| 川部 淳志 | M, 1回目発表 | サイバネティクス・リアリティ工学 | 清川 清, 加藤 博一, 酒田 信親, 磯山 直也 |
|
title: Study on work support in large space by IPD control
abstract: It is not easy for an instructor to give an instruction that grasps the entire large work space such as a large warehouse or building construction site. There is a need for a system that can provide an easy overview by looking down at the entire space. Therefore, in this study, the binocular parallax is controlled by increasing or decreasing the eye width to manipulate the perspective of the space and present a sense that a distant object is nearby. Aiming at application to work support in real environment, the methods and the results simulated with VR will be described. language of the presentation: Japanese 発表題目: IPD制御による広い空間での作業支援に関する研究 発表概要: 現場の指示者にとって、広い倉庫やビルの建設現場などの広大な作業空間全体を把握した指示を出すことは容易ではない。このような空間において全体を俯瞰し、容易に指示を行うことのできるシステムが求められる。そこで、本研究では眼幅の増減により両眼視差を制御することで空間に対する遠近感を操作し、遠くの物体が近くにあるような感覚を提示する。実空間での作業支援への応用を目指し、VRでシミュレーションした結果を手法とともに述べる。 | |||
| 西本 真由 | M, 1回目発表 | 数理情報学 | 池田 和司, 笠原 正治, 吉本 潤一郎, 久保 孝富, 福嶋 誠 |
|
title: Mother Dog Mediates Adaptation of Young Offspring to Human Society
abstract: Dogs play important roles in human society as companion animals or working dogs, e.g., guide dogs, service dogs, rescue dogs, etc. with their high social cognitive abilities. Dogs acquire social cognitive abilities to interact with human beings in their developmental processes. However, it is still unclear whether a mother dog mediates adaptation of her young offspring to human. The purpose of this study is to verify the mediation effect of a mother dog. The participants were an owner of dogs, an unfamiliar person to the dogs, and a group of six hunting dogs: a mother, an adult dog, and four young offsprings of the mother. The human subjects were instructed to walk along a specified course. During the walk, the location of each individual was recorded by a GPS system. We conducted multiple trials of the walk and changed the members for each trial. The distances between all possible individuals were calculated in the analysis. We compared the distances between the owner and the young dogs between the condition with the mother and that with the other adult dog (control condition). As a result, we found that the distance between the owner and young dogs under mother condition was smaller than that of the control condition. This result may suggest that the mother dog mediates adaptation of young offsprings to human society. I will present future directions using ANOVA and dynamic network analysis. language of the presentation: Japanese | |||
| 東 健太 | M, 1回目発表 | サイバネティクス・リアリティ工学 | 清川 清, 中村 哲, 酒田 信親, 磯山 直也 |
|
title: Study on improvement of conversational satisfaction by recognition and expansion of facial expressions
abstract: We can obtain a lot of information visually in interpersonal communication. In particular , It has been clarified that conversational satisfaction is greatly influenced by factors such as nodding and gaze of the conversation partner. In this study, we present visual information using AR technology around the conversation partner, and investigate whether and how it affects the conversational satisfaction. language of the presentation: Japanese 発表題目:顔表情の認識と拡張による会話満足度の向上の研究 発表概要:我々は対人コミュニケーションの中で視覚的に多くの情報を入手することが可能である．中でも，対話相手の頷きや視線といった要因により，会話満足度が大きく影響されることが明らかになっている． 本研究では，対話相手の周囲にAR技術を用いて視覚情報を提示し，会話の満足度に影響を与えられるか，またどのような影響が与えられるかを調査する． | |||
| 森田 大夢 | M, 1回目発表 | ソフトウェア工学 | 松本 健一, 飯田 元, 石尾 隆, Raula G. Kula |
|
title: Measuring the Impact of Industrial Programming Training on Source Code Quality
