| 小嵜 泰造 | M, 1回目発表 | ユビキタスコンピューティングシステム | 安本 慶一, 清川 清, 松田 裕貴 |
|
title: A Method of Supporting Abacus Learning with Game Elements to Overcome Difficulty in Abacus Operation
abstract: However, there are various types of abacus operations, and many students face " difficulty with abacus operations" in the process of learning. However, it is well known that overcoming these difficult bead manipulations requires repeated practice over a long period of time. Although a method to recognize the surface of an abacus board by analyzing the bird's-eye view of a commercial abacus board taken by a video camera and a method to present information necessary to support the learning of abacus on a desktop display in real time have been proposed, no system that takes into account the extraction of difficult operations and their reflection in the problems has been introduced. However, no system has been introduced that takes into account the extraction of difficult operations and their reflection in problems. In this study, I aim to establish a method to support the learning of abacus by incorporating a game element into the learning of abacus so that the student can learn repeatedly while enjoying the game. language of the presentation: Japanese | |||
| 筒井 巽水 | M, 1回目発表 | ユビキタスコンピューティングシステム | 安本 慶一, 藤川 和利, 諏訪 博彦, 松田 裕貴 |
|
title: Consideration of a method for sharing evacuation shelter data using an infrastructure-less evacuation support map in the event of a disaster
abstract: In recent years, local governments have been increasingly using web-based shelter sharing services such as Disaster Relief Map to prepare for natural disasters, but there is a problem that these services cannot be used if the communication infrastructure is cut off.This research examines the construction of an infrastructure-less evacuation support map to enable the sharing of evacuation center data even when communication infrastructure is cut off due to a natural disaster. | |||
| 渡邉 珠海 | M, 1回目発表 | サイバネティクス・リアリティ工学 | 清川 清, 安本 慶一, 内山 英昭, Perusquia Hernandez Monica, 平尾 悠太朗 |
|
title: The Effect of Real-Time Feedback During Light Exercise on Activities for Breaking Up Sedentary Posture
abstract: Prolonged sitting has emerged as a problematic behavior associated with increased health risks. To alleviate the health risks associated with prolonged sedentary behavior, frequent light exercise is considered effective. While there are existing services and studies that provide feedback before and after light exercise, the effect of providing feedback during light exercise has not been adequately explored. Therefore, this study investigates the effect of interactive real-time feedback during light exercise on the behavior of reducing sedentary posture. language of the presentation: Japanese | |||
| 渡部 宙 | M, 1回目発表 | サイバネティクス・リアリティ工学 | 清川 清, 安本 慶一, 内山 英昭, Perusquia Hernandez Monica, 平尾 悠太朗 |
| 大坪 寛武 | M, 1回目発表 | サイバネティクス・リアリティ工学 | 清川 清, 田中 沙織, 内山 英昭, Perusquia Hernandez Monica, 平尾 悠太朗 |
|
title: The perspective effect of awe in VR environments: a comparison of first-person and third-person perspectives
abstract: Awe is the feeling of wonder and fear that we experience when we encounter something amazing or beautiful in nature or art, and it has been shown to have many benefits for us. But the effect of perspective on complex emotions such as awe has not been empirically verified. Therefore, the purpose of this study is to measure the degree of awe felt when viewing different perspectives of static landscapes in VR environments, and to clarify the impact of perspective on awe. language of the presentation: Japanese 発表題目: VR環境におけるAweの視点効果: 第一人称視点と第三人称視点の比較 発表概要: Aweとは自然や芸術などにおいて、感動や恐れなどの混合した感情を指し、私たちに多くの効果をもたらすことが分かっている。 しかしAweのような複雑な感情に対する視点の効果は、実証的に検証されていない。そこで、本研究の目的は、VR環境で風景の異なる視点を見たときに感じるAweの度合いを測定し、視点がAweに及ぼす影響を明らかにすることである。 | |||
