Machel Reid (Google Research)
[Webサイト]He is a research scientist in the Brain Team at Google Research working on NLP research, with a focus on multilingual NLP. He also will be a PhD student at the University of Washington co-advised by Luke Zettlemoyer and Noah A. Smith. He was previously a visiting student at Carnegie Mellon University, advised by Graham Neubig and researcher at the University of Tokyo at Matsuo Lab with Yutaka Matsuo.
概要
In this talk, I will present DiffusER, a new text generation model that is based on denoising diffusion models. DiffusER is able to perform edit-based generation, allowing it to revise existing text, a capability that current models lack. In addition to being a strong generative model on its own, DiffusER can also perform other types of generation, such as allowing a user to condition generation on a prototype or incomplete sequence and revise based on previous edits. In comparison to existing models, DiffusER performs well on a range of tasks including machine translation, summarization, and style transfer.
※トークは英語 (QAは日本語)です。