ゼミナール発表

日時: 9月22日(月)4限 (15:10-16:40)


会場: L1

司会: Graham Neubig
大内 啓樹 1351014: M, 2回目発表 松本 裕治, 中村 哲, 新保 仁, Kevin Duh
title: Improving Dependency Parsers with Supertags
abstract: Transition-based dependency parsing systems can utilize rich feature representations. However, in practice, features are generally limited to combinations of lexical tokens and part-of-speech tags. We investigate richer features based on supertags, which represent lexical templates extracted from dependency structure annotated corpus. First, we develop supertags that encode information about head position and dependency relations. Then, we propose a transition-based dependency parser that incorporates the predictions from a supertagger as new features. On standard English Penn Treebank corpus, we show that our supertag features achieve parsing improvements of 1.38% in unlabeled attachment.
language of the presentation: Japanese
 
俵 雄貴 1351067: M, 2回目発表 松本 裕治, 中村 哲, 新保 仁, Kevin Duh
title: Joint Morphological Analysis and Dependency Parsing
abstract: In Japanese, morphological analysis has achieved high performance using sequential information. However there exist the cases which can not solved only with sequential information. For such cases, dependency information can be an important cue to analyze correctly. By jointly solving morphological analysis and depencency parsing, we aim to utilize dependency information in morphological analysis. This presentation shows details of the joint algorithms.
language of the presentation: Japanese
 
三浦 未来 1351101: M, 2回目発表 松本 裕治, 中村 哲, 新保 仁, Kevin Duh
title: Answering the true/false question by Recognizing Textual Entailment
abstract: Recognizing Textual Entailment(RTE) task is defined as recongnizing, given two text fragments, whether the meaning of one text can be entailed from the other. Recently, shared task of RTE is suggested. Paticipants of this task answer the true/false question by RTE. This presentation proposes new method by the data division for the RTE shared task.
language of the presentation: Japanese
 
皆川 正浩 1351103: M, 2回目発表 松本 裕治, 中村 哲, 新保 仁, Kevin Duh
title: Automatic Interpretation of Japanese Verb Metaphor
abstract: We try to investigate whether we can apply an existing approach to interpret English verb metaphors to Japanese ones. This approach is based on a proposal to define metaphor intepretation as a paraphrasing task. In this presentation, we first talk about the significance of interpreting metaphors in Natural Language Processing application, and then talk about the examples of outputs obtained from the approach. At the end of this presentation, we discuss what yet remains to be done.
language of the presentation: Japanese
 

会場: L2

司会: 玉井 森彦
坂井 昭仁 1351045: M, 2回目発表 伊藤 実, 安本 慶一, 楫 勇一, 関 浩之
title: Data authentication scheme suitable for broadcast in the sensor networks
abstract: Data broadcasting is often utilized in a sensor network system, but the authentication of the broadcast data is not a simple issue. The computational resources of sensor nodes are so restricted that common security techniques such as digital signatures may not be available in a sensor network system. In this study, a novel mech- anism for the authentication of broadcast data is discussed. The proposed mechanism makes use of several pairs of one-way hash chains, and realizes secure and light-weighted scheme for the authentication of broadcast data. This study also investigates the relation between the security and the complexity, and proposes a means to derive an optimum choices of parameters.
language of the presentation:Japanese
発表題目: センサネットワークにおける放送型通信に適したデータ認証方式
発表概要: 多くのセンサネットワークシステムでは,サーバからの放送型通信を有効に活用して各種の機能を実現し ている.通信の安全性を確保するには,放送型通信においてデータ認証を実現する必要があるが,センサノードは非 常に限られた計算資源しか持たないため,電子署名等の汎用技術を利用することは困難である.そこで本研究では, 複数のハッシュ連鎖を組み合わて利用することで,放送型通信におけるデータ認証を安全かつ軽量に実現する方式を 提案する.さらに,提案方式の安全性と計算量の関係について考察し,ある意味で最適なパラメータの選択方法につ いて議論を行う.
 
