발간년도 : [2022]
| 논문정보 |
|
| 논문명(한글) |
[Vol.17, No.5] Design and Implementation of an Interactive Search System Based on KoBERT |
|
| 논문투고자 |
Jong-Won Lee, Woo Seung Jo, Tae-Hyun Kim |
|
| 논문내용 |
Most of the 210,000 users using NTIS search for R&D information through NTIS integrated search and interactive search services. NTIS Integrated Search uses a word-based search function to morphemically analyze words or sentences entered by users. And it shows all R&D information including the derived words. As the number of input words increases, it becomes difficult to provide the search result desired by the user. To solve this problem, this paper proposes an interactive search system based on artificial intelligence technology. The proposed system utilizes the KoBERT model, a pre-learning language model, to provide more accurate responses than the existing interactive search services. The system analyzes the intention of a user query and provides two search functions when determined as a search intention. If there is one input word, a word-based search is performed using a search engine. If there are more than one input word, a title with high similarity is found in the task information through the pre-learning language model and the similarity comparison function, and R&D information above the threshold is shown. In addition, a step-by-step search using a search engine will be conducted together. The proposed system is expected to be an efficient search system that can help users obtain the desired search results. |
|
| 첨부논문 |
|
|
|
|
|
자료실 > 논문지 
