Two master students of the Department of Computer and Information Science passed their thesis oral defense at the University of Macau on 31th May 2012.

Mr. Tian Liang, a master student in Software Engineering, defended his Master thesis entitled “Word Alignment Based on Maximum Matching Method”. The research focus of his research is Word alignment which plays a very important role in statistical machine translation field. One contribution of his work is the deducing of a formula, which is to predict the number of phrase pairs given the alignment points. In this thesis, experiments show that the higher quality of word alignment the fewer numbers of phrase pairs will be extracted. Another contribution of the work is the implement of a word alignment system (The system is named as M3Aligner) running in a Windows personal computer, which is to provide some convenience to those researchers familiar with Windows environment. The results in his thesis have well explained the possibility of utilizing a simple model to produce the word alignment instead of complex mathematics models, and might be a valuable reference for further research in such field. Empirical study demonstrates that the proposed method gives a better alignment result than that of the GIZA++, especially for parallel sentences consist of phrases, idioms or expressions, which can be adapted easily in personal desktop computers.

Mr. Zeng Xiao-Dong(Samuel), a master student in E-Commerce technology, defended his Master thesis entitled “Research of iEnsembler Learning Algorithm and Its Applications”. The research focus of his research is Ensemble Learning methodology in machine learning community. One primary contribution of his research is the design and implementation of a novel good-performance ensemble learning algorithm named iEnsembler. The brilliant characteristic of proposed methodology is to integrate ensemble pruning process that pruning the redundant classifiers by weighted trading off accuracy and diversity to form optimal ensemble. The experimental results, evaluated by various strategies on a number of benchmark datasets from UCI machine learning repository, demonstrated the solid evidences that iEnsembler is superior to others in most cases. Another contribution in his research is that the proposed iEnsembler learning algorithm has been applied to two real-life classification applications: ECG heartbeat identification and part-of-speech tagging. The corresponding achievements again proved the superiority of the algorithm.

Both students are supervised by Dr. Derek F. Wong and Dr. Lidia S. Chao.