ゼミナール発表

日時: 09月27日(水)2限(11:00-12:30)


会場: L1

司会: 武富 貴史
阪口 栄穂 1651051: M, 2回目発表 環境知能学 萩田 紀博, 加藤 博一, 神原 誠之
title:Driver’s Visual Recognition Modeling based on Target Visual Recognition Difficulties
abstract: We consider about a method to estimate completing visual recognition of driver, to draw driver’s attention properly. Most previous studies only use the condition between targets and gaze positions, and do not consider differences in targets appearances. We define difficulty in visual recognition as “the Visual Recognition Difficulties” of the target. We analyzed the relationship between "the Visual Recognition Difficulties" and gaze movement in the visual search task to propose new estimation method.
language of the presentation: *** Japanese ***
:対象の視認難易度に基づいた運転手視認モデルの構築
:自動車運転者の注意状態を把握し適切に注意を誘導するため、対象物に対する運転者の視認完了を推定する手法について検討している。従来の視認推定手法の多くは対象と視線の方向に関する一律の基準で視認完了を判定しており、視認対象の視覚的特性による視認性の差異を考慮していない。我々は対象の視認の難しさを視認難易度と定義し、視認難易度と物体視認時の視線挙動との関係の分析した。
 
向井田 一平 1651107: M, 2回目発表 環境知能学 萩田 紀博, 加藤 博一, 神原 誠之
title: Evaluation of TV Chatting Robot's Utterance to Improve Motivation to Interact of User
abstract: We are developing a robot that watches a TV program and chats with a user for daily communicaiton. It is important to keep up motivation to interact with a robot for daily communicaiton. In conventional research, a robot was able to speak contents improvimg user's motivation using social media comments. However, the robot selected a spoken comment at random and no criterion of comment selection to evaluate the utterance. The robot need new selection criteria based on the motivation. We evaluate the social comment to improve motivation to interact of user.
language of the presentation: Japanese
発表題目: ユーザの対話意欲向上に向けたテレビ雑談ロボットのための発話文の評価
発表概要: 我々は,対話ロボットを利用して日常的な対話を実現するために,ユーザと共にテレビを視聴して雑談をするロボットを開発している. 日常的にロボットと対話を継続するためには,ロボットと対話を続けるモチベーションである対話意欲が重要である. 従来手法では,ソーシャルメディア上から取得したソーシャルコメントを利用してユーザの対話意欲を向上させる発話を可能にした. しかし,この手法では対話意欲を向上させるソーシャルコメントを選択して発話しているわけではなく,発話文の対話意欲を評価する基準もない. ロボットがより対話意欲を向上させる発話をするためには,ユーザの対話意欲を反映した発話文の選択基準が必要である. そこで本研究では,ソーシャルコメントがユーザの対話意欲に与える影響を評価し,より対話意欲を向上させるロボットを実現する.
 

会場: L2

司会: 畑 秀明
池田 祥平 1651008: M, 2回目発表 ソフトウェア工学 松本 健一, 飯田 元, 石尾 隆, 伊原 彰紀, Raula G. Kula
title: Mining the contents of readme files in OSS GitHub projects: A case study of NPM projects
abstract: Software development projects often make the own software document called ''README'' to share how to use it among the project developers. The developers often note the install guide and usage examples in the document. Nowadays, it is a crucial document because software projects often release the software on the public web service such as GitHub. We then use the public software as open source software (OSS) with the document. However, the software document is not always supported for users according to GitHub survey published in 2017. The released documents are sometimes incomplete and old. In this presentation, we introduce our study that analyzes whether the same type of software use similar items in README file. To analyze this, we conduct an empirical study on README items for 143,239 projects in GitHub. As a result, we could identify the trend items such as install, usage, and license in the README. For my dissertation, we analyze README items for the similar type of software.
language of the presentation: Japanese
 
山田 悠貴 1651117: M, 2回目発表 ソフトウェア設計学 飯田 元, 松本 健一, 市川 昊平, 崔 恩瀞
title:Support method for method inlining using agglomeration degree
abstract:Inlining a method is a work of unfolding to a side calling an obvious method without calling it.The developer is to inline the method to reduce unnecessary call relations However, it is difficult to find candidates for inlining methods from large-scale source code. Therefore, this research proposes a method to support method inlining using cohesion metrics. First, a function pair with calling relation is extracted from the source code and filtering is done.After that, we calculate the degree of agglomeration when inline with function pair and present it to the developer based on that value.
language of the presentation:Japanese
 
吉田 貴信 1651123: M, 2回目発表 ソフトウェア設計学 飯田 元, 松本 健一, 市川 昊平, 崔 恩瀞
Title: Investigation of Bugs Contained in Design-Pattern-applied source code
abstract:Design Pattern is leveraged for solving problems that frequently appear in software design. Builder pattern is one of the design patterns, but high bug rate has been reported in this pattern by previous study. We consider that if software developers are notified with bug-connected modifications in source code which are applied with design pattern, bugs could be avoided beforehand. In this research, by investigating the change logs of source code that are applied with design pattern, we classify the types of bugs as well as the factors that lead to bugs. As an example, among the classes related to design pattern, bugs can be caused by software developers's changing one class but forgetting to change another class. For conducting the above investigation, we analyze the commit logs of an open source project. As well, we also utilize design pattern detection tool in our work.
language of the presentation:Japanese
Body:
 

会場: L3

司会: 川原 純
原 崇徳 1651089: M, 2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: Path Selection Based on Road Blockage Risks for Automatic, Speedy and Reliable Evacuation Guiding
abstract: Evacuees have to evacuate to a refuge quickly and safely under large-scale disasters. An automatic evacuation guiding scheme has been proposed, which is a reactive approach based on the disaster information collected by implicit interactions among evacuees, their mobile devices, and networks. In this scheme, an evacuation route is obtained as the shortest path selection, which will contribute speedy evacuation but may not be safe. In this presentation, we propose speedy and reliable path selection, which is a proactive approach based on the road blockage probabilities, and combine it with the automatic evacuation guiding. The proposed path selection first calculates candidates of short paths from the evacuee's current location to the refuge. Then, it selects the most reliable one from the candidates by taking into account road blockage probabilities. Through the simulation experiments, we show that the proposed scheme can improve the safety of evacuation compared with the shortest path selection regardless of communication environments, in terms of the number of encounters with blocked road segments. In addition, the proposed scheme can also improve the evacuation time compared with the shortest path selection under inferior communication environments.
language of the presentation: Japanese
 
藤井 公大 1651092: M, 2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: Risk Analysis of Encountering Road Blockage during Disaster Evacuation
abstract: From the viewpoint of disaster prevention, risk analysis of road networks is important. In this presentation, we propose a scheme to analyze the risk of encountering road blockage during disaster evacuation. The risk of encountering road blockage can be analyzed based on the road demands and road blockage probabilities. The road demands can be regarded as the betweenness centrality, which depends on evacuation path selection schemes. The road blockage probability is the probability that the corresponding road is blocked after a disaster occurs. Through the analysis using actual data obtained from Nagoya City, we show some roads have high risk of encountering road blockage even if evacuees try to select safer evacuation route by taking account of road blockage probabilities.
language of the presentation: Japanese