No abstract

Understanding voluminous historical records provides clues on the past in various aspects, such as social and political issues and even natural science facts. However, it is generally difficult to fully utilize the historical records, since most of the documents are not written in a modern language and part of the contents are damaged over time. As a result, restoring the damaged or unrecognizable parts as well as translating the records into modern languages are crucial tasks. In response, we present a multi-task learning approach to restore and translate historical documents based on a selfattention mechanism, specifically utilizing two Korean historical records, ones of the most voluminous historical records in the world. Experimental results show that our approach significantly improves the accuracy of the translation task than baselines without multi-task learning. In addition, we present an in-depth exploratory analysis on our translated results via topic modeling, uncovering several significant historical events.

Playtesting is a lifecycle phase in game development wherein the completeness and smooth progress of planned content are verified before release of a new game. Although studies on playtesting in Match 3 games have attempted to utilize Monte Carlo tree search (MCTS) and convolutional neural networks (CNNs), the applicability of these methods are limited because the associated training is timeconsuming and data collection is difficult. To address this problem, game playtesting was performed via learning based on strategic play in Match 3 games. Five strategic plays were defined in the Match 3 game under consideration and game playtesting was performed for each situation via reinforcement learning. The proposed agent performed within a 5% margin of human performance on the most complex mission in the experiment. We demonstrate that it is possible for the level designer to measure the difficulty of the level via playtesting various missions. This study also provides level testing standards for several types of missions in Match 3 games.

In the field of structural-health monitoring, vibration-based structural damage detection techniques have been practically implemented in recent decades for structural condition assessment. With the development of deep-learning networks that make automatic feature extraction and high classification accuracy possible, deep-learning-based structural damage detection has been gaining significant attention. The deep-learning neural networks come with fixed input and output size, and input data must be downsampled or cropped to the predetermined input size of the networks to obtain desired output of the network. However, the length of input data (i.e., sensing data) is associated with the excitation quality of a structure, adjusting the size of the input data while maintaining the excitation quality is critical to ensure high accuracy of the deep-learning-based structural damage detection. To address this issue, natural-excitation-technique-based data normalization and the use of 1-D convolutional neural networks for automated structural damage detection are presented. The presented approach converts input data to predetermined size using cross-correlation and uses convolutional network to extract damage-sensitive feature for automated structural damage identification. Numerical simulations were conducted on a simply supported beam model excited by random and traffic loadings, and the performance was validated under various scenarios. The proposed method successfully detected the location of damage on a beam under random and traffic loadings with accuracies of 99.90% and 99.20%, respectively.

Historical documents refer to records or books that provide textual information about the thoughts and consciousness of past civilisations, and therefore, they have historical significance. These documents are used as key sources for historical studies as they provide information over several historical periods. Many studies have analysed various historical documents using deep learning; however, studies that employ changes in information over time are lacking. In this study, we propose a deep-learning approach using improved dynamic word embedding to determine the characteristics of 27 kings mentioned in the Annals of the Joseon Dynasty, which contains a record of 500 years. The characteristics of words for each king were quantitated based on dynamic word embedding; further, this information was applied to named entity recognition and neural machine translation.In experiments, we confirmed that the method we proposed showed better performance than other methods. In the named entity recognition task, the F1-score was 0.68; in the neural machine translation task, the BLEU4 score was 0.34. We demonstrated that this approach can be used to extract information about diplomatic relationships with neighbouring countries and the economic conditions of the Joseon Dynasty.