KR-20260066513-A - System and method for classifying technologies

Abstract

According to the present invention, a method and system for classifying technical literature by technology using a large language model are provided.

Inventors

• 변지호

Assignees

• 변지호

Dates

Publication Date

20260512

Application Date

20241104

Claims (7)

1. 1) a step of transmitting technical information and technical classification categories of specific technical literature within a population of technical literature to be classified to a generative pre-trained large language model (GPT); and 2) A method for classifying technology comprising the step of receiving one of the technology classification categories selected by the above-mentioned giant language model based on the above-mentioned technology information and recording it for the above-mentioned specific technology literature.
2. In paragraph 1, A method for classifying technology characterized by transmitting information including technical information regarding the aforementioned specific technical literature, the aforementioned technology classification category, and the aforementioned technology category selected by the aforementioned large language model through an API (Application Programming Interface) key.
3. In paragraph 2, A method for classifying technology characterized by the fact that the information exchanged with the above-mentioned large language model is in JSON (JavaScript Object Notation) format.
4. In paragraph 1, A method for classifying technology characterized by further including a repetition step of repeating the above steps 1) to 2) for two or more technical documents with a specific delay time difference.
5. In paragraph 1, A method for classifying technology characterized in that the above technology classification categories differ depending on each technology field and classification purpose.
6. In paragraph 5, The above technology classification category includes a category that can be selected when there is no corresponding technology classification for the technology information of a specific technology document, and a technology classification method including a noise document removal function.
7. A technical literature information storage unit that stores a population of technical literature to be classified generated by a technical literature database according to a search condition (query) entered by a user, and optionally, a technical classification category uniquely set according to the technical field and the purpose of technical classification; A technology classification category input unit that inputs the above technology classification category into the above technology document storage unit or classification information transmission unit; A classification information transmission unit that transmits technical information for each technical document within a population of technical documents to be classified, located in the technical document information storage unit, and the technical classification category to the technical classification execution unit; A technology classification execution unit that selects a technology category by analyzing the technology information and technology classification categories for each technology document received through the classification information transmission unit, as a generative pre-trained large language model (GPT); A classification result transmission unit that receives a technology category selected by the technology classification execution unit and transmits it to a technology classification result storage unit; and A technology classification system comprising: a technology classification result storage unit that receives and records a classification result from the above-mentioned classification result transmission unit.

Description

System and method for classifying technologies The present invention relates to a method and system for automatically classifying technical literature according to pre-set technical classification categories based on the technical field and the purpose of technical classification. With the recent rapid advancement of artificial intelligence technology, attempts are being made to introduce AI technology into technology classification tasks. For example, Korean Patent Publication No. 10-2022-0011522 discloses a technology classification technology using artificial intelligence, wherein a pre-configured clustering algorithm is used to cluster and classify multiple documents into multiple clusters for a population to be classified, multiple sample data representing each cluster are selected from the documents included in each cluster, and after a user inputs technology classification information into the selected sample data, the AI performs learning based on that technology classification information and performs user-defined technology classification on other documents in the population based on the results of the learning. However, this technology has a limitation in that it cannot reflect the user's intention to classify, as it classifies technical documents according to the characteristics of the clusters classified by the clustering algorithm rather than classifying documents according to the technology classification desired by the user. FIG. 1 shows an example of a technical classification system configuration according to the purpose of the present invention. FIG. 2 shows another example of a technical classification system configuration according to the purpose of the present invention. A method for classifying technology according to the first objective of the present invention 1) a step of transmitting technical information and technical classification categories of specific technical literature within a population of technical literature to be classified to a generative pre-trained large language model (GPT); and 2) A step of receiving one of the technology classification categories selected by the above-mentioned giant language model based on the above-mentioned technology information and recording it for the above-mentioned specific technology literature; is included. The aforementioned population of technical literature subject to classification can be generated by entering a query into a technical literature database (DB). The query may utilize unique search terms and operators depending on the type of DB, the type of technical literature, etc. The above population can be created in the form of a spreadsheet or a table. The above technology classification categories may be appropriately determined by the classifier according to the technology field and the purpose of technology classification, and can typically be created as a concept similar to a technology tree. For example, the subcategories of ‘energy storage devices’ may be divided into ‘primary batteries’ and ‘secondary batteries,’ and the subcategories of ‘secondary batteries’ may be divided into ‘lead-acid batteries,’ ‘LIB batteries,’ ‘lithium-sulfur batteries,’ etc. The aforementioned Generative Pre-trained Transformer (GPT) is a widely known concept in the field of artificial intelligence, and examples such as Google's Gemini and Bard, Microsoft's Copilot, Anthropic's Claude, Naver's HyperCLOVA X, and Meta AI's LLaMa are commercially available. The technical information used by the above-mentioned large language model for the selection of technical classification may be, for example, the title, abstract, and conclusion of a technical paper, or the title, abstract, drawings, and claims of a patent. In the method for classifying the above technology, information including technical information regarding the specific technical literature, the technical classification category, and the technical category selected by the large language model can be transmitted (sent and received) through an API (Application Programming Interface) key. The information exchanged with the above-mentioned large language model may be in JSON (JavaScript Object Notation) format. When repeating steps 1) and 2) above for two or more technical documents, it may be desirable to repeat with a specific delay time difference between each repetition cycle. In the above method of classifying technology, a function to remove noise literature may be added by including a category that can be selected when there is no technology classification corresponding to the technology information of a specific technology document in the technology classification category. The classification system of technology according to the second objective of the present invention A technical literature information storage unit that stores a population of technical literature to be classified generated by a technical literature database according to a search condition (query) entered by a user, a