no code implementations • EMNLP (ArgMining) 2021 • Takahiro Kondo, Koki Washio, Katsuhiko Hayashi, Yusuke Miyao
We propose a methodology for representing the reasoning structure of arguments using Bayesian networks and predicate logic facilitated by argumentation schemes.
no code implementations • 16 Feb 2022 • Koki Washio, Yusuke Miyao
Moreover, to evaluate code generation for unknown libraries and our framework, we extend an existing dataset of open-domain code generation and resplit it so that the evaluation data consist of only examples using the libraries that do not appear in the training data.
no code implementations • ACL 2021 • Ryo Ueda, Koki Washio
Our results suggest that noise on a speaker is one of the factors for ZLA or at least causes emergent languages to approach ZLA, while noise on a listener and a channel is not.
no code implementations • IJCNLP 2019 • Koki Washio, Satoshi Sekine, Tsuneaki Kato
Definition modeling includes acquiring word embeddings from dictionary definitions and generating definitions of words.
1 code implementation • NAACL 2022 • Ikuya Yamada, Koki Washio, Hiroyuki Shindo, Yuji Matsumoto
We propose a global entity disambiguation (ED) model based on BERT.
Ranked #1 on Entity Disambiguation on MSNBC
no code implementations • NAACL 2018 • Koki Washio, Tsuneaki Kato
Our proposed model of $P(path|w_1, w_2)$ can be learned in an unsupervised manner and can generalize the co-occurrences of word pairs and dependency paths.
no code implementations • EMNLP 2018 • Koki Washio, Tsuneaki Kato
This overcomes the critical data sparseness problem encountered in previous pattern-based models.