computer science

OPENS DOORS

Information Retrieval Research Group

 

Currently, the following members of academic staff are associated with this group :-

Andrew Trotman, Richard O'Keefe

 

The work of the Group

There is more electronic information available than we could possibly process in a lifetime, so we must rely on search engines and collaborative filtering tools to process and sift data in an effort to answer our questions.

Since knowledge is power, the value of these tools indisputable. A traditional search engine returns a list of links to documents: our group is working in the newly developing area of Focused Retrieval (FR). In an FR approach, not only does the search engine return a list of links to documents, it also identifies relevant sections within those documents. This provides the user with a valuable tool to support, guide, and focus search queries. Important challenges in FR are: how quickly can we identify relevant sections, and how should those results be ranked? This work is carried out as part of a large-scale international search engine evaluation forum called INEX which we co-chair.

Our Information Retrieval Research Group has interests in many aspects of Information Retrieval and Data Mining with a particular focus on user behavior and requirements. Returning results to a query beyond the reading level of the user is unhelpful, but how do you deduce the reading level of a specific user, and how can the legibility, readability, and academic reading level of a web page be determined? We have been working with a local high school to build a search engine that provides ranking not that user’s comprehension level.

The Architecture of a modern computer is quite different from that of only 5 years ago. We are seeing multi-core and hyper-threading on the desktop and it is not at all clear how to take advantage of this new technology. We are exploring fundamental aspects of what a computer is, and how to manage resources, so that the search engines of the future will be both fast and efficient. Our work on hard-disk performance has already led to new compression schemes. Collaborative Filtering is a branch of data mining in which the computer makes recommendations to users. This technology is well established as movie, book, and music recommenders, but is it also seen in traditional search engines with “more like this” functionality. Our group is interested in large scale collaborative filtering - millions of users, millions of items.

We are also interested in geo-temporal collaborative filtering (after all, recommending a coffee shop in Auckland is of little use to an Otago student). If search engine technologies, data mining, or any sort of information discovery (quantitative or qualitative) appeals to you, call by and have a chat with us - we’re happy to hear from you.