Earticle

초록

영어

This paper described phenomena related to the past tense in English, summarized the theories and empirical evidence to explain these phenomena. It aimed to present implications for the past tense and similar language phenomena in current deep learning models, such as ChatGPT, a modern large language model(LLM). The discussion related to the past tense in English dates back to when a special article called "The Past-Tense Debate" was published in Trends in Cognitive Sciences in 2002. As a result, seemingly simple language phenomena gained the interest of researchers in various fields such as linguistics, psychology, and computer science. Steven Pinker and Michael Ullman wrote a paper titled "The past and future of the past tense," to which James McClelland and Karalyn Patterson responded with a paper stating, "Words or Rules cannot exploit the regularity in exceptions." McClelland and Patterson published a paper titled "Rules or connections in past-tense inflections: what does the evidence rule out?" and Pinker and Ullman countered with "Combination and structure, not gradedness, is the issue." At this point, twenty years later, in 2023, ChatGPT, a neural network-based AI system, has attracted global attention. In this paper, we examine past tense theories and explore whether recent neural networks like ChatGPT resolve the past tense debate or if the issues raised in that debate are still relevant in the current era of artificial intelligence.

목차

1. 서론
2. 영어과거시제
3. 영어과거시제에 대한 이론
3.1 전통 생성 문법에서의 과거형 설명
3.2 신경망 이론에서의 영어과거시제
3.3 이중 통로 이론(Dual-route model)
4. 영어과거시제에 단서들
4.1 언어 습득
4.2 빈도수 효과(Frequency Effect)
4.3 이중어(Doublet)와 파생 동사
4.4 뇌신경과 동사의 과거형
5. 최근의 딥러닝 모형을 통한 동사의 과거형 예측
6. 결론
References
[Abstract]

저자

• 윤태진 [ Tae-Jin Yoon | 성신여자대학교 ]

참고문헌