped to a set of standard columns for each ERP module, and a ranking of candidate items is calculated based on context-based weights to determine the optimal item code; a consistency verification module in which the consistency of the results of the mapping module is verified and the data is excluded from ERP input if an error is detected; an autonomous correction module in which user correction or automatic remapping is performed by providing a recommended correction value when an error is detected; and an ERP linkage module in which only data that has passed the consistency verification module and the autonomous correction module is input into the ERP system, and traceability is ensured by applying idempotent keys and duplicate blocking logic during the input process and recording all processing logs.

[0018] In addition, the technical feature of the above preprocessing module is that it performs typo correction, quantity and unit conversion, date format normalization, and stop word removal of the above unstructured data.

[0019] In addition, the above consistency verification module is characterized by the technical feature that three stages of column-level format verification, mapped value completeness verification, and business rule-based conformity verification are performed sequentially, and if even one is unsuitable, the entire amount of the above data is blocked from ERP integration.

[0020] In addition, the above mapping means is characterized by the technical feature of defining a standard set of columns per ERP module as a data contract unit, and approving ERP input only when the said set of columns is fully satisfied.

[0021] In addition, the processing log recorded by the above-mentioned ERP integration module includes a decision log, and the technical feature is that the decision log includes the original text of the input data, the preprocessing result, the mapping candidate list, the final confirmed value, and the error and correction history. Effects of the invention

[0022] The present invention based on the above-described configuration can expect the following effects.

[0023] By automatically normalizing typos, unit inconsistencies, variety of date formats, and stop words contained in unstructured data and converting them to conform to ERP standard data formats, it becomes possible to significantly improve the accuracy of data entered into the ERP.

[0024] By mapping to a standard set of columns for each ERP module by considering business contexts such as customer, delivery date, and item attributes, rather than simple string matching, it becomes possible to reduce incorrect code mapping and automatically determine the optimal item code.

[0025] By performing column-level format verification, mapping verification with master data, and business rule-based conformity verification in stages, it becomes possible to ensure data integrity by blocking ERP input if a non-conformity is detected at any stage.

[0026] When an error is detected, instead of simply blocking it, by calculating and providing recommended correction values ​​to the user or performing automatic remapping, it becomes possible to reduce the user's correction burden and rapidly recover erroneous data.

[0027] By applying idempotent keys and duplicate blocking logic during the ERP integration process and logging all input and decision processes, it becomes possible to prevent duplicate reflection of the same data and facilitate subsequent auditing and tracking.

[0028] By minimizing unnecessary tasks such as manual data correction, error reprocessing, and repetitive entry, it becomes possible to increase the success rate of ERP implementation, improve business processing speed, and enhance company-wide operational efficiency.

[0029] Unlike simple data mapping or ETL methods, it possesses structural features including context-based mapping, consistency verification at the column set level, and feedback that suggests correction values, thereby securing differentiation from existing technologies and enabling expanded application to various ERP system environments and additional business areas. Brief explanation of the drawing

[0030] FIG. 1 illustrates the configuration and flow paths of an integrated system for unstructured data standardization mapping consistency verification and autonomous correction according to a preferred embodiment of the present invention. FIG. 2 illustrates the overall operation flowchart from the integration system for unstructured data standardization mapping consistency verification and autonomous correction according to a preferred embodiment of the present invention until unstructured data is input into an ERP system. FIG. 3 illustrates the detailed operational structure of a mapping module in an integrated system for unstructured data standardization mapping consistency verification and autonomous correction according to a preferred embodiment of the present invention, showing the process of generating candidate items, calculating context-based weights, evaluating rankings, and determining optimal item codes. Figure 4 illustrates the detailed structure of a consistency verification module in an integrated system for unstructured data standardization mapping consistency verification and autonomous correction according to a preferred embodiment of the present invention, showing a state in which the decision to input into ERP is made through format verification, mapping verification, and business rule verification processes. FIG. 5 illustrates the feedback process of an autonomous correction module in an integrated system for unstructured data standardization mapping consistency verification and autonomous correction according to a preferred embodiment of the present invention, showing a procedure for error detection, generation of recommended correction values, user correction, or re-verification after automatic remapping. FIG. 6 illustrates the log recording structure of an ERP linkage module in an integrated system for unstructured data standardization mapping consistency verification and autonomous correction according to a preferred embodiment of the present invention, showing the process in which processing logs and decision logs are recorded and finally reflected in the ERP system. FIG. 7 illustrates the detailed operation of a preprocessing module in an integrated system for unstructured data standardization mapping consistency verification and autonomous correction according to a preferred embodiment of the present invention, showing the process of outputting normalized data through typo correction, unit conversion, date normalization, and stop word removal after inputting unstructured data. Specific details for implementing the invention

[0031] The following is a detailed description of an integrated system for unstructured data standardization mapping consistency verification and autonomous correction according to a preferred embodiment of the present invention, with reference to the attached drawings.

