University of Thessaloniki and NCBI win the BioASQ prizes on large scale biomedical semantic indexing (task 1a)

In task 1a of the first BioASQ challenge, participants were asked to assign MeSH terms to MEDLINE articles, as they appear online. They were then gradually evaluated as the correct indexing was provided by the MEDLINE curators. The evaluation took place every week for 18 weeks, divided into three batches (6 weeks each). The whole effort was kindly supported by the Indexing Initiative team of NLM and in particular Lan Aronson and Jim Mork, both of whom we would like to thank.
 
Competition in this task was great, with several teams participating from various parts of the world. In addition to the competitors, we provided a number of baseline systems, two of which were particularly hard to beat, as they are based on the MTI system of NLM, i.e. the system that helps the MEDLINE curators in their indexing task!
 
The teams that topped the competition for each batch were:
 
Zhiyong Lu
First batch - first place: NCBI team (Yuqing Mao and Zhiyong Lu)
First batch - second place: AUTH-ATYPON team (Grigorios Tsoumakas, Manos Laliotis, Nikos Markantonatos and Ioannis Vlahavas)
 
Grigorios TsoumakasIoannis Vlahavas
Second batch - first place: AUTH-ATYPON team (Grigorios Tsoumakas, Manos Laliotis, Nikos Markantonatos and Ioannis Vlahavas)
Second batch - second place: NCBI team (Yuqing Mao and Zhiyong Lu)
 
Third batch - first place: AUTH-ATYPON team (Grigorios Tsoumakas, Manos Laliotis, Nikos Markantonatos and Ioannis Vlahavas)
Third batch - second place: NCBI team (Yuqing Mao and Zhiyong Lu)
 
It is worth noting that the AUTH-ATYPON team was the only one that managed to almost consistently outperform the MTI system after week 3!
 
We would like to thank and congratulate the winners, who will soon receive their awards and prizes.
 
If you want to find out how they did it, have a look at the workshop papers on the following page: http://www.bioasq.org/workshop/schedule.
 
Also, as a reminder, the tests are still running every week, in an out-of-competition fashion. Teams who would like to participate in the semantic indexing task of the second BioASQ challenge are strongly encouraged to take advantage of these tests and prepare their systems. More information is available on the BioASQ participants area: http://bioasq.lip6.fr/.