XML of Scientific Article

<algorithms version="110505">
<algorithm name="SectLabel" version="110505">
<variant no="0" confidence="0.001276">
<title confidence="0.984639">
Towards Event Extraction from Full Texts on Infectious Diseases
</title>
<author confidence="0.8597195">
Sampo Pyysalo∗ Tomoko Ohta∗ Han-Cheol Cho∗ Dan Sullivan† Chunhong Mao† Bruno Sobral† Jun’ichi Tsujii∗‡§ Sophia Ananiadou‡§
</author>
<affiliation confidence="0.99501">
∗Department of Computer Science, University of Tokyo, Tokyo, Japan †Virginia Bioinformatics Institute, Virginia Tech, Blacksburg, Virginia, USA ‡School of Computer Science, University of Manchester, Manchester, UK §National Centre for Text Mining, University of Manchester, Manchester, UK
</affiliation>
<email confidence="0.985733666666667">
{smp,okap,priancho,tsujii}@is.s.u-tokyo.ac.jp {dsulliva,cmao,sobral}@vbi.vt.edu [email protected]
</email>
<sectionHeader confidence="0.993163" genericHeader="abstract">Abstract</sectionHeader>
<bodyText confidence="0.999863">
Event extraction approaches based on ex- pressive structured representations of ex- tracted information have been a significant focus of research in recent biomedical nat- ural language processing studies. How- ever, event extraction efforts have so far been limited to publication abstracts, with most studies further considering only the specific transcription factor-related subdo- main of molecular biology of the GENIA corpus. To establish the broader relevance of the event extraction approach and pro- posed methods, it is necessary to expand on these constraints. In this study, we pro- pose an adaptation of the event extraction approach to a subdomain related to infec- tious diseases and present analysis and ini- tial experiments on the feasibility of event extraction from domain full text publica- tions.
</bodyText>
<sectionHeader confidence="0.999134" genericHeader="introduction">1 Introduction</sectionHeader>
<bodyText confidence="0.999883115384615">
For most of the previous decade, biomedical In- formation Extraction (IE) efforts have focused pri- marily on tasks that allow extracted information to be represented as simple pairs of related enti- ties. This representation is applicable to many IE targets of interest, such as gene-disease associa- tions (Chun et al., 2006) and protein-protein inter- actions (N´edellec, 2005; Krallinger et al., 2007). However, it has limited applicability to advanced applications such as semantic search, Gene On- tology term annotation, and pathway extraction, tasks for which and relatively few resources or sys- tems (e.g. (Rzhetsky et al., 2004)) have been intro- duced. A number of recent studies have proposed more expressive representations of extracted in- formation, introducing resources supporting ad- vanced IE approaches (Py