[0032] FIG. 1 illustrates the configuration and flow paths of an integrated system for unstructured data standardization mapping consistency verification and autonomous correction according to a preferred embodiment of the present invention; FIG. 2 illustrates the overall operation flowchart from the integrated system for unstructured data standardization mapping consistency verification and autonomous correction according to a preferred embodiment of the present invention until unstructured data is input into an ERP system; FIG. 3 illustrates the detailed operational structure of a mapping module in an integrated system for unstructured data standardization mapping consistency verification and autonomous correction according to a preferred embodiment of the present invention, showing the processes of candidate item generation, context-based weight calculation, ranking evaluation, and optimal item code determination; FIG. 4 illustrates the detailed structure of a consistency verification module in an integrated system for unstructured data standardization mapping consistency verification and autonomous correction according to a preferred embodiment of the present invention, showing the state in which the decision on whether to input into the ERP is made after undergoing format verification, mapping verification, and business rule verification processes; FIG. 5 illustrates an integrated system for unstructured data standardization mapping consistency verification and autonomous correction according to a preferred embodiment of the present invention Figure 6 illustrates the feedback process of the autonomous correction module, showing the procedure of error detection, generation of recommended correction values, user correction, or re-verification after automatic remapping; Figure 7 illustrates the log recording structure of the ERP linkage module in the integrated system for unstructured data standardization mapping consistency verification and autonomous correction according to a preferred embodiment of the present invention, showing the process in which processing logs and decision logs are recorded and finally reflected in the ERP system; and Figure 7 illustrates the detailed operation of the preprocessing module in the integrated system for unstructured data standardization mapping consistency verification and autonomous correction according to a preferred embodiment of the present invention, showing the process in which normalized data is output through typo correction, unit conversion, date normalization, and stop word removal processes after inputting unstructured data.

[0034] An integrated system for unstructured data standardization mapping consistency verification and autonomous correction according to a preferred embodiment of the present invention may be composed of a preprocessing module (100), an intent classification module (200), a mapping module (300), a consistency verification module (400), an autonomous correction module (500), and an ERP linkage module (600), as illustrated in FIG. 1.

[0036] First, we will examine the above preprocessing module (100).

[0037] The preprocessing module (100) may be responsible for the function of normalizing unstructured data collected through various channels such as KakaoTalk, messenger, email, voice recognition, and OCR into a standard data format required by the ERP system.

[0038] The preprocessing module (100) may perform the role of structuring data and minimizing errors so that mapping between ERP modules can be performed smoothly, rather than simply transmitting data input to the ERP as is.

[0039] The preprocessing module (100) includes a dictionary-based correction function, which is a function that automatically corrects typos or non-standard expressions that may occur in user input, for example, it may be possible to convert an unstructured expression such as “pair-up instruction” into a standard expression such as “work instruction”.

[0040] This correction process is not limited to simple string substitution; it may be possible to perform more precise conversions by comparing with standard item names and code systems linked to the ERP business context.

[0041] Additionally, the preprocessing module (100) includes a rule-based normalization function, which performs a procedure to unify key data such as quantity, unit, and date format into an ERP standard format, for example, it may be possible to convert a natural language expression “next Friday” into an accurate date in YYYY-MM-DD format, or to convert a “BOX” unit to “EA” by applying a conversion table.

[0042] Additionally, quantity data is corrected to satisfy minimum order units or multiple rules according to ERP business rules, which may prevent data inconsistencies during ERP input.

[0043] In addition, the preprocessing module (100) includes a stop word removal function, and expressions unrelated to ERP input, such as greetings, interjections, and special characters, are automatically removed so that only core information can be transmitted to the subsequent module.

[0044] For example, it may be possible to perform a process that removes unnecessary expressions such as “Hello” from the input “Hello, please order 100 tapes” and extracts only the core data “order 100 tapes”.

