コロキアムB発表

日時: 9月28日(金)2限(11:00~12:30)


会場: L1

司会: Soufi Mazen
北村 勇希 M, 2回目発表 ロボティクス 小笠原 司, 佐藤 嘉伸, 高松 淳, 丁 明, Gustavo Garcia
title: Development of Soft Hand with Temperature Controllable and Passive Mechanism for Automation of Touch Care
abstract: "Touching" is an important communication method for human beings. There are various methods such as reflexology and acupressure for this care. Touch care has attracted attention in the field of nursing care and medical care in recent years, and it is a method eases stress, pain and dementia. However, due to the lack of nursing care and medical staff, it is difficult to do thoroughly. If we can automate touch care, we can expect the reduction of burden on nursing care staff. In this research, we aim to develop a flexible hand having a temperature controllable and passive mechanism for the automation of touch care. First, a torsion spring is used to reproduce the contact force when touching like a human being. Second, in order to reproduce the warmth, a heater and sensor are equipped to control the temperature of the hand. Currently, as a first step making toward a hand, we made prototype fingers. In the future we plan to work on making the hand.
language of the presentation: Japanese
 
尾﨑 隼平 M, 2回目発表 数理情報学 池田 和司, 佐藤 嘉伸, 吉本 潤一郎, 中野 高志
TItle: Developing a decoder to estimate emotional states of the mice from whisking movement.
Abstract: It has been reported that mice's whiskers characteristically move after the stimulus related to the reward in conditioning tasks. It means that the internal state (positive emotion) of the mice may be output as mice's whisking movement. In this study, we verify its possibility. therefore we developed a decoder to estimate states of the mice from whisking movement in the Hearing Go/No-Go Task. As a result, we can find a difference between a reward anticipation and an non reward anticipation in whisking movement.
Language of the presentation: Japanese
 

会場: L2

司会: 吉野 幸一郎
村山 太一 M, 2回目発表 自然言語処理学 松本 裕治☆, 中村 哲, 荒牧 英治, 鈴木 優
title: The analysis of resources and new model for the influenza prediction.
abstract: Predicting the epidemics of infections can reduce the time of public health professionals and medical institutions, resulting in more efficient service for people. We focused on one of the epidemics, influenza. When the model for influenza prediction is made generally, UGC(user-generate-content) are used as predictors but, we don't know which UGC content is userful. In this research, we analyse the resources for flu prediction and make the new model for flu prediction considering seasonal movement.
language of the presentation: Japanese
 
山本 英弥 M, 2回目発表 自然言語処理学 松本 裕治☆, 中村 哲, 荒牧 英治, 田中 宏季
title : Improvement on Estimating Difficulty of Words and Applying to Medical Terms
abstract : In clinical, communication between patient and doctor is important for good mediation.
When you talk with doctors, however, you maybe hear medical technical terms which is difficult to understand. To recognize the existence of these technical terms each other is important to achieve smooth communication. To attack this problem, for estimating difficulty of medical terms, we improved difficulty estimating method by using frequency in Twitter and applied to medical terms. To ask difficulty of each word and validate this method, we employed crowdsourcing. As a result, list of difficult terms became larger than before, but this method works not so good. This estimation method has margin for improvement and validation is not easy.
language of the presentation : Japanese
 
石原 敬大 M, 2回目発表 自然言語処理学 松本 裕治, 中村 哲, 新保 仁, 進藤 裕之
title: Text Reperesentation Using Distributed Tree Kernel
abstract: Structures of sentence such as word order and parse tree are important feature in natural language processing. Recent neural network approaches reflect such structure in their model architecure, which causes high computational costs. To avoid such issues, we propose the way to create parse tree representation using distributed tree kernel (DTK).
language of the presentation: Japanease
 
渡邊 賢也 M, 2回目発表 自然言語処理学 松本 裕治, 中村 哲, 新保 仁, 進藤 裕之
title: Relation Extraction for Biomedical text
abstract: In some domain, there are too many thesis to read by human these days. So, I analysis thesis by named entity recognition (NER), entity linking (EL) and relation extraction (RE) for finding thesis we want. I used biomedical text data (BioCreative V). In this time, I improved existing muti-task model for named entity recognition and relation extraction (BRAN).
language of the presentation: Japanese
 

会場: L3

司会: 藤本 まなと
德田 佐紀 D, 中間発表 計算システムズ生物学 金谷 重彦, 安本 慶一, 小野 直亮, MD.ALTAF-UL-AMIN, 黄 銘
title: Differences in energy intake and body mass index among the traditional Chinese medicine constitutions.
abstract: Previous studies showed that obesity ratios or body mass index (BMI) differed between the types of the traditional Chinese medicine constitution (TCMC). I examined the differences in energy intake between TCMC types using data from the crowdsourcing survey and try to discuss its contribution to BMI. There were not significant differences in energy intake between TCMC types. But BMI was significantly lower in XX than in YY. Significant differences in sex, height, and exercise habits are not considered yet. Accordingly, this result does not completely suggest that energy intake might not relate with the difference in BMI.
language of the presentation: Japanese
 
石谷 勇人 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, 黄 銘, 小野 直亮
title: Relationship between 25 items in Back Pain Evaluation Questionnaire by Japanese Orthopaedic Association (JOABPEQ) and the Physical Findings Based on Multivariate Analysis.
abstract: The purpose of this study was to investigate the possibility of constructing an expression model using the physical findings and answers of 25 questions in JOABPEQ. We reviewed 301 patients with low back pain. The scores for the five categories were calculated. We measured the degrees of pain using a visual analog scale and also measured the angle of the hip joint (ROM). All answers and physical function data were subjected to cluster and partial least squares (PLS) regression analyses. The five categories and low back pain were explained by answers to the 25 questions, as revealed by PLS analysis. However, the other items could not be modeled using answers to the questions.
language of the presentation: Japanese
 
井川 優太郎 M, 2回目発表 知能システム制御 杉本 謙二, 安本 慶一, 松原 崇充, 小林 泰介
title: Development of Biomechanical Energy Harvester Applicable to Various Movements
abstract: We have tried creating knee mount type energy harvester equipped with a belt type continuously variable transmission (CVT) and investigated the amount of electric power and the influence on the human motion due to the gear ratio. In this time, we made improvements to increase the electric power of the harvester and propose a gear ratio optimization strategy using Bayesian optimization to search the optimum gear ratio for each motion with a small number of experiments.
language of presentation: Japanese