중의성 문제 해결을 위한 데이터 및 모델 개발 기반 인공지능 방법론 연구

Data and AI based Model Development to Solve the Ambiguity Problem: focusing on Bulgarian and Korean application cases

In this paper, I briefly review the examples of ambiguity problem in Korean and Bulgarian, and examine methods for solving the ambiguity problem, focusing on previous studies. Resolving the ambiguity of the word stage has progressed to some extent through the dictionary work (off-line work) so far, and it is also true that the importance of understanding the meaning of the ambiguity of a word alone in the actual dialogue act process is relatively low. Therefore, solving the problem of ambiguity in the units of phrases, sentences, and texts is the central part of this thesis. Since there are quantitative and temporal limitations in the database that researchers collect and create, the method of computerizing this and the method of solving(processing) when a phrase or sentence to be identified appears in the main study from the point of view of artificial intelligence processing. The first step in the research for resolving ambiguity in Bulgarian is to continuously discover ambiguous expressions that appear in words, vocabulary, sentences, and texts, and create a data base. Theoretical research on such matters should proceed. Collection of examples (DB) at individual stages can be continuously performed off-line and on-line. Off-line, it will be individual dictionaries such as homonyms and Bulgarian dictionary, and on-line, it will be a web crawling collection of data from Bulgarian Wikipedia, Wordnet, and online portals. WordNet, an English-centered thesaurus, is not just a digitized dictionary, but is organized into a Synset in which nouns, verbs, adjectives, and adverbs are a set of analogous words. WordNet also provides antonyms, metonymy, pertainymy, homonymy, and entailment. It can also be an efficient and good way to secure additional data in Bulgarian by using an artificial intelligence automatic translation technique for WordNet, which is mainly provided in English. In particular, considering that the linguistic characteristics of Bulgarian are syntactically similar to English and the role of function words including prepositions is strong due to the analytic language structure, this method is very meaningful. To add one more thing, this method is very meaningful because English is the language resource that currently holds the most ambiguity data in the world. This paper is a theoretical focus on how to solve the ambiguity problem through data and model development. Accuracy analysis and parameter adjustment through actual data processing remain as future tasks to be attempted by introducing an engineering methodology.