[0045] In addition, the preprocessing module (100) includes a format normalization function, and even if the data is the same, Korean and English characters, and uppercase and lowercase characters may appear mixed depending on the input user's habits, but the preprocessing module (100) can convert this into a unified format according to the ERP standard field format.

[0046] For example, an expression such as “opP tape” can be converted into the standard item name “OPP TAPE”.

[0047] In addition, the preprocessing module (100) may be able to perform intelligent processing that reflects contextual meaning by combining a machine learning-based pattern recognition function, moving away from simple rule-based processing.

[0048] For example, even if the item name is the same, it may be linked to a specific supplier code in purchasing operations and to a work process code in production operations, and the preprocessing module (100) can analyze the context of the input data and normalize it so that the correct connection can be made in the subsequent mapping step.

[0049] Accordingly, the preprocessing module (100) of the present invention can convert various forms of unstructured data into standard data required by the ERP system, remove errors in advance, and provide a foundation for the intention classification module (200) and mapping module (300) described later to operate accurately.

[0051] Next, we will examine the above intention classification module (200).

[0052] The intention classification module (200) plays the role of accurately classifying unstructured data into specific business areas such as purchasing, sales, and production in the ERP system, and can operate through complex contextual analysis and rule-based judgment rather than relying solely on simple keyword matching.

[0053] After receiving normalized input data from the preprocessing module (100), the intention classification module (200) may first perform a primary classification based on predefined business representative words.

[0054] For example, if keywords such as “order,” “order,” and “work order” are included, it may be possible to initially map them to relevant ERP modules such as purchasing, sales, and production based on this.

[0055] However, the intention classification module (200) may not stop at merely checking simple keywords, but may perform secondary classification by analyzing the structural features and context of the input data together.

[0056] For example, if a pattern such as “Customer Name + Delivery Date + Quantity” is identified within the data, it may be classified as a sales order, and if “Item Code + Order Quantity + Supplier” exists, it may be classified as a purchase order.

[0057] This makes it possible to distinguish the correct business module depending on the context, even if the same word exists.

[0058] In addition, the intention classification module (200) may additionally apply a machine learning-based classifier to make it possible to properly determine ambiguous input data.

[0059] By reflecting rules and patterns learned based on a large amount of data samples accumulated in the past, it is possible to reliably classify even complex inputs that are difficult to process with rule-based methods by utilizing context and the learned model.

[0060] For example, inputs containing a mix of vendor, item, and quantity, such as “Order 10 motor pumps from the supplier,” can be correctly classified as purchase orders through a trained model.

[0061] Furthermore, this module may be capable of separating different intentions even when they are mixed within a single input and transmitting them to the respective modules.

[0062] For example, if input such as “Order 100 steel plates and issue processing instructions at the same time” is received in a single sentence, it may be possible to route it by separating it into purchasing and production tasks.

[0063] Through this processing method, it is possible to structure unstructured inputs to suit the actual ERP business flow.

[0064] The intention classification module (200) may be able to assign a reliability score to the classification result by considering the variability and ambiguity of the input data and transmit it to a subsequent module based on this.

[0065] If the reliability is below a certain standard, it may be designed to be marked as ambiguous and undergo user correction or additional verification in the autonomous correction module (500) described later.

[0066] In addition, logs are recorded throughout the classification process, making it possible to utilize them for future error analysis or system improvement.

[0067] As a result, the intention classification module (200) performs context-based precise routing beyond simple keyword-based classification, and can be supported to accurately connect to business data for each ERP module by combining machine learning techniques and multi-layered discrimination procedures.

[0068] This enables unstructured data to be systematically classified according to business intent, and strengthens the reliability and accuracy of the entire ERP integration process.

[0070] Next, we will examine the mapping module (300) above.

[0071] The mapping module (300) may be capable of performing the function of mapping the data provided by the preprocessing module (100) and the intent classification module (200) to a set of standard columns required by each business area of ​​the ERP, and calculating the ranking of candidate items based on context-based weights to determine the optimal item code.

[0072] The mapping module (300) may be configured to define a set of standard columns required in the purchasing area, sales area, and production area of ​​the ERP as a data contract unit.

[0073] For example, it may be possible to include item codes, order quantities, supplier codes, and delivery dates in the purchasing area of ​​the ERP, item codes, order quantities, customer codes, and requested delivery dates in the sales area of ​​the ERP, and item codes, work quantities, process codes, and start dates in the production area of ​​the ERP.

