Integrated and Contrastive Text and Data Mining

The amount of scientific, medical and/or business data that is available in open-access or private repositories and databases is enormous, and is increasing at an unprecedented pace. Many domains have much more data than they could possibly analyse and use to improve their understanding of the domain. Therefore, the identification of contradicting information is of particular value. This is especially dramatic in domains that have vast amounts of knowledge in heterogeneous sources (both structured e.g. databases, and semi- and un-structured, e.g. scientific literature, health records, stock reports, Web pages, etc.). Integrating and contrasting findings originated from and within heterogeneous resources is a challenge that will be the main aim of this project.

The project will aim to provide a framework for integrating results of data and text mining in order to combine or contrast findings drawn from heterogeneous sources. Data mining refers to discovering new patterns from large structured data (e.g. databases), while text mining is about discovering new information from text (unstructured data), including extraction of information and representation of semantics, contradictions and contrasts. Discovery of outliers and correlations between textual and non-text information is a very powerful hypothesis generation method. For example, entities that appear similar from the results of text mining might behave very differently under a particular set of experimental conditions or in a given business environment or for a given patient; this suggests the data is uncovering something that was previously unknown and is worthy of further investigation. At present, there are no good tools for detecting these types of interesting patterns and this project will endeavour to develop such tools.

The project can be placed in the context of biomedicine, medical/health informatics or business intelligence (applications for e-commerce).
More details on related research at http://gnode1.mib.man.ac.uk/