CoNLL-2016 Shared Task Results

The formal detailed description of the task and the analysis of the results will be available after the CoNLL workshop. The results across all three datasets can be downloaded in the excel format: discourse parsing task | sense classification subtask

The results in the tsv format for each task, language, and dataset can also be downloaded individually:

The best performing system for each task and each subcomponent based on the blind test set F1 score is as follows:

English discourse parsing

  • Overall: Stephan Oepen, Jonathon Read, Tatjana Scheffler, Uladzimir Sidarenka, Manfred Stede, Erik Velldal and Lilja Øvrelid from University of Oslo, University of Potsdam, Teesside University
  • Argument extraction subcomponent: Evgeny Stepanov and Giuseppe Riccardi from University of Trento
  • Connective extraction subcomponent: Zhongyi Li from Shanghai Jiao Tong University
  • Explicit relations only : same as overall
  • Non-Explicit relations only : same as overall

Chinese discourse parsing

  • Overall: Xiaomian Kang, Long Zhou, and Haoran Li from Institute of Automation, Chinese Academy of Sciences
  • Argument extraction subcomponent: Lan Man and Wang Jianxiang from East China Normal University
  • Connective extraction subcomponent: Fang Kong, Sheng Li, and Muhua Zhu from Soochow University
  • Explicit relations only: Lan Man and Wang Jianxiang from East China Normal University
  • Non-explicit relations only: same as overall

English sense classification

  • Overall: Todor Mihaylov from Heidelberg University
  • Explicit relations only: Rohit Jain from IIIT-Hyderabad
  • Non-explicit relations only: Te Rutherford and Nianwen Xue from Brandeis University

Chinese sense classification

  • Overall: Gregor Weiss from University of Ljubljana
  • Explicit relations only: Niko Schenk, Christian Chiarcos, and Samuel Rönnqvist from Goethe University Frankfurt
  • Non-explicit relations only: same as overall