Yuki Yasuda

Yuki Yasuda received his B. H. and M. H. degrees in Human Science from Waseda University in 2015 and 2017, respectively. After his graduation, he joined NHK (Japan Broadcasting Corporation) and has been working at the Science and Technology Research Laboratories since 2019. He is in charge of developing technology to support production work by applying natural language processing. Especially, He is working on multi-label classification task that target news articles. He has developed automatic labelling system with novel loss function combining various weights to suppress label imbalance. His system has been introduced into production process and provide accurate metadata to large amounts of text in less time than is required for conventional manual metadata addition. Applying this technology to the production process could promote various richer broadcasting programs and online contents at lower cost. In addition, He tackled the task to extract valuable information from social networking site under the outbreak of COVID-19. He has developed the neural network model using graph attention networks to encode syntactic dependencies and word positions in the sentence, and loss function based on connectionist temporal classification. His paper on the technology had been presented at Workshop on Noisy User-generated Text 2020(WNUT 2020) that is one of Empirical Methods in Natural Language Processing (EMNLP)’s workshop and had been added to the list of Awesome Papers Using DGL. He is a member of The Association for Natural Language Processing (ANLP).