[행사/세미나] How Good Are GPT Models at Machine Translation?; Why?(23.4.19. 13:00)
- 인공지능융합학과(일반대학원)
- 조회수2308
- 2023-04-05
안녕하세요, 대학원 인공지능융합학과 사무실입니다.
아래와 같이 세미나 안내드립니다.
Title: How Good Are GPT Models at Machine Translation?; Why?
Speaker: Dr. Young Jin Kim @ Microsoft
Time: 2023 April 19th 13:30 ~ 15:00
Location: Hybrid
- In-person: 26310
- Online: https://us02web.zoom.us/j/88081137117?pwd=OTJaN25qbWQyTCszYUEvandyRXRFZz09
- (Passcode: 0419)
Language: English speech & English slides
Abstract:
Generative Pre-trained Transformer (GPT) models are known for their exceptional natural language generation abilities, but their machine translation performance has not been thoroughly examined. In this presentation, I will provide a comprehensive evaluation of GPT models for machine translation and discuss interesting phenomena observed in GPT-based translations. I will also discuss how GPT models have acquired machine translation capabilities without explicit supervision.
The presented evaluation encompasses various aspects, including the quality of different GPT models compared to state-of-the-art research and commercial systems and the effect of prompting strategies. We have conducted experiments involving 18 translation directions, comprising high and low resource languages, as well as non-English-centric translations, to evaluate the performance of three GPT models: ChatGPT, GPT3.5 (text-davinci-003), and text-davinci-002.
The results indicate that GPT models achieve highly competitive translation quality for high resource languages, while their capabilities for low resource languages are limited. I will also demonstrate how hybrid approaches, combining GPT models with other translation systems, can further enhance translation quality. Moreover, I will share a comprehensive analysis and human evaluation to better understand the characteristics of GPT translations, thus helping to identify the potential and limitations of GPT models for translation.
Bio:
Young Jin Kim is a Principal Researcher at Microsoft where he develops machine learning models with state-of-the-art techniques. His recent research focus includes designing efficient and effective algorithms and model architectures for large scale language models. Young received his Ph.D. from Georgia Institute of Technology for his research in deep learning and high-performance computing.
감사합니다.
대학원 인공지능융합학과 사무실 드림