Extracting events of biomedical relevance from text

Date

August 21, 2012

Speaker

Sophia Ananiadou

Affiliation

University of Manchester

Overview

The understanding of e.g. multifactorial diseases requires the discovery of entities and their relationships from several domains: proteomics, metabolomics, toxicology, etc. The evidence to generate hypotheses for comprehensive diagnostics, interventions, treatments, etc is hidden in text. In addition, the type of evidence needed is complex, requiring techniques beyond statistical keyword search mechanisms. Event extraction techniques can represent such complex associations which can trace cause and effect across multiple levels of biological organisation while also providing other contextual information such as negation, speculation, etc. Event extraction results have been incorporated into recent text search engines to support advanced tasks such as pathway curation. Curation environments such as Argo, faciliate the task of annotation.

Speakers

Sophia Ananiadou

Professor Sophia Ananiadou is director of the National Centre for Text Mining (NaCTeM), which was established in 2004 by JISC, BBSRC and EPSRC to provide leadership in text mining for the UK academic community, with particular focus on biological and biomedical science. She is also Professor of Computer Science in the School of Computer Science, University of Manchester. She is leading a team of 12 text miners providing scalable text mining services all accessible via the NaCTeM portal to the community. Her research includes the development of a large-scale terminology resource for biology (BioLexicon, FT6-028099), data integration using text mining (ONDEX, BB/F006012/1), automated biological event extraction for drug discovery (BB/G013160/1 and AZ), text mining for chemistry and systematic reviews (JISC). She is leading the text mining work of UKPMC funded by Wellcome Trust, providing text mining advanced search functionalities for biomedicine. She has authored over 230 publications in journals, conferences and books.