The Government of Japan has taken several measures to secure human resources for long-term care, includingimproving work efficiency by care workers, improving a working environment, and providing effective training fornew employees. However, there are major barriers in resolving the issues, such as difficulties in public surveys due toa variety of situations involving personal data in long-term care facilities, and difficulties in quantifying informationand features on care workers and care receivers in actual nursing care processes. In this research, specifically focusingon excretion care, which is regarded to be a particularly high burden on care workers, we newly construct a virtualnursing care process simulator based on a multi-agent model. At first applying it to simulate care processes in anactual day-service facility, we have confirmed that the simulated results agree with observed ones reasonably well.Next, we evaluated the work efficiency of care workers quantitatively, parametrically varying the number of elderlypeople and that of care workers. As the results, it is confirmed that the total amount of assistance by care workersincreases as the number of elderly people increases, and that the work time per care worker decreases as the numberof care workers increases. We can conclude that such a simulation-based approach will be a powerful tool to discussissues on nursing care processes quantitatively and to search for new solutions.

In the study of criminal law, the analysis of a large number of criminal cases is essential. However this taskrequires considerable human costs and time. In our research, we have created a system that automates case lawanalysis using generative AI, specifically utilizing the GPT-4. This system is designed to automatically extract factsfrom criminal cases that fulfill the requirements for establishing a crime, which are called the elements of crime.To evaluate the performance of our system, we conducted experiments with two types of crimes and asked legalexperts to create the correct answer data. As a result, over 80 percent of the extractions overlapped with the correctanswers, adaptively reflecting the characteristics of each legal expert. Furthermore, the quality of extraction wasfound acceptable from the perspectives of legal experts. On the other hand, the cases where the extracted parts aresummarized or insufficient extraction need improvement.

This study aims to create a robot capable of diverse behaviors. To achieve this, we propose a behavior generationframework for robots and evaluate whether the generated behaviors result in diverse movements. The proposedframework learns cat behaviors, which are familiar to humans, through a large language model (LLM). The datasetwas created by extracting 2D keypoints from cat videos, converting these 2D keypoints into 3D keypoints, and calculatingjoint angles using inverse kinematics based on the joint positions. The data, consisting of joint positions andangles, was then converted into a language-like format, called ”motion language.” This converted data was used totrain the gMLP. The gMLP’s loss function reached its minimum validation loss after two epochs, so the model trainedup to this point was used to generate motion language. Evaluation experiments assessed whether the generated behaviorsequences could sustain diverse behaviors over time. The evaluation method used Multivariate Multiscale Entropy.The results confirmed that the gMLP-generated data exhibited a level of complexity comparable to that of the trainingdata. Additionally, the generated data demonstrated greater complexity than random data. gMLP’s output not onlyachieved short-term complexity but also generated long-term sequences with a level of complexity unattainable byrandom behavior generation.

In this paper, we propose a novel model that combines two types of textual data, newspaper articles and stockbulletin board data, to predict the rise of the Volatility Index (VI). VI is a crucial indicator of market risk and iswidely used to forecast future price fluctuations. Often referred to as the fear index, VI is closely linked to socialconditions and investor sentiment. Previous studies have primarily focused on either mass media or social mediawhen predicting financial indicators. Mass media provides valuable information about social events that influence theeconomy and financial markets. In contrast, social media reflects the interests, opinions, and sentiments of investorsand contributors, making it an important data source. To comprehensively capture both social conditions and investorsentiment, this study utilizes newspaper articles and stock message board posts to predict VI. Specifically, featurevectors are extracted from these textual data and combined with financial time series data to build a machine learningbased prediction model. The proposed method was evaluated by comparing its prediction accuracy with baselinemodels and conducting trading simulations. The results demonstrate that the proposed model outperforms existingmethods in terms of prediction accuracy and shows the potential to generate profits.

In the BDI model, a well-known model of autonomous agents, the BDI logic is used for formalization. It extends the classical logic by introducing modal operators representing mental states and tense operators. However,real-world agents may obtain inaccurate information, and retain inconsistent beliefs. Extending the classical logic which has the explosion law cannot handle such situations well. In this paper, we propose a logic system for agents that can handle such situations, and provides a discussion of its relevance to agent development environments.

In recent years, inpainting of human face images has attracted a great deal of attention among various inpainting tasks that aim to complete missing or obscured regions in digital images. Compared to inpainting of abstract objects such as landscapes, buildings, or patterns, inpainting of human face images presents unique and considerable challenges. This is due to the fact that human faces have an complex visual structure, where even subtle differences in facial features or proportions can cause a sense of discomfort or seem unnatural. As such, faithfully completing missing regions in face images while maintaining a convincing and realistic appearance is an extremely difficult problem. Although some inpainting techniques have been developed to fill in missing facial parts based on analyzing and extrapolating from surrounding available facial information, most existing methods struggle to reproduce facially coherent results. To address this, we propose leveraging supplementary voice data, which contains cues strongly correlated to an individual’s facial structure and expressions, to guide and enhance face image inpainting. Specifically, our proposed method uses voice segments as additional conditioning inputs when generating missing facial regions, with the aim of improving fidelity and perceptual realism of completed faces. To rigorously evaluate this voiceaugmented face inpainting approach, we constructed a large test dataset consisting of around 20,000 pseudo-masked face images paired with corresponding preprocessed voice samples of each individual. In comparative experiments, our method attained significantly higher face completion quality versus a baseline model without any voice inputs. Additionally, we conducted challenging real-world verification tests using actual masked face images and raw voice data as inputs. Although performance remained insufficient for reliably handling real occluded faces, these experiments confirmed that voice conditioning clearly improves results on artificial test data, demonstrating its viability as a supplementary signal to guide generative face inpainting systems.