KR-20260062513-A - Strategic goods classification support system and method thereof

KR20260062513AKR 20260062513 AKR20260062513 AKR 20260062513AKR-20260062513-A

Abstract

The present invention relates to a strategic goods determination support system and a method thereof. The present invention relates to a technology that supports preliminary preparation for subsequent tasks related to determination by utilizing existing strategic goods determination data to predict the determination results of new applications and searching for and providing similar cases for new applications, and supports the decision-making of reviewers for institutions performing self-determinations.

Inventors

류문영
원병희
김현조

Assignees

한국원자력연구원

Dates

Publication Date: 20260507
Application Date: 20241029

Claims (12)

A preprocessing unit that receives text-based technical data for which strategic material determination is desired and performs preprocessing; A first analysis processing unit that inputs the above-mentioned technical data preprocessed into a stored LLM (Large Language Model)-based learning model and outputs a predicted strategic material determination result; A second analysis processing unit that uses a stored similarity algorithm to perform a similarity analysis between a stored dataset related to past strategic material determination and the preprocessed technical data, and extracts past strategic material determination related data of a preset ranking or higher; A post-processing unit that utilizes LLM to perform a similarity analysis between extracted past strategic material determination-related data and the preprocessed technical data, and readjusts the similarity ranking of the extracted past strategic material determination-related data; and An additional processing unit that generates past strategic material judgment-related data corresponding to the highest similarity ranking up to a preset similarity ranking, based on the output result of the first analysis processing unit and the similarity ranking readjusted by the post-processing unit, as judgment support information for the technical data; Strategic material determination support system including
In Article 1, The above preprocessing unit A strategic material determination support system that performs data rearrangement, morphological separation, and size adjustment on the above-mentioned technical data received as input.
In Article 1, The above-mentioned dataset related to past strategic material determination is A strategic goods determination support system comprising text-based technical data from past strategic goods determinations, determination history data for the corresponding technical data, and strategic goods list data notified by law.
In Paragraph 3, The above strategic material determination support system is Prior to this, a learning preprocessing unit that performs preprocessing on the above-mentioned past strategic material determination related data set; and A learning processing unit that generates an LLM-based learning model by performing learning processing on the preprocessed data set related to the determination of past strategic materials using a stored LLM; A strategic materials determination support system that further includes
In Paragraph 3, The above second analysis processing unit A strategic material determination support system that performs preprocessing on a stored dataset related to past strategic material determination and performs similarity analysis of the preprocessed dataset related to past strategic material determination based on the preprocessed technical data.
In Paragraph 5, The above post-processing unit Using LLM, vector embedding is performed on the preprocessed technical data and the preprocessed data set related to the past strategic material determination, and A strategic material determination support system that performs a similarity analysis between extracted past strategic material determination-related data and preprocessed data by applying the results of vector embedding.
A strategic material determination support method by a strategic material determination support system in which each step is performed by a computational processing means, A preprocessing step (S100) in which text-based technical data for which strategic material determination is desired is received in the preprocessing unit and preprocessing is performed; In the first analysis processing unit, the technical data preprocessed by the preprocessing step (S100) is input into a stored LLM (Large Language Model) based learning model, and a prediction processing step (S200) is output to receive a predicted strategic material determination result; In the second analysis processing unit, a first similarity analysis step (S300) performs a similarity analysis between a previously stored past strategic material determination related data set and the technical data preprocessed by the preprocessing step (S100) using a previously stored similarity algorithm; In the second analysis processing unit, a ranking extraction step (S400) for extracting past strategic material determination-related data of a preset ranking or higher according to the analysis result of the first similarity analysis step (S300); A second similarity analysis step (S500) in the post-processing unit, utilizing LLM to perform a similarity analysis between the past strategic material determination data extracted by the ranking extraction step (S400) and the technical data processed by the pre-processing step (S100); In the post-processing unit, a ranking readjustment step (S600) for readjusting the similarity ranking of past strategic material determination-related data extracted by the ranking extraction step (S400) according to the analysis result of the second similarity analysis step (S500); and In an additional processing unit, a judgment support step (S700) generates past strategic material judgment-related data corresponding from the highest similarity ranking to a preset similarity ranking as judgment support information for the technical data, based on the output result from the prediction processing step (S200) and the similarity ranking readjusted by the ranking readjustment step (S600); A method for supporting strategic material determination, including
In Article 7, The above preprocessing step (S100) is A method for supporting strategic material determination, which performs data rearrangement, morphological separation, and size adjustment on the above-mentioned technical data received as input.
In Article 1, The above-mentioned dataset related to past strategic material determination is A method for supporting strategic goods determination, comprising text-based technical data on which strategic goods determinations were previously performed, determination history data for the corresponding technical data, and strategic goods list data notified by law.
In Article 9, The above strategic material determination support method is Before performing the above prediction processing step (S200), To save an LLM-based learning model, a learning preprocessing step (S10) in which preprocessing is performed in advance on the data set related to the past strategic material determination in the learning preprocessing unit; and In the learning processing unit, a learning processing step (S20) is performed on the past strategic material determination related data set preprocessed by the learning preprocessing step (S10) using the previously stored LLM to generate the LLM-based learning model; A method for supporting strategic material determination, further including
In Article 9, The above first similarity analysis step (S300) is A strategic material determination support method that performs preprocessing on a stored dataset related to past strategic material determination and performs similarity analysis of the preprocessed dataset related to past strategic material determination based on the preprocessed technical data.
In Paragraph 11, The above second similarity analysis step (S500) is Using LLM, vector embedding is performed on the preprocessed technical data and the preprocessed data set related to the past strategic material determination, and A strategic material determination support method that performs a similarity analysis between extracted past strategic material determination-related data and preprocessed data by applying the results of vector embedding.

