인공지능 시대의 사전과 데이터 ‒ 한국역사인물사전(데이터)을 위한 탐색 ‒

Abstract

This paper concludes that the future of the artificial intelligence era dictionary is in the ‘machine readable data’, which is a collection of various things in a way that computers can understand for human beings and arranged in a certain order and commentary on each of them. In order to explore the possibility of transition to the data of dictionary studies, researcher examined the advantages and disadvantages of labeled data, parallel corpus data, and squad data released on AIHUB, an artificial intelligence integration platform. In data design and data construction in the specialized domain area, it was decided that it could inherit the research results of existing dictionary studies. And researcher searched the structure and contents of semantic data, another artificial intelligence data to overcome the limitations of data for deep learning, and decided that the results of existing dictionaries could be fully utilized in ontology design and semantic data construction. Based on this, for practical case study, data for deep learning and semantic data through ontology design were experimentally constructed on the theme of Korean historical character dictionary. First, researcher constructed and experimented with the learning data of object name identification (NER) based on the annals data of the Joseon Dynasty using the object name identification (NER) method of deep learning. The problem that object identification of object name identification (NER) through deep learning only identifies the attribute of the object and does not recognize the uniqueness of the object is confirmed. Therefore, the possibility of semantic data was examined through experimental ontology design to build semantic data as knowledge that has various uniqueness through the relationship with the object’s time, space, and characters. In conclusion, if the concept and understanding of machine-readable data can be added in traditional dictionary studies, it is considered to be competitive enough in both deep learning data and semantic data through ontology.