[0074] The mapping module (300) may be able to operate to approve ERP input only when all of the standard column set units defined in this way are satisfied.

[0075] The mapping module (300) may include a function to generate item candidates by comparing the data corrected in the preprocessing module (100) with ERP standard information.

[0076] The generated item candidates are assigned scores based on the context of the input data, customer attributes, past order history, and the characteristics of the relevant ERP business area, and the candidate with the highest score may be confirmed as the final item code.

[0077] The mapping module (300) may be able to calculate a score by considering factors such as item name similarity, whether the name is exclusive to a customer, and recent usage frequency when calculating the ranking of item candidates.

[0078] This allows the most suitable item code to be derived even in the case of ambiguous input data, and if the data falls below a certain threshold or multiple candidates compete for the top score, it can be classified as ambiguous data and excluded from ERP input.

[0079] The mapping module (300) may be configured to record all mapping results in the decision log.

[0080] The decision log includes the original input data, preprocessing results, a list of candidate items, the final confirmed item code, and mapping basis information, which can be utilized for subsequent verification or error analysis.

[0081] These recording functions can enhance the transparency of ERP data entry and enable administrators or users to verify the legitimacy of mapping results.

[0082] Therefore, the mapping module (300) can ensure the accuracy and completeness of ERP data input through mapping based on a set of standard columns defined in each business area of ​​the ERP, context-based weighted ranking, and decision log recording functions.

[0084] Next, we will examine the above consistency verification module (400).

[0085] The consistency verification module (400) performs the function of verifying formal, logical, and business rule consistency for ERP input candidate data derived through the mapping module (300), and it may be possible to verify the completeness and consistency of the data in multiple layers before it is transmitted to the ERP.

[0086] The consistency verification module (400) systematically checks whether the input data conforms to the standard information and policies required by the ERP system, and data that does not pass the verification may be automatically blocked during the ERP integration process.

[0087] The consistency verification module (400) may first perform format verification at the column level.

[0088] For example, if a required input column is empty, the date does not conform to the YYYY-MM-DD format, or quantity data is entered as a negative value, it may be classified as a format error and immediately determined as abnormal data.

[0089] This format verification step can serve as a preprocessing role that satisfies the minimum requirements of the ERP system.

[0090] Next, it may be possible to perform a verification of the completeness of the mapped values.

[0091] This is a step of verifying whether the item code, unit, and customer code, etc. confirmed in the mapping module (300) exist in the actual ERP standard information, and if an unregistered item code or an invalid unit conversion is found, the data may be classified as an error processing target.

[0092] This makes it possible to prevent data that is inconsistent with ERP master data from entering the system in advance.

[0093] Additionally, it may be possible to perform conformity verification based on business rules.

[0094] For example, in the case of order data, if the order quantity is listed below the minimum order quantity, the delivery date violates the company's business day rules, or the price exceeds the allowed discount rate range, it may be determined as a violation of business policy and blocked from ERP input.

[0095] This conformity verification step goes beyond simple data matching and can ensure data reliability by substantially reflecting the company's internal operating rules and policies.

[0096] The consistency verification module (400) may apply a total blocking principle that excludes the entire data from the ERP integration target if even one of the verification procedures is deemed unsuitable.

[0097] Even if partially valid data exists, the entire case is treated as a failure, making it possible to prevent incomplete or erroneous data from being reflected in the ERP.

[0098] This can serve as an important technical means to enhance data integrity and ERP reliability.

[0099] In addition, the consistency verification module (400) may be able to record a clear error code and reason for errors that occurred during the verification process.

[0100] For example, it may be possible to assign systematic error codes such as V-REQ for missing required columns, M-AMBIG for insufficient item candidate scores, M-NOTFOUND for discrepancies in master data, M-UOM for inability to convert units, and B-DATE for violations of delivery date rules.

[0101] This error information may be utilized in a subsequent processing step for user correction or an automatic remapping procedure through an autonomous correction module (500).

[0102] The consistency verification module (400) stores the error data in an isolation queue so that it is not mixed with normal data, and it is possible to provide feedback on the error details to the user or administrator while ERP transmission is blocked.

[0103] This makes it possible to structurally block incorrect data from being entered into the ERP system, while simultaneously ensuring the possibility of auditing to trace the causes of errors.

