챗GPT 프롬프트 엔지니어링을 활용한 상식 추론결과에서 나타난 언어 자질들 분석
The Analysis on the Linguistic Features of Common Sense Inference Using Chat-GPT Prompt Engineering
- 발행
- 2023
- 소속·발행
- 이화여자대학교
- 출처
- 국내 KCI
- DOI
- 10.18855/lisoko.2023.48.3.009
- 원문등록
- 2023-09-30
- 원문
- 원문 보기 ↗
개념
키워드
ChatGPT, common sense inference, questions, formal semantics, argument structure, linguistic features, prompt engineering, ChatGPT, common sense inference, questions, formal semantics, argument structure, linguistic features, prompt engineering
초록
The purpose of this paper is how linguistic features operate prompt engineering in common sense inference using OpenAI ChatGPT. The operation of AI based on the Large Language Models is widely known as prompt engineering, which induces the common sense inferences of GPT via reinforcement learning from human feedback. Prompt engineering per se is a query to induce desired responses from the generative AIs. It is essential in creating better services and results from generative power of AI tools. In this paper, we analyze what aspect of linguistic features bring out better performance queries for the common sense inference. We deeply look into the features based on the theoretical linguistics.