abstract: Programming skill is crucial to the success of software development projects. In software industry, since a lack of consistency in programming skillsets may exist among engineers, those engineers should be educated about the basic knowledge and programming language(s) that a company requires for their engineers through a training course. This study aims at reporting the impact of the training course in which 23 entry-level engineers of a company took. To do so, we compare those engineers’ code quality before and after training. In a nutshell, we find that programs written after the training tend to have less amount of unnecessary code than before. In other words, the training enabled participants to write programs more concisely. language of the presentation: Japanese 発表題目: 新人研修がソースコード品質に与える影響の評価 発表概要：プログラミングスキルはソフトウェア開発において重要である．ソフトウェア開発者の育成を目的としたプログラミング研修では，プログラミングスキルの向上が期待される. 本研究の目的は，研修前後でのプログラミングスキルの変化をソフトウェア品質の観点から評価する手法を確立することである. | |||
| 前田 健登 | M, 1回目発表 | ソフトウェア設計学 | 飯田 元, 藤川 和利, 市川 昊平, 高橋 慧智 |
|
title: VM Placement Optimization in Multiplayer Cloud Gaming
abstract: Recently, cloud gaming, which is a service that allows a game to be run on a cloud server and played over a network, has attracted attention in the game industry. In cloud gaming, minimizing / fairness of screen display and operation delays is an issue. In particular, when implementing a multiplayer game in a cloud gaming environment, the fairness of all participating players is important. In this research, we try to solve this problem by optimizing the placement of VMs for game servers and game clients for multiplayer games running on a cloud server. language of the presentation: Japanese 発表題目: マルチプレイヤークラウドゲーミングにおけるVM配置最適化 発表概要: 昨今、ゲームをクラウドサーバ上で動作させ、ネットワーク越しにプレイすることを可能にするサービスであるクラウドゲーミングがゲーム産業で注目されている。クラウドゲーミングおいて、画面表示や操作の遅延の最小化/公平性というものが課題である。とりわけマルチプレイヤーゲームをクラウドゲーミングの環境で実現する場合には参加する全てのプレイヤーの公平性が重要となる。これに対し、本研究ではクラウドサーバ上で動作するマルチプレイヤーゲームのゲームサーバ及びゲームクライアントのVMの配置最適化による解決を試みる。 | |||
| 中部 仁 | 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, 諏訪 博彦, 水本 旭洋（阪大） |
|
title: A method for scheduling optimal cooking procedure for multiple recipes considering user’s skill
abstract: Recent development of Information and Communication Technology (ICT) has enabled cooking beginners to easily learn about cooking. On the other hand, there is little support for the cooking beginner to acquire cooking skills taking into account his/her strengths and/or weaknesses in each of tasks (such as washing and cutting) required for cooking. In addition, it is not easy to schedule cooking procedures (consisting of cooking tasks) for cooking multiple recipes in parallel with limited number of cooking tools. In this study, we propose a method for scheduling a cooking procedure that reflects the user's strengths and/or weakness in cooking tasks when cooking multiple recipes in parallel. Since there are multiple different cooking tasks that take different time depending on the user’s skill, In the proposed method, the time spent for cooking the recipes in the past by the cooker is used to schedule the cooking procedure. In this presentation, we formulate the problem for scheduling optimal cooking procedure and show a use case of cooking multiple recipes in parallel. language of the presentation: Japanese 発表題目: 調理者の得手不得手を考慮した複数レシピに対する最適調理手順の作成 発表概要: 近年，ICT技術の発展により調理初心者であっても調理に関する知識を得ることは容易になってきている．一方で，調理者が調理技術を身につけるために得手不得手を考慮した個人に対するサポートは行われていない．また限られた調理器具の中で，複数レシピに対する調理手順を考えることは容易ではない．そこで本研究では，複数のレシピに対するユーザの得手不得手を考慮した調理手順の作成を行う．調理作業には「洗う」「切る」といった調理者によって時間が変化する調理作業があることから，調理者が過去にその調理作業にかけた時間を調理手順作成の際に用いることで調理時間の効率的な短縮を目指す．本発表では，調理手順最適化のための問題設定を行い，本研究のユースケースを提示する． | |||
| MUTINDA FAITH WAVINYA | M, 1回目発表 | 知能コミュニケーション（ソーシャル・コンピューティング） | 中村 哲, 松本 裕治, 荒牧 英治, 若宮 翔子, 田中 宏季 |
|
