Den 21. februar holdt Dr. Theodora Tsikrika oplæg om ImageCLEF, en årlig evaluering af visuelle søgesystemer, som har til formål at støtte flersprogede brugeres adgang til den hastigt voksende mængde af visuelle informationer.
Bedre adgang til visuel information
ImageCLEF blev
introduceret i 2003 som en del af Cross Language Evaluation Forum
(CLEF). Evalueringskampagnens primære formål er at støtte
udviklingen af feltet visuelle medier, herunder analyse,
indeksering, klassifikation og genfinding, ved at udvikle den
nødvendige infrastruktur til vurdering af visuelle
informationssøgesystemer.
"Evaluation campaigns have been widely credited with contributing
tremendously to the advancement of information access systems by
providing the infrastructure and resources that support researchers
in the development of new approaches, and promoting the exchange of
ideas," fortæller Theodora Tsikrika. "Measuring the impact of such
benchmarking activities is crucial for assessing which of their
aspects have been successful, and thus obtain guidance for the
development of improved evaluation methodologies and information
access systems. Given that their contribution to the field is
mainly indicated by the research that would otherwise not have been
possible, it is reasonable to consider that their success can be
measured, to some extent, by the scientific impact of the research
they foster. The scientific impact of research is commonly measured
by its scholarly impact, such as the publications derived from it
and the citations they receive."
Nye domæner
I sit foredrag præsenterede Theodora Tsikrika en oversigt
over de ImageCLEF evalueringsopgaver, som er toneangivende for
genfindingen af forskellige billeder, såsom generelle fotografiske
og medicinske billeder, samt domæne-specifikke opgaver. Hun
fortalte om de ressourcer, der er blevet bygget i forbindelse med
evalueringsaktiviteterne, hvordan man undersøger deres pålidelighed
og genanvendelighed og om en foreløbig undersøgelse af ImageCLEFs
videnskabelige indvirkning.
Efterfølgende fortalte Tsikrika om baggrunden for sit besøg: "My
visit here is funded by ELIAS [Evaluation
Information Access Systems] and the aim is to get some knowledge on
how bibliometric analysis is performed and also, because this is in
your domain, we want to assess the impact of European projects or
domains of research. Previously, the focus has been on assessing
the impact of individuals. So for them it's a new domain, and we
want to be able to benefit from each other. One of the aspects we
are trying to assess is the scientific impact of the evaluation
campaign, which can be measured through bibliometrics. In that
context, I came here to seek guidance and advice from Birger
Larsen. Furthermore, your school has just joined the Promise project. Also,
there are people here who are doing Information Retrieval research
and I want to get to know them and exchange ideas about possible
collaborations for the future."
Den underliggende usikkerhed
Ifølge Tsikrika er der store udfordringer forbundet med
det arbejde, hun udfører, men også mange fordele: "The biggest
challenge in working with image retrieval is the underlying
uncertainty of it all, because image retrieval has developed and
progressed a lot during the last ten years, though not to the same
extent as textual retrieval. And the challenge now is how to be
able to combine evidence from the different media and to do that
effectively. And to be able to scale up to larger scale data,
and to do that within the different domains where the impact is
going to be substantial, like the medical domain or scientific
multimedia data. What I always find really interesting about my
work is first of all personalization - how to be able to deliver
retrieval support to individuals - and secondly, how to be able to
combine the different evidence."
Af Helle
Saabye