冨永 拓也 1251069: M, 2回目発表 伊藤 実, 安本 慶一, 柴田 直樹
title: A Proposal of evacuation guidance system using the light of the smartphones
Abstract: In the underground shopping center when there is a power outage, we cannot see floors or walls. In this context, we have to use evacuation lights in order to evacuate from it. However, according to the existing study, there are only twenty percent of people using evacuation lights, and it does not serve as the role of the evacuation instruction. In this paper, we propose a evacuation guidance system using smartphone's light (smartphone's light is a generic name of backlight and flash light of the smartphone). This method controls each smartphone's light so that light seems to spread in the direction where you should evacuate when an evacuee looked at the floor. In other words, look at the floor, the evacuees should evacuate to the direction where the light seemed to spread. A proposed system consists of evacuation guidance devices and smartphones which evacuees carry. The evacuation guidance devices to be built into the evacuation light, and its power uses a battery of evacuation light. In addition, in order to know the smartphones situation requiring evacuation, the evacuation guidance devices have a wireless LAN. When the underground city is blacked out, each evacuation guidance device broadcast packet that notify the start of the evacuation algorithm (start packet) to neighbors smartphone's. Each smartphone that received a start packet, control it's smartphone's light built-in autonomous distributed algorithm so that light seems to spread in the evacuation direction. We conducted simulations using 3D movies in order to evacuate the proposed method by questionnaire. Results of the questionnaire, confirms that light seems to spread in the evacuation direction when evacuees look at the floor.
language of the presentation: Japanese
発表題目: 地下街におけるスマートフォンの光を用いた避難誘導方式の提案
発表概要: 停電した地下街では,壁や床等が見えず,避難者は唯一の目印である避難誘導灯を用いて避難することになる. しかし先行研究によると避難誘導灯を利用する避難者は2割程度であることがわかっており,避難誘導の役割を十分に果たしているとは言えない. 本稿では,避難者の携えるスマートフォンの発する光(バックライトとフラッシュライト,総じてスマホライトと呼ぶ)を用いた避難誘導方式を提案する. 避難者が床を見た際に,避難すべき方向(避難方向)に光の帯が流れるように見えるよう,各スマホライトを制御する手法を取る.すなわち,避難者は光が流れるように見えた方向に避難すればよい.提案するシステムは避難誘導装置と避難者の携えるスマートフォンからなる.避難誘導装置は避難誘導灯にビルトインし,電源は避難誘導灯の蓄電池を利用する.また,避難誘導が必要な状況をスマートフォンに知らせるために,避難誘導装置は無線LANを具備することとする. 停電が発生すると,避難誘導装置は,避難誘導アルゴリズムの開始を知らせるパケット(開始パケット)を近隣のスマートフォンにブロードキャストする. 開始パケットを受け取った各スマートフォンは,あらかじめ設定されている自律分散型アルゴリズムにしたがって,避難者から見て避難方向に光が流れるように見えるよう,スマホライトを制御する. 提案手法を評価するため,避難者の目線による3D動画を用いたシミュレーションを行い,アンケートにより評価した. アンケートの結果,避難者が床を見ることで,光が避難方向に流れるように見えることを確認した.
 
BABATUNDE OJETUNDE SEGUN 1351116: M, 2回目発表 伊藤 実, 安本 慶一, 柴田 直樹
title: Endorsement based Mobile Payment System for A Disaster Area
abstract: Payment system in a disaster area is essential for people to buy necessary amenities, like groceries, clothing, medical supplies. However, existing payment systems require communication infrastructures (like wired networks and cellular networks) to enable transactions, and thus cannot function well in disaster areas where these communication infrastructures may be likely destroyed. In this presentation, we propose a mobile payment system by adopting infrastructureless mobile ad-hoc networks (MANETs), which can allow users to purchase amenities in disaster areas while providing secure transactions. Specifically, we propose an endorsement-based scheme to guarantee each transaction and a location information based monitoring scheme to achieve transaction validity and reliability. By employing ecoin, blind signature and one-time session token techniques, our mobile payment system can also prevent collusion, reset and recovery attacks.
language of the presentation: English
 
徐 家興 1351125: M, 2回目発表 伊藤 実, 安本 慶一, 柴田 直樹
title: GreenSwirl: Combining Traffic Signal Control and Route Guidance for improving efficiency of the vehicle traveling
abstract: Serious traffic congestion is a major social problem in large cities. Inefficient setting of traffic signal cycles, especially, is one of the main causes of congestion. GreenWave is a method for controlling traffic signals which allows one-way traffic to pass through a series of intersections without being stopped by a red light. GreenWave was tested in several cities around the world, but the results were not satisfactory. One of the problems with GreenWave is that since it still stops the crossing traffic, and it forms congestion in the traffic turning into or out of the crossing streets. To solve these problems, we propose a method of signal control method GreenSwirl, in combination with a route guidance method GreenDrive. GreenSwirl controls traffic signals enabling so that "green waves" of smooth-flowing traffic on various circular routes throughout the city: "swirls" of traffic, as it were, extending the GreenWave technology. We also use navigation systems to optimize overall control of the city's traffic. We did a simulation using the traffic simulator SUMO and the road network of Manhattan Island in New York. We confirmed that our method shortens the average travel time by 10%-60%, even if not all cars on the road are equipped to use this system.
language of the presentation: Japanese
発表題目: GreenSwirl--車両走行効率の向上を目指した信号制御および経路案内方式
発表概要:近年,大都市で深刻な交通渋滞が社会的問題となっている.特に渋滞を引き起こす原因の一つとして非合理的な交通信号サイクルがある. 信号制御の技術としてGreenWaveが中国の複数の都市で実験されてきたが,結果は満足できるものではなかった.GreenWaveは一定速度で走行する車両は連続する交差点を常に青信号で通過できる技術である. GreenWaveの問題点として幹線道路のみに生成されるため,対向車線と横断道路の妨害、入口と出口の渋滞などを引き起こしてしまうことが挙げられる.この問題点を解決するために本稿では信号制御方式GreenSwirlおよび経路案内方式GreenDriveを提案する. 提案手法では複数のGreenWave道路を渦巻き状に発生させ,GreenDrive案内方式で道路を走行する時間を見積もり,車両の平均走行時間を最小化する.提案手法の性能を評価するために交通流シミュレータSUMOを用いてシミュレーションを行った. ニューヨーク市マンハッタン島の道路網で車両の走行時間短縮効果を計測した結果,従来の手法と比べて提案手法は平均10〜60%程度,平均走行時間が短縮できたことを確認した.