[0104] In conclusion, the above consistency verification module (400) sequentially proceeds through three stages of format verification, mapping completeness verification, and business rule verification to ensure the consistency of input data in various ways, and it is possible to significantly improve the quality of data input into the ERP system.

[0106] Next, we will examine the above-mentioned autonomous correction module (500).

[0107] The autonomous correction module (500) may be capable of performing a function that supports the derivation of final data suitable for the ERP system by performing an automatic or semi-automatic correction process on data in which an error was detected during the verification step of the consistency verification module (400).

[0108] The autonomous correction module (500) can go beyond simply blocking errors to provide recommended correction values ​​and execute an automatic remapping procedure, thereby maximizing the success rate of ERP integration.

[0109] The autonomous correction module (500) may first be able to apply different correction strategies depending on the type of error.

[0110] For example, if a required value is missing, it may be possible to automatically generate candidate values ​​by referring to ERP master data, or to suggest the most suitable correction value based on the history of existing similar data.

[0111] In addition, if an item code does not exist, it may be possible to recommend top candidate item codes through an item name dictionary and context-based similarity analysis.

[0112] If unit conversion is not possible, it may be possible to configure the system to convert to an alternative unit using a standard unit conversion table or to send a correction request to a senior administrator.

[0113] In addition, the autonomous correction module (500) may be able to provide immediate feedback to the user when an error is detected.

[0114] The feedback content is configured to include the column where the problem occurred, the error reason code, the suggested correction value, and a correction guide, making it possible for the user to intuitively decide whether to correct it.

[0115] If the user accepts the proposed correction value, re-verification and re-mapping are performed with the same RUN_ID, and it may be possible to ensure idempotency.

[0116] Furthermore, this module may be capable of performing automatic correction under certain conditions without user input.

[0117] For example, simple typo correction or frequently occurring standard conversions (e.g., “pair-up instruction” → “work instruction”, BOX → EA unit conversion, etc.) may be corrected and remapped by the system itself.

[0118] On the other hand, in the case of complex business rule violations or multiple candidate ambiguities, instead of automatic correction, a list of top candidates may be provided to the user, making it possible to perform a correction procedure based on the selection.

[0119] The autonomous correction module (500) can record the history of all correction processes in a decision log, making it possible to facilitate future auditing and tracking.

[0120] This record includes the original data, error type, proposed correction value, final confirmed value, and user intervention status, which can be utilized for future data quality improvement or the advancement of machine learning-based correction rules.

[0121] In addition, the autonomous correction module (500) may reflect the same error as a learning data when it occurs repeatedly, so that an automatically improved correction result may be provided when similar data is input thereafter.

[0122] This makes it possible to gradually improve correction accuracy and efficiency over time.

[0123] In conclusion, the autonomous correction module (500) can overcome the limitations of a simple error blocking method and may be formed into a cyclic structure of providing recommended correction values, user or system correction, re-verification and re-mapping, log accumulation and learning.

[0125] Next, we will examine the above ERP integration module (600).

[0126] The ERP integration module (600) controls the safe input of only data that has passed through the consistency verification module (400) and the autonomous correction module (500) into the ERP system, and it is possible to ensure data integrity and traceability during the ERP integration process.

[0127] The ERP integration module (600) acts as an interface with the ERP, and when inputting data, it is possible to apply an idempotent key and operate a duplicate blocking logic so that even if the same data is requested multiple times, it is managed to be reflected only once.

[0128] The above idempotent key is generated by combining the original message ID, processing module, and data hash value, and it is possible to prevent duplicate reflection in the ERP even if the same data is repeatedly entered.

[0129] The duplicate blocking logic can validate new input requests based on already recorded idempotent keys, block ERP input if a duplicate exists, and perform only retry processing if necessary.

[0130] This structure makes it possible to maintain the integrity and consistency of ERP data.

[0131] In addition, the ERP integration module (600) records all input data processing processes as logs, and the recorded logs may include decision logs.

[0132] The above decision log includes the original input data, preprocessing results, mapping candidate list, final confirmed value, and error occurrence and correction history, making it possible to ensure traceability throughout the entire data transformation and input process.

[0133] This makes it possible to clearly record not only whether data has been reflected in the ERP, but also the entire process of correction, mapping, and verification performed in the pre-input stage.

