ゼミナール発表

日時: 09月28日(木)2限(11:00-12:30)


会場: L1

司会: 藤本 まなと
加藤 明彦 1661004: D, 中間発表 自然言語処理学 松本 裕治, 中村 哲, 新保 仁, 進藤 裕之, 能地 宏
Title: Construction and Analysis of MWE-aware Dependency Corpus
Abstract : Because syntactic structures and spans of multiword expressions (MWEs) are independently annotated in many English syntactic corpora, they are generally inconsistent with respect to one another, which is harmful to the implementation of an aggregate system. In this work, we construct a corpus that ensures consistency between dependency structures and MWEs, including named entities. Further, we explore models that predict both MWE-spans and an MWE-aware dependency structure. Experimental results show that our joint model using additional MWE-span features achieves an MWE recognition improvement of 1.35 points over a pipeline model.
language of the presentation: Japanese
 
太田 瑶子 1651023: 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, 能地 宏
title: Create of an Automatic Haiku Generator Using Deep Learning
abstract: Haiku is the shortest Japanese poem in the world. Haiku is composed of three phrases, each of them is called the top five, the middle seven, the bottom five. Also, haiku has rules of mora constraints, constraints of a season word, cutting word restrictions. In this research, I want to create an automatic haiku generator that can follow the rules of haiku and express expressions flexibly by using deep learning. In this presentation, I will describe the previous research, the experiment to select the last phrase (bottom five) as a preliminary experiment,its result, and future plan. language of the presentation: Japanese
 
久保 大輝 1651046: 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, 能地 宏
title: Detection of Japanense Functional Expressions Using Deep Learning
Abstract: A Functional Expression is an expression in which a plurality of words has one grammatical function as a whole. A Functional Expression is characterized by non-compositionality of meaning. Therefore, it is important to consider Functional Expressions in applied tasks such as Information Retrieval. In this research, we attempted to detect Functional Expressions using deep learning. We will report on the results and future issues.
language of the presentation: Japanese
 

会場: L3

司会: Raula G. Kula
田中 雄大 1651070: M, 2回目発表 モバイルコンピューティング 伊藤 実, 安本 慶一, 柴田 直樹, 川上 朋也
title : Delivery Scheduling for Relief Supplies Considering Realistic Situations in a Disaster Area
abstract : Due to earthquakes and typhoons, various parts of Japan are suffering severe damage. When the general administrative agencies that do not normally deliver are also involved in the delivery due to the damage, there is a problem that information sharing between each agency and transportation companies is not done. Great disarray in delivery of relief goods is caused by these and many problems that delivery of goods to evacuation centers is not sufficient are caused.Various researches have been done on these problem, but it is pointed out that the number of items handled is not necessarily large, and there is a difference in the quantity of goods at each delivery base. There is a possibility that this can be improved by rebalancing which adjusts the quantity of goods between bases. Also, in the event of a disaster, it is difficult to deliver goods due to shredding of the road and a place where traffic is difficult.In this paper, we propose a logistics model to accurately deliver goods to evacuation centers, assuming that supply of supplies is insufficient immediately after a disaster. In order to solve the above problem, we optimize delivery routes that rebalance goods and trucks between delivery bases and incorporate demand amount and supply amount as information, so that shortage of goods does not occur.
language of the presentation : Japanese
 
堺井 大地 1651050: M, 2回目発表 モバイルコンピューティング 伊藤 実, 向川 康博, 柴田 直樹, 川上 朋也
title: Examination of encoding method enabling dimming in visible light communication using rolling shutter effect
abstract: A dedicated module was required for conventional visible light communication. Recent studies have realized visible light communication using LED illumination and mobile phone camera by using rolling shutter effect. However, these existing studies have problems such as poor energy because of dimming by voltage. Therefore, in this reseach, we will consider coding method which enables dimming without changing voltage.
language of the presentation:Japanese
 
政野 博紀 1651097: M, 2回目発表 モバイルコンピューティング 伊藤 実, 藤川 和利, 柴田 直樹, Gao Juntao
title: A Study for Operating Cost-Efficient On Demand Bus System
abstract: On demand bus has been increasingly attractive for many applications, such as transportation in sparsely populated areas and self-driving vehicle based public transport services. However, most of current on demand bus systems are not efficient in providing transport services and making profit. In this research, we propose an efficient on demand bus scheduling algorithm that dynamically determines points of pick-up and drop-off to reduce detour route of on demand bus. Users need to walk to pick-up points from their starting location and walk from drop-off points to their destinations.
language of the presentation: Japanese
 
宮村 純真 1651105: M, 2回目発表 ソフトウェア設計学 飯田 元 ☆, 松本 健一, 片平 眞史(客員), 石濱 直樹(客員), 川口 真司(客員)
title:Classification of problem detection for incomplete software requirements
abstract: Software faults in spacecraft software can largely halter the success chance of missions. In this research, we investigate the lack of requirements, which is considered to be a main cause leading to software faults. We approach the above problem from the viewpoints for malfunction types and process. In this presentation, we will explain about the retrieved current results.
language of the presentation: Japanese