| 津曲 優子 | 1561016: D, 中間発表 | 計算システムズ生物学 | 金谷 重彦,佐藤 嘉伸,MD.ALTAF-UL-AMIN,佐藤 哲大 |
| 白石 磨貴男 | 1561012: D, 中間発表 | 計算システムズ生物学 | 金谷 重彦,安本 慶一,MD.ALTAF-UL-AMIN,小野 直亮 |
| 豊嶋 章宏 | 1651078: M, 1回目発表 | 知能コミュニケーション | 中村 哲 |
|
title: Dialogue Management for Non-task-oriented Dialogue Systems by Considering User Engagements
abstract: Non-task-oriented dialogue systems are used to make a relationship with users. In non-task-oriented dialogue systems, engagements of users, motivations of users to talk with the system, is important to continue the conversation. We discuss two important factors to construct a non-task-oriented dialogue system that attracts user interests. The first one is a topic control, to make a natural conversation. The second one is considering the user interest to the current conversation content. language of the presentation: Japanese | |||
| 寺澤 直人 | 1651074: M, 1回目発表 | 知能コミュニケーション | 中村 哲 |
|
title: Detection of temporal liking during watching videos using short-term EEG data
abstract: EEG-based emotion recognition have been frequently researched because conventional feature, facial expression and prosody may not be able to utilized. However, in a case of emotion recognition during watching videos, EEG and subjective evaluation values are divided into each scene and detailed information in the scene is lost. Our aim is to detect temporal liking in the scene with relatively short-term EEG data. In this study, a method that EEG data and subjective evaluation are analyzed by not dividing the scene, is proposed. language of the presentation: Japanese | |||
| 沖本 雄大 | 1651025: M, 1回目発表 | ネットワークシステム学 | 岡田 実 |
|
title: Proposal of beamforming method using leaky coaxial cable
abstract: Recently, an increase in communications traffic due to popularization of smartphones and IoT devices is remarkable. In particular, the lowering of communication efficiency in areas with high population density such as underground shopping areas has become a problem. We aim to solve the problem by beamforming method using a leaky coaxial cable. language of the presentation:Japanese | |||
| 河中 祥吾 | 1651037: 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 |
|
title: A Method for Detecting Approaching Vehicle with Smartphone to Support Elderly People's Bicycle Operation
abstract: More than 50% of the number of fatal accidents during bicycle riding are occupied by elderly people aged 65 and older. As a cause, their physical ability and cognitive ability are decreased due to aging, and detection and avoidance of danger such as approach of a vehicle is delayed. Therefore, in order to reduce the number of fatal accidents during bicycle riding for elderly people, it is important to support cognitive abilities by tools that detect approaching cars and notify them to bicycle users. In this study, we propose a system to detect the change pattern of the frequency component of the unique sound which occurs when approaching the car from sound data recorded by microphone of smartphones. Then, we developed a proximity vehicle detection algorithm by collecting feature quantities and carrying out machine learning. Finally, the developed vehicle detection model was evaluated by 10-fold cross validation, and average F-value of 97.4 [%] was achieved. language of the presentation: Japanese | |||
| 鈴木 大介 | 1651064: M, 1回目発表 | 光メディアインタフェース | 向川 康博 |
|
title: Estimating photosynthetic activity of plants by measuring leaves
abstract: Crop plants is evaluated by measuring nutrition after harvest. We assume to photosynthetic activity at leaf affect nutrition of crop plants. If photosynthetic activity can be measured, that can be to contribute to grow high quality plants. In this study, we aim to estimate photosynthetic activity of plants by measuring leaves. In this presentation, we report on the results that we measured leaf photosynthesis and analyzed about photosynthetic activity. language of the presentation: Japanese | |||
| 佐藤 祐紀 | 1651054: M, 1回目発表 | 光メディアインタフェース | 向川 康博 |
|
title:Making of the optical replica for purpose of the appearance of the Raden
abstract:Recently, the digital archives of many cultural assets are widely present to preserve Japanese local custom and culture. However, the displaying archived data has a limitation to reproduce the appearance especially for Raden (mother-of pearl inlay work) because Raden is difference view performance by viewpoint. In this research, we produce the optical replica of the Raden to reproduce the appearance. Our method can control colors of varying viewpoint position with lenticular lens. We confirm the efficiency of our method experimentally. language of the presentation:Japanese | |||
| 香山 侑槻 | 1651034: M, 1回目発表 | 大規模システム管理 | 笠原 正治 |
|
title: On Impact of Peer Departure on Average File Retrieving Time in Tit-for-Tat-based P2P File Distribution