[0134] The ERP integration module (600) can be formed in a structure that performs a call to the ERP system in a single request unit and allows exactly one input for each data set.

[0135] When a response is received from the ERP, the result is immediately reflected in the log, and in the event of failure, it is stored in an isolation queue along with the reason for the error, and it may be possible to perform administrator notifications or automatic retry procedures.

[0136] Even during this process, duplicate reflection is prevented, and it may be possible to manage error data separately so that it is not transmitted to the ERP.

[0137] In addition, the above ERP integration module (600) can be extended and applied in the same way to integration with external systems other than ERP, and it is possible to reliably transmit data through a REST API gateway and check the success or failure status in real time.

[0138] Such a structural design can fundamentally prevent duplicate, omitted, and incomplete data entries that may occur during the ERP input stage, while simultaneously ensuring the reliability and audit traceability of ERP data.

[0140] Additionally, we examine the possible implementation forms of the above modules.

[0141] The above preprocessing module (100) may be implemented as a processor mounted on a server device or a cloud-based data processing engine, and may be formed as a software configuration in which a text normalization algorithm, a dictionary-based typo correction program, a unit conversion and a date conversion program are executed.

[0142] In addition, it may be possible to provide real-time large-scale data processing performance by implementing it as a dedicated hardware device combined with an input gateway.

[0143] The above intention classification module (200) may be implemented as an artificial intelligence server equipped with a machine learning model, or formed as a keyword matching program and a natural language processing-based classification program.

[0144] In addition, it may be possible to perform the function of classifying multiple input data simultaneously by combining it with dedicated hardware, including GPU acceleration devices.

[0145] The above mapping module (300) can be implemented as a database mapping engine operating on an ERP integration gateway server, and can be formed as software including a candidate item generation algorithm, a context-based weight calculation program, and a standard column set mapping program.

[0146] In addition, it is implemented as a dedicated device at the front end of the ERP system, making it possible to determine the optimal item code through real-time integration with master data.

[0147] The above consistency verification module (400) may be implemented as a verification server or an ERP security gateway device placed immediately before ERP input.

[0148] This module can be formed as a software configuration including a format verification script, a master data comparison program, and business rule verification logic, and it may be possible to verify in multiple layers whether the input data matches ERP policies.

[0149] The above autonomous correction module (500) can be implemented as a correction server linked with a user interface (UI) terminal, and can be formed as software including a recommended correction value generation program, an error correction algorithm, and a user input reflection program.

[0150] In addition, it may be implemented as a dedicated hardware device including an automatic remapping engine, making it possible to perform the correction process in real time after error detection.

[0151] The above ERP integration module (600) may be implemented as a gateway server or network interface device directly connected to the ERP, and may be formed as software including an ERP API calling program, an idempotent key generation program, and a log recording program.

[0152] In addition, it is implemented with an extensible structure linked to external systems, so that all processing history is recorded in the decision log, and it is possible to prevent duplication and ensure stable integration during the ERP input process.

[0154] Based on the configuration described above, the operating state of the present invention is examined.

[0155] The present invention makes it possible to perform an integrated operational process of converting unstructured data flowing in through various channels into standard data suitable for an ERP system, verifying it, and safely inputting it.

[0156] First, unstructured data collected from KakaoTalk, email, OCR, etc., passes through a preprocessing module (100) to perform typo correction, unit conversion, date format normalization, stop word removal, etc., and can be processed into basic structured data required for ERP input.

[0157] The preprocessed data is transmitted to the intention classification module (200) so that it can be automatically classified according to business intentions such as purchasing, sales, and production.

[0158] In this process, not only simple keyword matching but also context and format analysis are utilized, making it possible to route data to the appropriate ERP module.

[0159] Afterwards, the mapping module (300) maps the classified data to a standard set of columns for each ERP module, generates candidate items, and applies context-based weighted ranking to determine the optimal item code.

[0160] During the mapping process, ERP input is approved only when all required columns are satisfied, and incomplete or uncertain data can be marked as an error.

[0161] The consistency verification module (400) sequentially performs format verification, mapping completeness verification, and business rule verification on the mapping result, so that if there is even one unsuitable item, the entire data can be blocked.

[0162] When an error is detected, the data is separated and isolated from normal data, and it becomes possible to record the reason for verification failure along with the error code.

