Supervision
Research Areas
Fei welcomes research students (either Masters by research or PhD) in the following areas Sentiment Analysis Question Answering Systems Automated Reasoning Semantic Web
Completed PhD Students
Below is a list of completed PhD students and their research projects
Dr. Van Hung LE
Year of Completion	2010
Title of Thesis	Fuzzy Linguistic Logic Programming
Description of Project	In the real world, there are situations in which information may not be given in a quantitative form, but rather in a qualitative one, i.e., by linguistic terms. This may arise from different reasons. In some cases, due to its nature, information may be unquantifiable and thus can be stated only in linguistic terms. In other cases, precise quantitative information may not be stated due to either its unavailability or a high cost of computation, so a linguistic ``approximate value" can be acceptable. Moreover, humans reason mostly in terms of linguistic terms rather than in terms of numbers and often use linguistic hedges (modifiers) to express different levels of emphasis. Therefore, it is necessary to investigate logical systems that can directly work with linguistic terms and make use of linguistic hedges since such systems make it easier to represent and reason with linguistically-expressed human knowledge and can be a foundation for applications dealing with linguistic information. The thesis introduces fuzzy linguistic logic programming (FLLP) for representing and reasoning with linguistically-expressed human knowledge. FLLP is a many-valued logic programming framework without negation. In FLLP, truth values are linguistic ones such as very true and approximately false, and linguistic hedges, for example, very and approximately, can be used as unary connectives in body formulae. Most concepts and results of traditional definite logic programming can have a counterpart in the framework. More precisely, it has all notions of declarative semantics, procedural semantics and fixpoint semantics. The sound and complete procedural semantics directly manipulates linguistic terms to compute answers to queries. The framework is shown to have applications in deductive databases, threshold computation and fuzzy control. Logic programs of FLLP can be translated into Prolog ones for being safely executed inside any standard Prolog interpreter.
Dr. Gang LI
Year of Completion	2013
Title of Thesis	Clustering-based Sentiment Analysis Approach
Description of Project	Sentiment Analysis has become a hot research topic, as it is a technology which can provide decision makers with online opinions. Online opinion expressing documents can be regarded as a huge and dynamic natural language database from which it is challenging to extract useful information. Traditional approaches have produced good analysis results. However, they have drawbacks in real applications. This thesis will introduce a new sentiment analysis framework–a clustering-based sentiment analysis approach. By applying a TF-IDF weighting method, a voting mechanism and importing term scores, an acceptable and stable clustering result can be obtained. The methodology has competitive advantages over the two existing types of approaches: symbolic techniques and supervised learning methods. It is a well-performed, efficient and non-human participating approach to solving sentiment analysis problems. Furthermore, several extension techniques for promoting the capability of clustering-based sentiment analysis system are introduced. Namely, applying opposite opinion contents processing and non-opinion contents processing techniques to further enhance the accuracy level; and using a modified voting mechanism and a distance measurement method to conduct three-class sentiment analysis. These plug-in techniques not only improved the performance of the clustering-based sentiment analysis framework, but also demonstrate the extendibility of the framework, which illustrates the possibility of further research along this track.
Dr. Manh Thang DO
Year of Completion	2013
Title of Thesis	ASR: ASP-Based Stream Reasoning For Semantic Sensor Data Processing and Activity Recognition
Description of Project	This thesis proposes a logic-based framework using Answer set programming for Stream Reasoning (ASR). The framework fills the gap between high speed but limited knowledge stream reasoning mechanisms, and “static” but rich knowledge reasoning techniques. The ASR language is a dialect of dlv with new functions: handling data streams, reasoning continuously about sensor data, reasoning about ambiguous situations and integrating with machine learning techniques to enhance reasoning. An implementation of the ASR framework is provided, together with a case-study called “AutoDiary”, which automatically recognizes user’s daily activities and situations, and provides machine understanding of human lifestyles.
Dr. Youseef Aid D ALOTAIBI
Year of Completion	2013
Title of Thesis	Business Process Modelling Challenges and Solutions: An Empirical Study
Description of Project	In recent years, research on Information Technology and business has identified key challenges in creating effective business process model (PM). Previous studies have pointed to three particular challenges: (1) difficulty of deriving IT goals from business goals and misalignment between business and IT; (2) security challenges; and (3) managing customer power and the rapidly changing business environment and business process (BP) challenges. A gap in the field is that almost all studies are mechanistic and fail to capture BPs as they apply to real business environments. Few studies address the difficulties for IT analysts when it comes to BPs, the result often being difficulties for IT professionals in developing information systems (IS) by BPs developed by non-IT managers or simply following business expectations. Furthermore, existing methods of IS development often do not meet the requirements to resolve security-related IS problems. Likewise, they frequently fail to successfully integrate security measures during all development process stages and only deal with specific security requirements, goals and constraints. In addition, few studies have been conducted in the area of classifying customers into different priority groups to provide services according to their required delivery time, payment history and feedback, which will enhance the company’s performance and profit. The focus of this study is to solve these challenges by proposing a business PM framework that has five stages: (1) modelling BP towards derivation of IT goals; (2) modelling a secured BP; (3) modelling BP managing customer power; (4) implementing the proposed business PM; and (5) evaluating the proposed business PM. A number of technical tools are utilised in the research. These include i* model technique, goal tree, and UML use case which are used to model BP towards derivation of IT goals. The i* model technique, UML sequence diagram, UML state chart diagram, and UML class diagram are used to model the secured business PM. The research study uses a case study on a mobile phone order management process in a telecommunication company and hotel booking management process to validate the proposed framework. Java, PHP and MySQL have been used to implement our proposed business PM. To evaluate the model, an empirical analysis based on data from 130 business and IT managers was used and investigated if and, the extent of the impact on business process performance. The study has proposed various solutions to challenges facing business PM. The results indicate that: (1) the proposed business PM has a positive influence on the implementation of a system according to business expectations; (2) by considering IT at the time of business PM, better cultural and social relationships result between the business and IT staff; (3) the proposed business PM has a positive impact on business process performance. A conclusion that can be drawn is that this proposed business PM framework can be followed to create a good business PM. Further research could explore testing this business PM framework with more BPs in different business sectors using different modelling techniques, and expanding the survey and employing large sample sizes using more complex statistical methods, such as structural equation model.
Experiment DataSets by Murtadha Talib Neamah Al-Sharuee
	Datasets
Application and Scholarship
Information on postgraduate research degree application and scholarship can be found from research services, La Trobe University.