| LI KAIFAN | M, 1回目発表 | ソーシャル・コンピューティング | 荒牧 英治, 渡辺 太郎, 若宮 翔子, 矢田 竣太郎 |
|
title: Long text summarization with recurrent memory transformer
abstract: Text summarization is an important task in the field of natural language processing, a good summary can help us quickly understand the core content of the text. Past researchers have mainly focused on understanding short texts, such as news datasets like CNN-Dailymail or XSum. However, in practical application scenarios, it is very common to face the issue about handling long texts or even super-long texts, such as conference papers or medium/long novels, which cannot be handled by existing models limited by the input length. To this end, we propose a model based on RMT:Recurrent Memory Transformer, where We first sliced the long text and then force the model to process the text sequentially with memory mechanism, in the expectation of generating coherent and concise summaries in each section. language of the presentation: English | |||
| 藤田 一天 | M, 1回目発表 | 自然言語処理学 | 渡辺 太郎☆, 吉野 幸一郎（客員教授）, 河野 誠也（客員助教） |
|
title: Tailored Scheduling and Performance Management for Individuals
abstract:Background Due to the COVID-19 pandemic, more people are spending time at home, leading to issues such as lack of exercise and disrupted lifestyles. Some individuals address this by going to the gym, but doing so requires not only money but also the need to balance work and family. In fact, 42.6% of people withdraw from the gym due to family or work reasons. The background stems from the idea of whether scheduling and performance management tailored to an individual's environment and personality can be solved through image processing, an area of personal interest. Problem Wanting to perform optimal schedule creation and management without investing excessive time and effort. Solution Exploring the possibility of using tools that caption human behavior or instantly create schedules anytime to address this issue. Research Objectives Reduction of effort in schedule creation/modification Reduction of effort in performance management Creation of schedules tailored to individual personalities and preferences Past Related Research Scene Graph Generative Agent Methods to Achieve Research Objectives (Solution/Experiment) Using ChatGPT to create schedules Recording actions within the home using scene graphs and captions Current Progress Utilizing ChatGPT to create optimal schedules Inputting scene graphs and action captions into ChatGPT to describe performed actions Future Challenges Determining prompts needed to create schedules tailored to individual personalities Establishing criteria to judge whether a created schedule was adhered to or not Developing a method for comparing action caption texts with schedule content texts Deciding on prompts for ChatGPT to determine the next day's schedule based on previous day's actions Figuring out how to collect necessary language data and scene graphs for input. language of the presentation: Japanese 発表題目: 個人に合ったスケジューリングと履行管理 発表概要: * 研究動機（背景・着想に至った経緯） * 背景 * コロナ禍により、在宅で過ごすことが多くなってきている。その分、運動不足や生活習慣が乱れることもある。それを改善するためにジムに通う人もいる。だがしかし、ジムに通うためにはお金はもちろん、仕事や家庭との両立をする必要がある。事実、家庭事情や仕事の都合でジムを退会する人は42.6%も存在する。もっと個人の環境や性格に合った予定作成や履行管理を自分が興味ある画像処理で解決できないか。そう思ったのが背景です。 * 問題 * 手間と時間ををかけずに、最適な予定作成と予定管理を行いたい。 * 解決策 * 人の行動のキャプショニングや24時間いつでも、瞬時に予定を代わりに作成してくれるツールをつかって解決できないか。 * 研究目的 * 予定作成/修正の手間削減 * 履行管理の手間削減 * 個人の性格や都合に合った予定の作成 * 過去の関連研究 * シーングラフ * Generative Agent * 研究目的を達成するための方法（解決・実験） * ChatGPTを用いて、予定を作成する * シーングラフとキャプショニングを用いて、家の中の行動を記録する * 現在の進捗状況 * ChatGPTを使用して、最適な予定を作成する * シーングラフと行動キャプションをChatGPTに入力し、履行したことを記述する * 今後の課題 * 個人の性格に合わせた予定を作成するためにどんなプロンプトが必要か * 作成された予定のうち、履行したかしていないかの判断基準をどうするか * アクションのキャプション文と予定の内容の文章比較をどのように行うか * 前日の行動を基に、次の日の予定を決めるためには、どんなプロンプトをChatGPTに行うべきか * 入力に必要な言語データやシーングラフをどのように収集するか | |||