title: Clinical Semantic Textual Similarity abstract: Clinical Semantic Textual Similarity (ClinicalSTS) measures the semantic similarity between clinical notes. ClinicalSTS minimizes redundancy in clinical notes, so as to improve the clinical decision making process. ClinicalSTS can also be used in various NLP applications such as text summarization, question answering, information retrieval, etc. Our fundamental strategy for this task is to explore general approaches such as deep learning and traditional machine learning which have been previously used in similar tasks. Thus, we employ neural networks and regression based models, consisting of three different approaches: 1) An ensemble model using Deep Learning and Gradient Boost, 2) An ensemble model leveraging traditional NLP features and word embeddings, and 3) A two-stage model using SVM and regression models. language of the presentation: English
| |||
| 大谷 珠有 | M, 1回目発表 | 大規模システム管理 | 笠原 正治, 安本 慶一, 笹部 昌弘 |
|
title: On Path Reachability Considering Bypass Possibilities during Evacuation abstract: When a large-scale disaster occurs, speedy and reliable evacuation is essential for evacuees. In the existing work, a geographical risk information based path selection for an automatic evacuation guiding has been proposed to achieve speedy and reliable evacuation. However, the existing scheme only focuses on the reachability of the path itself, called path reliability. In actual evacuation, an evacuee may be forced to bypass the recommended route when he/she encounters a blocked road segment. In this study, we propose a new metric of a path, "path robustness," which measures the probability that an evacuee can reach the destination not only through the path but also through possible bypasses. We first show the path robustness is exactly the same as the network reliability. In addition, we further propose path robustness with hop count limitation where the length of each candidate path is limited. Through numerical results, we show the fundamental characteristics of the path robustness. language of the presentation: Japanese | |||
| 宮本 佳奈 | M, 1回目発表 | 知能コミュニケーション | 中村 哲, 小笠原 司, 作村 諭一(BS), 田中 宏季 |
|
title: Adaptive music generation with EEG for emotion induction
abstract: Music is known to be effective for emotion induction. Emotion induction techniques are important in the treatment of mental disorders such as depression and mood disorder. Although emotion induction using music has been studied, there are individual differences in the degree of emotion induction depending on the user's favorite music and mental state. In this study, we propose a music generation system that reflects the user's current emotion to reduce individual differences. Electroencephalogram (EEG) is used to objectively evaluate user’s emotion in the system. The estimated emotion and the desired emotion are compared and the result is used to generate music. In this presentation, we show outline of proposed system and the experimental design. language of the presentation: Japanese | |||
| 二宮 大空 | M, 1回目発表 | 知能コミュニケーション（データ駆動知識処理） | 中村 哲, 松本 裕治, 宮尾 知幸(MS), 鳥澤 健太郎, 飯田 龍 |
|
title:Conditional Why-Question Answering with Background Knowledge
abstract: Conditional Why-Question Answering (why-QA) is a task to retrieve answers from a given text archive for a conditional why-question. A conditional why-question (a special type of why-questions) asks causes or reasons of certain events or phenomena under a “specific condition”. For example, the following why-question “Q: Why does the sea level rises as global warming progresses?” asks the reason why “the sea level rises” under the condition that “global warming progresses”. The main issue in this task is to consider the causality chains among events such as “A: global warming progresses” ⇒ “B: Antarctic ice melts” ⇒ “C: the sea level rises”, where B would correspond to the answer to the above question “Why does the sea level rise as global warming progresses?”