CS Team Wins SemEval-2015 Semantic Textual Similarity Competition
Posted on 2015-08-04
The team, named NeRoSim, consisted of Ph.D. students Rajendra Banjade, Dipesh Gautam, Nabin Maharjan, and Nobal Niraula, along with postdoctoral fellows Dr. Mihai Lintean and Dr. Dan Stefanescu. They participated in two tasks: the English semantic textual similarity (STS) task and the interpretable STS task, achieving winning results in both. For the English STS task, the NeRoSim team ranked 10th out of 74 system runs but there was no significant difference among the top 10 system runs, as stated by the organizers' report. For the interpretable STS, the team ranked 1st for the 'Headlines' data subset and 2nd for the 'Images' data subset, which was the best performance overall (for both datasets) of any team.
Competitors at the event included teams from the University of Colorado at Boulder, Samsung Research America, Hong Kong University of Science and Technology, East China Normal University, Japan Advanced Institute of Science and Technology, Australian eHealth Research Centre, Budapest University, and the University of Trento.
Detailed information can be found in the organizers' full report, available at http://alt.qcri.org/semeval2015/cdrom/pdf/SemEval045.pdf.