[0163] The autonomous correction module (500) can provide recommended correction values ​​or perform automatic remapping for data where errors occurred during the verification stage, thereby enabling the ERP input success rate to be improved.

[0164] Simple typos and unit conversions are automatically corrected, complex errors are reflected based on user feedback, and all correction history is recorded in a log, making it possible to utilize it for future re-verification.

[0165] Finally, the ERP integration module (600) controls that only data that has passed both consistency verification and autonomous correction is entered into the ERP system, and during the input process, an idempotent key and duplicate blocking logic are applied so that the same data is not duplicated.

[0166] In addition, the processing log records a decision log containing the original text, preprocessing results, mapping candidate list, final confirmed value, and error and correction history, making it possible to trace the entire process.

[0167] Therefore, this technology automates and systematically manages all stages from the collection of unstructured data to preprocessing, classification, mapping, verification, correction, and ERP input, thereby making it possible to dramatically improve the reliability and integrity of ERP-linked data.

[0169] The above-described embodiments are merely exemplary, and various other embodiments modified therefrom are possible for those skilled in the art.

[0170] Therefore, the true technical scope of protection of the present invention should include not only the above embodiments but also other embodiments that are variously modified according to the technical concept of the invention described in the following claims. Explanation of the symbols

[0171] 100 : Preprocessing module 200 : Intent Classification Module 300 : Mapping module 400 : Consistency Verification Module 500 : Autonomous Correction Module 600 : ERP Integration Module

Claims

Claim 1 In an integrated system for performing mapping from unstructured data to ERP standard data, the integrated system comprises: a preprocessing module in which unstructured data is collected and normalized by performing preprocessing including dictionary-based correction, rule-based normalization, stop word removal, and pattern recognition to enable consistency verification and automatic remapping of the unstructured data; an intent classification module in which the preprocessed data is classified according to business intents such as purchasing, sales, and production; a mapping module in which the classified data is mapped to a standard column set for each ERP module, and the optimal item code is determined by calculating a ranking of candidate items based on context-based weights; a consistency verification module in which the consistency of the results of the mapping module is verified and the corresponding data is excluded from ERP input if an error is detected; an autonomous correction module in which user correction or automatic remapping is performed by providing a recommended correction value when an error is detected; and a control module that ensures that only data passing the consistency verification module and the autonomous correction module is input into the ERP system, applies idempotent keys and duplicate blocking logic during the input process, and records all processing logs to ensure traceability Secured ERP integration module;This is included, wherein the intent classification module is configured to determine subsequent mapping and consistency verification criteria by identifying one of the business areas of purchasing, sales, and production through primary classification of the preprocessed data, and specifying an ERP module or transaction type corresponding to the identified business area through secondary classification; the consistency verification module is configured to perform consistency verification stepwise according to a plurality of different consistency verification criteria including format verification, value range verification, and verification of standard information consistency, and to block the corresponding data (or data set) from being an ERP input target if any of the plurality of consistency verification criteria is not satisfied; the autonomous correction module is configured to generate at least one correction candidate for items where errors are detected based on the verification results of the consistency verification module, and to perform automatic remapping according to the selected correction candidate; the consistency verification module is configured to perform consistency verification again according to the plurality of consistency verification criteria on the data remapped by the autonomous correction module; and the ERP integration module is configured to... An integrated system for unstructured data standardization mapping consistency verification and autonomous correction, configured to control the automatic input of the data into an ERP system only when the consistency verification result satisfies all of the aforementioned multiple consistency verification criteria, and to prevent duplicate transaction confirmations for the same data by assigning an idempotent key to the confirmed data, thereby forming a closed-loop data confirmation control structure in which data blocking, automatic remapping based on correction candidates, re-verification, and permission to input into the ERP system are continuously controlled starting from the consistency verification result, and the uniqueness of transaction confirmation is guaranteed. Claim 2 delete Claim 3 An integrated system according to claim 1, characterized in that the consistency verification module sequentially performs three stages: column-level format verification, verification of the completeness of mapped values, and business rule-based conformity verification, and if any one is unsuitable, the entire amount of the said data is blocked from ERP integration. Claim 4 delete Claim 5 An integrated system for unstructured data standardization mapping consistency verification and autonomous correction, characterized in that, in claim 1, the processing log recorded by the ERP integration module includes a decision log, and the decision log includes the original text of the input data, the preprocessing result, the mapping candidate list, the final confirmed value, and the error and correction history.

Images