| LIU SHANSHAN | M, 1回目発表 | 光メディアインタフェース | 松本 裕治, 中村 哲, 新保 仁, 進藤 裕之 |
|
Title: Information Extraction of Chemical Material Synthesis Based On Neural Network Methods
Abstract: For today's researchers, how to efficiently access the latest knowledge has been plaguing them. Researchers always spend a lot of time searching and reading in order to use this new knowledge to help their research. The same is true in the field of chemical synthesis. This study refers to using neural network methods to process the experimental texts in chemical papers to obtain the process of material synthesis. The operations performed on the substance will be considered as an event, while the raw materials and conditions of the operation will be the arguments of this event. And we will try to get the relationship between events to form a complete chemical synthesis process. The current idea is to extract events using the Bi-LSTM (Bidirectional Long Short-Term Memory) network and the GCN (Graph Convolutional Network) method, and then use the Bi-LSTM network with the attention mechanism to obtain the relationship between events. Language of the presentation: English | |||
| PHAM HOAI LUAN | M, 1回目発表 | コンピューティング・アークテクチャ | 中島 康彦, 笠原 正治, 中田 尚, Tran Thi Hong, 張 任遠 |
|
Title:
A Secure Remote Patient System for Hospital Using Blockchain Smart Contract
Abstract: Nowadays, a combination between Internet of Things (IoT) technology and remote mtient monitoring is extensively researched due to its efficiency and convenience for human life. When the number of IoT devices in health care system is increased exponentially, the privacy and security issues of patients are becoming a concern. In order to protect personal and device-generated information, we propose to use blockchain-based smart contracts for managing patients' information and medical devices. We have verified the proposed smart contract on Ethereum test environment called TESTRPC and implemented the system on an experimental environment with real devices. This system works well at small scale. Language of the presentation: English | |||
| SASHI NOVITASARI | M, 1回目発表 | 知能コミュニケーション | 中村 哲, 松本 裕治, 作村 諭一(BS), Sakriani Sakti |
|
Title: Sequence-to-Sequence Incremental Speech Recognition by Attention Transfer
Abstract: Attention-based sequence-to-sequence automatic speech recognition (ASR) requires a significant delay to recognize long utterances. Several studies proposed mechanisms with neural network for incremental speech recognition (ISR), an ASR with a short delay, but resulted in more complicated frameworks than the standard attention-based ASR. In this work, we aim to construct ISR based on the attention-based ASR. To enable the ISR to output the transcription that aligned to a short speech segment, we propose to utilize attention-based alignment by full-utterance ASR to teach the ISR. In the character-level recognition task with small corpus, the proposed ISR with the best configuration shows comparable performance to the full-utterance ASR. Towards automatic simultaneous speech translation (SST), we also aim to integrate the proposed ISR into the SST system, in which the ISR output will be used by machine translation (MT) module. However, MT generally requires sub-word or higher representation as input. Therefore, we plan to construct a sub-word level model by also exploring further ISR architecture. Language of the presentation: English | |||
| THONGLEK KUNDJANASITH | M, 1回目発表 | ソフトウェア設計学 | 飯田 元, 藤川 和利, 市川 昊平, 高橋 慧智 |
|
Title: A Long Short-Term Memory model for Efficient Resource Utilization in Data Centers
Abstract: Data centers are centralized locations where computing and networking equipments are aggregated to handle large amounts of data and computation efficiently. In a data center, computing resources such as CPU and memory are managed by a centralized resource manager. The resource manager accepts resource requests from users and allocates resource to applications. A commonly known problem is that users often request more resource than their applications actually use, which degrades the overall resource utilization in the data center. The objective of this work is to improve the resource utilization in data centers by predicting the optimal resource allocation for each job. We apply Long Short-Term Memory (LSTM), which is a time-series deep learning technique, to predict the optimal resource allocation of jobs based on historical resource usage data. Language of the presentation: English | |||
| 内嶺 佑太 | M, 1回目発表 | 自然言語処理学 | 加藤 博一, 清川 清, 太田 淳(MS), 神原 誠之, Alexander Plopski |
|
Title: Evaluation of The Imaging Method in The Prototype of Micro-Lens Array HMD
Abstract: Most commercial VR HMDs have a large form-factor that limits the use of the device. This is because a lens is placed inside of the device to help human's accommodation for a near-eye display. To design the device with a smaller form-factor, the shorter focal length lens should be employed. Although the HMD concept that employs micro-lens array with very thin design has been proposed, it produces undesired images such as ghost images and artifacts that lose image quality. Since the effectiveness of the proposed imaging method that addresses those undesired images has been confirmed only in a simulator, the aim of this study is first to evaluate the imaging method on the actual prototype model of this micro-lens array concept. As the first steps for the prototyping process, understanding the concept from mathematical theories, fabricating a camera head mount and demonstrating eye-tracking in a program have been done. Language of the presentation: Japanese 発表題目: レンズアレイを用いたHMDにおける提案イメージング手法の実システムでの評価 発表概要: 非透過型のVR-HMDの多くの市販製品は，大きなフォームファクタを持っており，用途を制限する要因となっている. これは，接眼ディスプレイに人間の目が焦点を合わせるために用いられるレンズが要因となっている. デバイスを小さくするために，焦点距離が小さいレンズを用いてディスプレイとレンズ間の距離を縮める手法があり，複数の小さなレンズから成るマイクロレンズアレイを用いて3Dイメージを投影する手法が提案されている. このコンセプトにおいてはGhost imageやArtifactsといった望まれないイメージが発生するため，これを除去することのできる瞳孔条件により変化するイメージング手法が既に提案されている. 一方，それらは全てシミュレーション上での評価であり，実際のシステムにおいて検証されていない. そこで，本研究では，まずプロトタイプを作成し，実システムでのイメージング手法の評価を行うことを目的としている. プロトタイプ作成に向けて，コンセプトの理解のために理論面の学習，3Dプリンタでのカメラヘッドマウントの試作，アイトラッキングのテストを行った. | |||