Information Retrieval without queries

The traditional paradigm of users presenting queries to represent an information need has been the dominant model in information retrieval for decades. There is a growing realisation of the limitations of this approach and a move towards realising timely retrieval of relevant information in a manner that does not require explicit querying by the user. The aims of this research are to infer information needs on the users’ behalf based on the current task or upcoming tasks or meetings; to classify these needs (informational, transactional etc.) and to represent these needs as suitable queries; to retrieve information from collections (web, desktop) and to present these results to the user. The approaches and models are evaluated for a range of task.

Design of suitable weighting schemes for graph based representations in Information Retrieval

Traditional information retrieval approaches adopt a range of techniques to represent documents and users’ queries; these include sets, bags and vectors. The use of graph representations can help overcome some of the limitations of previous approaches (e.g. term independence assumption) and can help capture more intuitively and expressively the information in a document or a query. There have been a number of recent studies into the ’correctness’ of a weighting scheme using an axiomatic framework. This research aims to adopt such a framework to graph representations, to derive a family of weighting schemes and to evaluate their performance on standard benchmarks.