I can't find the evidence files in the public data.
Posted by: niksss @ Aug. 8, 2021, 2:42 p.m.Hi niksss,
You will find the training and development data in the "Public Data" section under "Participate". Here is the link: https://competitions.codalab.org/competitions/33739#participate-get_starting_kit. Once you unzip the file, you will find this:
- trainset.txt: the annotated data. It looks like this (below are the first lines):
00001 C 9149180 3781329
00002 B 4396080 9207819 9757979 344087 361152 2099731
00003 B 1519858 11734712
- trainset.zip: XML files for each of the documents. There you fill find files with names 9149180.xml, 3781329.xml, and so on.
- devtestset.txt, devtestset.txt: the development data.
We will make the test set available in phase 2.
I hope this answers your question!
Diego
Where can I find the evidence files,0001,0002 etc ?
Posted by: niksss @ Aug. 9, 2021, 4:18 p.m.Calrifying on the text I posted earlier in this discussion, if the first line of trainset.txt has this information:
00001 C 9149180 3781329
This means that the evidence files for 00001 are 9149180.xml and 3781329.xml. These two files should be in the zipped file trainset.zip. The target label is C.
I hope it is clearer now.
Diego
Posted by: dmollaaliod @ Aug. 10, 2021, 12:38 a.m."This means that the evidence files for 00001 are 9149180.xml and 3781329.xml. These two files should be in the zipped file trainset.zip. The target label is C."
Hi, what do 00001 represent in this case? From my understanding 00001 represents the ID of some claim, but I don't know where to find the claim corresponding to this ID
Posted by: martinfunkquist @ Sept. 11, 2021, 6:44 a.m.If you mean that you would like to see additional information such as what medical question was asked and what medical evidence was found based on these documents, this information is not available for this shared task.
We are providing the same information as in the 2011 shared task, so that we can compare results. We can release more information after the shared task has finished if you want to continue further research on this.
Thanks for the response, that is what I mean.
Posted by: martinfunkquist @ Sept. 13, 2021, 6:19 p.m.