. However, the previous works seldom consider such causality chains and simply regard all the events expressed in a given question are related to the effect part of causality. This may lead to select as answers some wrong answer candidates irrelevant to the condition such as “global warming progresses” in the given question such as “Because a big ship pass by.” The main idea of this work is to leverage as background knowledge such binary patterns「“global warming” leads to “Antarctic ice melts”」 and 「“the sea levels” rises as “Antarctic ice melts”」for bridging the causality chain between question and answer in conditional why-QA. We assume that the binary patterns related to the condition is informative to the conditional why-question, and use the binary patterns as tools to narrow down many answer candidates. We work on the task in the following two steps: (1) extracting answer candidates to a given why-question, where each answer candidate is a passage consisting of several sentences and (2) classifying whether each answer candidate describes a proper answer to the given question. References Why-question answering using intra- and inter-sentential causal relations, Jong-Hoon Oh, Kentaro Torisawa, Chikara Hashimoto, Motoki Sano, Stijn De Saeger, and Kiyonori Ohtake. In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1733–1743, 2013 Open Domain Why-Question Answering with Adversarial Learning to Encode Answer Texts, Jong-Hoon Oh, Kazuma Kadowaki, Julien Kloetzer, Ryu Iida and Kentaro Torisawa, In the Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (ACL 2019), pp.4227--4237, Florence, Italy, July, 2019 Multi-column Convolutional Neural Networks with Causality-Attention for Why-Question Answering, Jong-Hoon Oh, Kentaro Torisawa, Canasai Kruengkrai, Ryu Iida and Julien Kloetzer, In the Proceedings of The tenth ACM International Conference on Web Search and Data Mining (WSDM2017), pp.415-424, Cambridge, UK, February, 2017. A Semi-Supervised Learning Approach to Why-Question Answering, Jong-Hoon Oh, Kentaro Torisawa, Chikara Hashimoto, Ryu Iida, Masahiro Tanaka and Julien Kloetzer, In the Proceedings of the 30th AAAI Conference on Artificial Intelligence (AAAI-16), pp.3022--3029, Phoenix, USA, February, 2016. language of the presentation: Japanese | |||
| 渡辺 佳宏 | M, 1回目発表 | 光メディアインタフェース | 向川 康博, 清川 清, 舩冨 卓哉, 田中 賢一郎, 久保 尋之 |
|
title: Rain Removal using Synthetic Aperture Method
abstract: Autonomous driving technology is greatly affected by poor visibility due to raindrops, rain streaks and water layer. Using camera with a large aperture such as F1.4, raindrops and rain streaks attached to windshield are removable by blurring them greatly. However, camera with a large aperture has limited aperture size due to hardware performance, thus a large subject such as water layer is impossible to remove. In this study, we propose synthetic aperture method that virtually generates larger lens from a multi-viewpoint image photographed by camera with a small lens. The effectiveness of our proposed method is evaluated in experimental environment used actual raindrops, simulated rain streaks and water layer. language of the presentation: Japanese | |||
| 中野 佑哉 | M, 1回目発表 | 知能コミュニケーション | 中村 哲, 松本 裕治, 宮尾 知幸(MS), 吉野 幸一郎 |
|
title: Multi-hop Reasoning for Question and Answering System
abstract: Reading Comprehension, or the ability to read texts and answer questions, is a challenging task for machines, requiring understanding of natural language. Question and answering tasks are a thriving topic that is attracting increasing attentions from researchers. In actual situations with using QA systems, there are some problems to treat huge amount of data for answering questions. For example, when a QA system is about to answer a question regarding how to use amachine’s multiple switches to help it work properly, usually it might be confused to answer it because the system has to seek reasonings from multiple documents representing ways how to use the switches. So, we propose Multi-hop reasoning model for this problem. Multi-hop reasoning model can refer multi document including reasonings of given qestions. In this presentation, we show the overview of our tackling QA and possibility of improving outputs quality with dialogue. language of the presentation: Japanese | |||