Description

Strategic goods classification support system and method thereof The present invention relates to a strategic goods determination support system and method, and more specifically, to a strategic goods determination support system and method that can support preliminary preparation for subsequent tasks related to determination by utilizing existing strategic goods determination data to predict the determination results of new applications and searching for and providing similar cases for new applications, and can support the decision-making of reviewers for institutions performing self-determinations. Under international export control regimes and domestic laws (such as the Foreign Trade Act, the Atomic Energy Safety Act, and the Defense Industry Act), any person who intends to export goods or technology to a foreigner or a foreign company must apply to the government for an expert determination to determine whether they are Strategic Goods, and if the result of the application indicates that they are Strategic Goods, they must obtain an export license. In Korea, jurisdiction is divided according to the type of strategic materials; the Korea Institute of Nuclear Safety and Security (KINSC) handles assessments and export licensing for items exclusively for nuclear power, the Defense Acquisition Program Administration (DAPA) handles military materials, and the Korea Trade Security Agency (K-SSA) under the Ministry of Trade, Industry and Energy handles all other items. The determination of strategic goods involves reviewing whether an item intended for export falls under Annexes 2 and 3 of the Strategic Goods Import and Export Notification. This process is typically carried out by expert reviewers at the determination agency, and the determination procedures and criteria follow the agency's internal regulations. Furthermore, specifications or descriptions of the goods that serve as the basis for the determination are managed as confidential materials without the consent of the exporter (the person intending to export goods or technology). For this reason, applicants (those intending to export goods or technology, businesses, users, exporters, etc.) face the problem of difficulty in continuously tracking relevant cases that were previously determined to be strategic goods and difficulty in predicting the determination results. In particular, in the nuclear power sector, securing public data is much more difficult compared to other fields due to the unique characteristics of the industry, and applicants face even greater challenges as access to the application of artificial intelligence technology is highly restricted due to reasons such as security. In this regard, Korean registered patent No. 10-0927553 ("Strategic Material Management Network and Method for Managing Strategic Material Using the Same") discloses a technology that allows a user to easily determine whether a material qualifies as a strategic material. However, its purpose is to improve the current determination procedures of judgment agencies and enhance operational accuracy, and it cannot predict the judgment results. Korean registered patent No. 10-1620631 ("System and method for searching similar technical documents related to nuclear power") searches for similar technical documents required for examination among technical documents that have completed review by utilizing the similarity of text and images in the field of nuclear technology, and may be limited in learning and detecting technologies other than those in the nuclear system. As such, a similar case program not limited to the nuclear system is required to enable operators or institutions possessing sufficient judgment history data on export items to improve their work environment by increasing the utilization of relevant data. Furthermore, if the program learns from past judgment results, identifies target information to predict judgment outcomes, and outputs past judgment cases, it can effectively reduce the administrative burden on operators. FIG. 1 is a configuration example diagram showing a strategic material determination support system according to one embodiment of the present invention. FIG. 2 is a flowchart illustrating a method for supporting the determination of strategic materials according to one embodiment of the present invention. Hereinafter, a strategic material determination support system and method according to the present invention, having the configuration as described above, will be explained in detail with reference to the attached drawings. The drawings presented below are provided as examples to ensure that the concept of the present invention is sufficiently conveyed to those skilled in the art. Accordingly, the present invention is not limited to the drawings presented below and may be embodied in other forms. In addition, throughout the specification, the same reference numerals indicate the same components. Unless otherwise defined, technical and scient