| 0151206 | KHIAT ABDELAZIZ | ロボティクス |
|
発表題目：Robot Architecture for Human-Robot Interaction: Case of the Humanoid Robot ASKA.
発表概要：Robots that can interact and communicate with human beings have always been a target for many researchers. Through this presentation, I will try to explain about the main technologies that contribute in this kind of applications. In the first half, I will introduce robot architectures and their necessity for the coherent development of any robot's software. Then, I will consider the Bayesian Networks technology that plays an important role in dealing with the inherent uncertainty in the act of communication. Finally, I will give the main steps for the realization of this project. | ||
| 0151207 | GOH CHOOI LING | 自然言語処理学 |
|
発表題目：Unknown Word Detection and Segmentation for Chinese Using a Hybrid Approach
発表概要：Word is defined as the minimal meaningful unit that can function independently of any natural language, and it plays a primary role in any further processing in Natural Language Processing system such as Machine Translation. Unlike languages such as English where the spaces identify the word boundaries, there are no such delimiters in Chinese text as in the case in Japanese. A sentence is written as a string of characters without separation between words. Therefore, before we can process a text further, we must identify the word boundaries in the text. This process is referred as text segmentation and has to resolve problems such as segmentation ambiguities and occurrence of unknown words. Sometimes there are several ways to segment a sentence with each segmentation result yielding a different meaning. We must segment the text taking the context into consideration. Unknown words are defined as the words that are not in the dictionary. In this situation, the segmentation system will not be able to identify it correctly. Words in Chinese cannot be all defined in a dictionary as may be done in European languages. There is flexibility of creating new words by combining characters or words such as names of person, places or companies. It is neither possible for a dictionary to contain all words in Chinese, nor to specify all the rules for word formation. So a segmentation process should be able to recognize possible unknown words. In previous research, rule-based approach, statistical approach and hybrid approach have been used to solve the problem. In this research, we would like to apply the hybrid approach and further improve on the results. | ||
| 0151208 | SHARIFF BIN ABDUL RAHMAN | 情報コミュニケーション |
|
発表題目：<題目>Multiuser Detection Based on Radial Basis Function for a Multicode DS/CDMA System
発表概要：<概要> DS/CDMAシステムの準最適受信方法の一つとして近年、Neural Networkの学習を適用した受信手法が注目されている。本発表ではNeural Networkの一種であるRadial Basis Function Networkを適用したDS/CDMA受信方式を紹介する。そしてマルチレート伝送システムでのシミュレーションの結果からこの方式の優位性を考察する。 | ||