Hiroyuki Shindo

TL;DR: A method of data augmentation is proposed to increase data size by converting existing resources into synthesized non-standard forms using handcrafted rules and improves the performance of Japanese text normalization.

TL;DR: This paper proposes an approach that utilizes this information by jointly modeling multiple morphosyntactic tagging tasks with a multi-task learning framework and proposes a method of incorporating tag dictionary information into the authors' neural models by combining word representations with representations of the sets of possible tags.

TL;DR: The authors restores interpretability to adversarial training by restricting the directions of perturbations toward the existing words in the input embedding space, which can straightforwardly reconstruct each input with adversarial text to an actual text by considering the perturbation to be the replacement of words in sentence while maintaining or even improving the task performance.

TL;DR: A new contouring technology is developed that executes contour re-alignment based on a matching of the measured contour with the design data, which minimizes contouring errors and pattern roughness effects to the minimum and enables contouring that represents the contour across the wafer.

TL;DR: A neural network model that jointly learns distributed representations of texts and knowledge base (KB) entities to be generic with the ability to address various NLP tasks with ease is described.