SemEval 2010 Task 10

Available Downloads

Trial data:
The trial data for the FN-version of the task is available from the SemEval 2010 page. You can also take a quick look at the data format for the FN-version of the task here.

Training data:
The training data are now also available from the SemEval 2010 page. The training set is annotated with:

• gold standard semantic argument structure (FrameNet and PropBank style)
• linking information for null instantiations
• co-reference chains (for evaluation purposes)

The data were independently annotated by two experienced annotators (in FrameNet format) and then adjudicated. The training set is available both with FrameNet- and PropBank-style annotations (in SalsaTiger-XML and a CoNLL-like format, respectively). Participants can choose with which data set they want to work (the evaluation is done separately). The PropBank annotations were derived semi-automatically from the FrameNet annotations. To find out more about the two different annotations read our Info for Participants and see:

• the README for the FrameNet-style annotation
• the README for the PropBank-style annotation

Note: both README files are part of the data distribution downloadable from the SemEval 2010 shared task website.

Scorer:
The evaluation software for the task is still being tested and will be available soon.

Contact:
We invite you to join our Google group for Task-10 to receive further information and for discussion and questions.