CV: A very short CV adapted from my ABET CV.

Overview of my research: I am interested in deriving intelligence from corpora using text mining, information extraction, natural language processing, machine learning and information retrieval approaches. My work has been applied in distance learning student performance evaluation, representation of research expertise for personalized uMining, finding similar people from the web using a personal web site as a search query, personzlied query refinement, etc. The following is a list of my current projects. (Current as of June 20, 2013)

This framework utilizes multiple representative articles provided by a user as positive samples to represent a complex information need without the user composing any search query. The system learns from the user samples and ranks all documents in a document base (such as a digital library), based on their relevance to the information need representing the sample documents provided by the user, using a semi-supervised Positive and Unlabeled Learning (PU Learning) approach. To achieve a high level of learning performance even with only very few positive samples, the system utilizes under-sampling, which is especially beneficial when relevant documents similar to the samples are not evenly distributed in the document base.

This project uses the user’s prior search sessions to model his or her evolving task-based search interests with long- and short-term, and positive and negative descriptors. To reduce the noise in the dataset, the clicked pages in the user’s search sessions are represented using associated social tags to form a pseudo user representation, from where the descriptors in the user ’s profile are derived.

This project proposes a query enhancement mechanism that augments a user’s queries by leveraging the user’s query log, which provides more useful context for the user’s interests and hence reduces the ambiguity in the inferred user ’s intent.

This project trains a support vector machine classifier to label digital library resources by subject and reading level automatically.

This project utilizes semantic and linguistic content categorization which will facilitate improved access methods for digital library resources.