Information theory and retrieval modeling

This research aimed to study existing information theories as well as potential new information measures that can be used for IR modeling, among other applications. I have developed a new theory, namely the Least Information Theory (LIT), and conducted several studies to evaluate its application in IR, which produced very strong empirical results as compared to classic methods derived from existing theories. One representative publication (ASIST'12 in appendix) reports on initial experimental results of text classification based on LIT. Two related articles on the application of LIT for text clustering and probabilistic retrieval modeling are currently under review to the information retrieval journal and the journal of data mining and knowledge discovery (see the work in progress section in CV). The theory has several interesting properties which I expect to study further.
 

Related Publications