| 國吉 房貴 | 1551040: M, 2回目発表 | 光メディアインタフェース | 向川 康博,佐藤 嘉伸,舩冨 卓哉,久保 尋之 |
|
title:Visibility enhancement for cell imaging
abstract: （企業との共同研究のため非公開） language of the presentation:Japanese 発表題目:細胞イメージングのための視認性向上手法の提案 発表概要: （企業との共同研究のため非公開） | |||
| 櫛田 貴弘 | 1551039: M, 2回目発表 | 光メディアインタフェース | 向川 康博,佐藤 嘉伸,舩冨 卓哉,久保 尋之 |
|
title: Measuring reflectance distribution through light scattering in participating media
abstract: We aim to measure reflectance distribution of real objects. There are various objects in our lives and each object has peculiar reflectance. Measuring the reflectance distribution of such objects is difficult because it requires many measurements of each incident and reflected light directions. We observe the reflected light outgoing to every direction through scattering in participating media, and estimate the reflectance distribution from the scattered light distribution. language of the presentation: Japanese | |||
| 日朝 祐太 | 1551083: M, 2回目発表 | 生体医用画像 | 佐藤 嘉伸,向川 康博,大竹 義人,横田 太 |
|
title: Constrained piecewise rigid 2D-3D registration for patient-specific analysis of rib cage motion using X-ray video
abstract: The respiratory function has been commonly evaluated by a spirometer in clinical routine. Although the spirometer is non-invasive and provides useful information about overall respiratory function, diseases caused by local lung dysfunction are not easy to diagnose. The purpose of this study was to provide an alternative rib motion diagnosis tool that achieves higher time resolution and three-dimensional analysis while keeping the radiation dose at the conventional level. For this purpose, we developed a robust 2D-3D registration algorithm between x-ray video and a one-time-phase CT using constraints based on the anatomical knowledge of the articulation at the costovertebral joint. language of the presentation: Japanese | |||
| 須藤 広大 | 1551055: M, 2回目発表 | 自然言語処理学 | 松本 裕治,中村 哲,新保 仁,能地 宏 |
|
title: Distant Supervision for Relation Extraction abstract: Distant Supervision technique has been widely used to find novel relational facts from text. However, Distant Supervision assumption is include noise texts, which means acoompanies with the wrong labelling problem, and these noisy data will substantially hurt the performance of relation extraction. To alleviate this issue, we propose hierarchal attention-based model for relation extraction. In this model, we employ convolutional neural networks to embed the semantics of sentences. Afterwards, we build sentence-level attention over multiple instances based on previous research. Finally, We build bag-level weight-sum model, which can reduce bag-level noise. language of the presentation: Japanese | |||
| 田口 雄哉 | 1551058: 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: Optimizing Distributed Representation for Document Summarization
abstract: Document summarization aims to capture the important information from single or multiple set of documents. In recent years, distributed representation of words is used in many natural language processing tasks including summarization. That can well capture the meaning of words which have different surfaces but same meanings. However, previous studies only used distributed representation of words which is already trained. In this presentation, we will introduce the optimization of distributed representation of words for document summarization. language of the presentation: Japanese* | |||
| 永井 優城 | 1551065: M, 2回目発表 | 自然言語処理学 | 松本 裕治,中村 哲,新保 仁,能地 宏 |
|
title: Hungarian Alignment for monolingual sentence alignment
abstract:Sentence Simplification is the task to make texts more readable one for humans. Many researches make simpler sentence using the statistical machine translation. Wikipedia and Simple Wikipedia is often used as a parallel corpus. In this presentation, I will talk about hungarian alignment for constructing parallel corpus . language of the presentation: Japanese | |||
| 西本 慎之介 | 1551074: M, 2回目発表 | 自然言語処理学 | 松本 裕治,中村 哲,新保 仁,能地 宏 |
|
title: Sentiment analysis for different domains
abstract:Sentiment Analysis is an important research topic in NLP. Sentiment Analysis is defined as determining whether the author's attitude is positive or negative on a given text. However, it is more useful to find the opinion towards specific targets and positions than simply determining whether the author has a positive or negative attitude in general. For this reason, I deal with the detection of stances in tweets on my research. I define stance detection to mean automatically determining from text whether the author is in favor of the given target, against the given target, or whether neither inference is likely. Because of the fast changes in the political theme on the Internet, it often happens that the number of data is limited. Therefore, it is necessary to use external resources. In my presentation, the application of different methods of sentiment analysis and external resources will be shown. language of the presentation: Japanese | |||