These are my graduate students. I'm proud of them. I try to keep them very busy.
Ian
Shaoqun Wu is seeking automatic ways to construct lexical acquisition systems based on a collection of text that teachers or learners provide. She is looking at how to identify important and typical lexical items for language learning from a given corpus by using human language, artificial intelligence, and information retrieval technologies. She is exploring the pedagogical value of these lexical items and using them to construct a computer assisted learning system that facilitates lexical acquisition.
You can play with some of the language games she has developed in our Flexible Language Acquisition (FLAX) project.
Michael Walmsley is designing second language learning activities that utilise time spent reading online to refresh and extend your language skills. He is aiming to combine research in second language acquisition and computational linguistics to develop systems for automatically creating vocabulary lists and second-language reading texts from document collections. The lists and texts must be tailored specifically to the interests, goals and abilities of individual learners.
Software for Japanese and Spanish language learners will be developed to evaluate the systems and activities in longitudinal user studies.
Anna Huang is interested in document clustering. She is investigating algorithms for interactive clustering, and the scatter-gather approach for facilitating the browsing of digital library collections. She is also interested in improving the efficiency of clustering, particularly incremental clustering, by using keyphrases extracted from free texts as descriptors of the document content. She would like to investigate the potential theoretical and practical performance improvements due to the use of automatically extracted keyphrases as the basic document representation.
Craig Schock is interested in improving the development and maintenance of medium and large-scale software systems. Software systems are subject to constant change pressures and because of this, they must remain flexible. The evolution of a software system is heavily dependent on its structure. The field of network theory has been used to analyze complex systems in a variety of different fields and has shown that specific structural characteristics contribute greatly to the evolvability of the system.
Craig's goal is to validate network theory as a viable mechanism for evaluating the evolvability of software systems.
Veronica Liesaputra is working on our Realistic Book project for her PhD. Her goals are (a) to produce a three-dimensional book model that is natural and interactive; (b) represent the Wikipedia as a huge three-dimensional book; and (c) produce a three-dimensional visualization of a personal digital library.
She has produced a lightweight implementation of realistic books that provides a quick, easy-to-use, and responsive page-turning mechanism, and combines the ability to include hyperlinks and animated media. You can see sample books or make your own from a HTML or PDF file.
Daniel McEnnis's task is simple to describe: separate the world's music into music they will like and music they will not like. The complexity of music recommendation systems has increased rapidly in recent years, drawing upon different sources of information: content analysis, web-mining, social tagging, etc. However, the tools to scientifically evaluate such integrated systems are not readily available; nor are the base algorithms available.
He has produced the Relational Analysis Toolkit, which provides a large library of graph-analysis routines within a framework that seamlessly integrates both flat and graph-based algorithms.
Olena Medelyan is interested in natural language processing techniques applied to information retrieval, information extraction and text mining. Her PhD research is on automatic indexing with controlled vocabularies.
The central hypothesis of her thesis is that
Olena has produced the latest version of the KEA algorithm for keyphrase extraction.
The goal of David Milne's PhD is to develop a framework for automatically generating highly accurate, concise thesauri for domain specific document collections. He hypothesizes that (a) for any document set, automatically-created thesauri suit users' searching needs better than manually defined ones, and (b) appropriately crafted thesauri can be integrated into the searching process to improve retrieval without placing an undesirable cognitive load on the user.
David's Koru is an example of interactive query expansion: a highly responsive web application based on AJAX. He has also produced the Wikipedia Miner toolkit.
Rob Akscyn is proposing a radical shift in software tools for knowledge workers: from highly-fragmented applications using flat data models with pull-down intensive interfaces to lattice-structured hypermedia, rich with knowledge schemas, in concert with extreme direct manipulation user interfaces, digital library search technology, and personal agent assistants.
The central hypothesis of his thesis is that the productivity of knowledge work will be significantly improved by transitioning from current computer-based tools to this new knowledge work paradigm.
Kathryn Hempstalk's PhD topic is Continuous Typist Recognition Using Machine Learning. She is trying to figure out to what extent typists can be identified by the patterns they exhibit while typing at keyboard. She is focused on a setting where the user is continuously monitored, rather than password hardening where only the login identification is monitored.
Kathryn produced the Digital Invisible Ink Toolkit, a Java steganography tool that can hide any sort of file inside a digital image, and the Digital Image Resizer Toy, an implementation of Avidan and Shamir's algorithm for "content aware" image resizing.
