Programme

Wednesday 17th September


13:30-15:30 - Session 1


13:30-13:45: Opening remarks

13:45-14:20: Task 1 overview “Visual-Interactive Search and Exploration of eHealth Data”  – Tobias Schreck, University of Konstanz, Germany

14:20-14:55: Task 2 overview “Information extraction from clinical text” – Danielle Mowery, University of Pittsburgh, USA

14:55-15:30: Task 3 overview “User-centred health information retrieval” - Liadh Kelly and Lorraine Goeuriot, Dublin City University, Ireland



16:00-18:00 - Session 2 (Session Chair: Danielle Mowery, University of Pittsburgh, USA)


16:00-17:00: Keynote – Maria Kvist "Professional language as an obstacle in information retrieval from patient records."

Abstract: 

Health record systems were originally created for billing purposes, and for that reason such information is easy to retrieve. Other types of information, such as much of the clinical information, reside in the free text sections and need NLP to be extracted. However, irrespective of the task, off-the-shelf NLP-tools will encounter problems in information retrieval as the text differ quite substantially from normal language use. Patient records are written under time pressure, are rich in terminology but lack words of minor importance, are full of jargon and abbreviations in order to speed up both the writing and the reading of notes, and are rarely edited for errors. IR systems will have to handle not only semantic but also syntactic peculiarities. As an example, we found that over 60% of sentences in Swedish radiology reports lacked a predicate/verb, which of course will impair the performance of NLP-tools and IR. The professional sublanguage needs to be characterized to improve IR from patient records. I will give examples of different clinical situations at the hospital where IR systems could help to ease the health care process for me as a clinician. Also, I will discuss IR from health records from the patients’ point of view, for the emerging patient portals. A third party with interest of IR from patient records are national health institutions, craving health statistics that now to a large extent is reported manually. We have examined aspects on IR for the above health care actors, using a large corpus of over 1 million Swedish patient records. 

17:00-17:20: Task 1 team presentation

17:20-17:40: Task 2 team HPI presentation - Mariana Neves, Hasso Plattner Institute at the University of Potsdam, Germany

17:40-18:00: Task 2 team HITACHI – Nishikant Johri, Hitachi, India



Thursday 18th September


10:00-12:00 - Session 3 (Session Chair: Gareth Jones, Dublin City University, Ireland)


10:00-10:20: Task 3 team GRIUM presentation – Jian-Yun Nie, Université de Montréal, Canada

10:20-10:40: Task 3 team RePaLi presentation - Sébastien Le Maguer, INRIA, IRISA, France

10:40-11:00: Discussion session on CLEF eHealth 2014 highs and lows

11:00-11:50: Panel session. Panel Chair: Allan Hanbury (Vienna University of Technology, Austria).  Panellists: Maria Kvist (Karolinska Institutet and DSV Stockholm University, Sweden), George Paliouras (NCSR "Demokritos", Greece), Angus Roberts (University of Sheffield, UK))

11:50-12:00: Wrap-up/closing