| 佐々木 俊樹 | 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: Answer Sentence Selectoion for Machine Reading
abstract: Machine Reading is major goal in Natural Language Processing. In recent research, it is close to or beyond human performance in both dataset which required only simple inference and multiplie-sentence reasoning. However, it is important to be able to use it as a supplement to human beings in the real world. We propose a system to select not only the answer to the question but also the sentences within the context necessary to guide the answer. language of the presentation: Japanese | |||
| 芝原 隆善 | 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: Japanese Predicate Argument Structure Analysis Coping with term preference of predicates by Using Similar Usage
abstract: I plan to talk about Japanese Predicate Argument Structure(J-PAS) Analysis coping with term preference of predicates by referring similar usage. In this presentation, I'll talk about a problem setting of J-PAS, details of a proposal method, analysis accuracy and issues of preliminary experiment. language of the presentation: Japanese 発表題目: 類似用例から述語の選択選好性を利用した日本語述語項構造解析 発表概要: 本研究では，類似用例を参照することで，述語の選択選好性(ある述語のある格として実現しやすい単語の傾向)を利用した日本語述語項構造解析について紹介する．本発表では，J-PASの問題設計，先行研究と予備実験について述べる． | |||
| 吉川 将司 | 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 |
|
title: A* CCG parsing with category and dependency factored model
abstract: I pursue natural language processing based on Combinatory Categorial Grammar (CCG), a linguistic theory that is also known for its usefulness in tasks such as question answering. In this presentation, I propose a new A* CCG parsing model in which the probability of a tree is decomposed into factors of CCG categories and its syntactic dependencies both defined on bi-directional LSTMs. The factored model allows the precomputation of all probabilities and runs very efficiently, while modeling sentence structures explicitly via dependencies. The model achieves the state-of-the-art results on English and Japanese CCG parsing. language of the presentation: Japanese | |||
| NGUYEN HUU HOANG | 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: Hierarchical Attention Network for Text classification
abstract: In SNS, classifying and extracting information of user opinion is very important task. In this research, we try to leverage the Hierarchical Information of document. The baseline method use RNN(recurrent neural network), so it needs so much time to process documents. So, we tried to use CNN(convolutional neural network) to reduce the processing time. language of the presentation: Japanese | |||
| 上田 朝己 | M, 2回目発表 | 光メディアインタフェース | 向川 康博, 加藤 博一, 舩冨 卓哉, 久保 尋之, 田中 賢一郎 |
|
title: Acquisition of Light Transport by Controlling the Synchronization and the Scan Speed in Projector-Camera System
abstract: To extract scene characteristics, it is effective to acquire light transport which is the relation between the incident illumination and outgoing radiance from the scene. In this research, I realized the acquisition of light transport with a different aspect than conventional way by synchronizing the rolling shutter of the camera with the raster scanning of the laser projector, and by controlling the scan speed of the rolling shutter. In my presentation, I show the result of acquiring the light transport by using the original hardware and show the effectiveness of this method and its application. language of the presentation: Japanese 発表題目: プロジェクタカメラシステムの同期遅延と走査速度の制御にもとづく光伝播の計測と応用 発表概要: プロジェクタでシーンを照明したときの様子をカメラで撮影することで，シーンの光伝播を計測し，様々な性質を推定することが可能である．本研究では，レーザプロジェクタとローリングシャッタカメラを用いて撮像タイミングを同期し，さらにローリングシャッタカメラの走査速度を制御することによって，従来とは異なるモダリティで光伝播の観測を可能にした．本発表では，実際に構築したシステムを用いて取得した光伝播を示し，本手法の有効性とその応用について述べる． | |||
| 中西 由希子 | M, 2回目発表 | 光メディアインタフェース | 向川 康博, 加藤 博一, 舩冨 卓哉, 久保 尋之, 田中 賢一郎 |
|
title: Mathematical Model for Pop-up Effect of ChromaDepth
abstract: ChromaDepth is a micro prism glass which produces pop-up effect by combining refraction and diffraction. Unlike other 3D glasses, the parallax angle of the ChromaDepth is independent to the distance from the display.Throughout many past experiments, we unintentionally realized that the pop-up distance changes depending on the distance between screen and observer viewpoint.In this paper, we formulate the relation between the pop-up distance and viewpoint distance using a mathematical model. Based on our model and optical measurement, we evaluate the human perception of the pop-up effect for ChromaDepth using a distant large display. language of the presentation: Japanese | |||