abstract: Peer-to-Peer (P2P) file distribution systems are effective for disseminating massive files from a server to many users. However, the increase of peers without uploading pieces, called free riders, deteriorates the systems' performance. To tackle this problem, P2P file distribution systems can adopt Tit-for-Tat (TFT) strategy in game theory, where each peer has to upload pieces to others so as to retrieve its demanding pieces from them. Recently, the optimal piece flow in TFT-based P2P file distribution has been analyzed under static environments where peers sojourn in the system before finishing file retrieval. Since peers may leave the system during file retrieval in actual systems, we reveal how such peers' departures affect the performance of TFT-based P2P file distribution. We first model the determination of optimal piece flow as an integer linear programming (ILP) and obtain the optimal solution by using existing solver, i.e., CPLEX. Through the analysis of the optimal piece flow, we show that the larger the upload capacity of the peer, the worse the performance will be. Furthermore, if the upload capacity of the left peer is not large, the slower the withdrawal time is, the better the performance will be. language of the presentation: Japanese | |||
| 児矢野 晋太 | 1651049: M, 1回目発表 | 数理情報学 | 池田 和司 |
|
title: Communication Network in Internet Stock Forum and Its Application to Stock Return Prediction
abstract: Stock return prediction have been studied for a long time and many predictive methods have been developed. Generally speaking, however, the performance of those methods is limited and unstable because of a lot of external factors. To overcome the difficulties, online social media have been used recently. Previous studies found that the amount of bullish opinions in an Internet stock forum was significantly correlated with stock return. In this research, we analyze the network of messages contributed on the forum to evaluate the influence of bullish or bearish opinions of individual users on the other users, and examine the relationship between the overall opinions and stock return. As metrics to calculate the influence of users, we use major network centrality measurements and engagement capacity measurement. language of the presentation: Japanese | |||
| 佐藤 元紀 | 1651053: M, 1回目発表 | 自然言語処理学 | 松本 裕治 |
|
title: Sequential Labeling Task with Lattice
abstract: In natural language processing, some tasks are solved as Sequential Labeling Task. We use Lattice and recent neural network technique for Named Entity Recognition task (NER). Our proposed method has two step. First, training Bi-LSTM-CNN model for BIO-tagging in NER task. Second, define threshold value and build Lattice from Bi-LSTM-CNN's output layer softmax distribution, and search gold path in Lattice. We overcome F-score from first step Bi-LSTM-CNN score. language of the presentation: Japanese | |||
| 小比田 涼介 | 1651047: M, 1回目発表 | 自然言語処理学 | 松本 裕治 |
|
title: Multilingual Back-and-Forth Conversion between Content and Function Head for Easy Dependency Parsing
abstract: Universal Dependency (UD) is becoming a standard annotation scheme cross-linguistically, but it is argued that this scheme centering on content words is harder to parse thant the conventional one centering on function words. To improve the parsability of UD, we propose a back-and-forth conversion algorithm, in which we preprocess the training treebankto increase parsability, and reconvert the parser outputs to follow the UD scheme as a postprocess. We show that this technique consistently improves LAS across languages even with a state-of-the-art parse, in particular on core dependency arcs such as noun modifier. We also provide an in-depth analysis to understand why our method increases parsability. language of the presentation: Japanese | |||
| 清川 拓哉 | 1651045: M, 1回目発表 | ロボティクス | 小笠原 司 |
|
title: Grasp Control of Deformable Objects based on Slip Prevention Strategy considering Profile Variation
abstract: Tactile sensing feedback manipulation has been verified for the purpose of developing robotic system and prostheses. In fact, their algorithms which take into consideration the characteristics of objects is useful for objects with unexpected mass variation, different friction properties and deformable objects. However, in their experiments, objects whose profile variation with the movement of the contents are not tested, and the grasp force are regulated after the slippage occurred besides. Therefore, in this research, we propose an estimation model of contact surface profile variation and a grasping force regulation algorithm by visual feedback based on the model in order to realize stable grasp suppressed slippage occurrence. language of the presentation: Japanese | |||
| 塩竃 実 | 1651058: M, 1回目発表 | ロボティクス | 小笠原 司☆ |
|
title: Measurement and Evaluation of Foot Status in Running
abstract: Recently, a running boom is breaking out by rise in health awareness, low cost and free goal setting. However, many novice runners who don’t have knowledge about running are injured and are not enjoyable. One of injuries is “overpronation”. It causes pain of knee or Achilles tendon. Neverthless, many of them don’t notice their foot status. And so I propose feeding back foot status to runners by using multiple IMU sensors. language of the presentation: Japanese | |||