Frequently asked questions

Can I only participate in Subtask 1 (Identifying Number of Statements)?

Yes, you can participate in both subtasks or only in Subtask 1. If you only want to participate in Subtask 1, please add an empty statement_spans column to your prediction file. In this case, the statement spans will be considered as null.

Can I only participate in Subtask 2 (Identifying Segments of Statements)?

No, Subtask 2 is dependent on Subtask 1. You cannot only participate in Subtask 2. But, it is possible to submit your results only to Subtask 1.

What tools are allowed to solve the task?

All technological solutions are welcome. To foster scientific reproducibility, we encourage you to use open-source software and data solutions whenever possible. Nevertheless, we don’t prohibit using any tool, software, or model.

Why do you split sentences like Eine Angel ist eine Schnur mit einem Haken into two statements?

Our statements focus mostly on chunks of information that need to be understood by the readers. From a content perspective, a fishing rod always consists of a string and a hook, so splitting these into two statements makes no sense. However, the reader still has to parse two separate statements independently: 1) there is a string, and 2) there is a hook. Therefore, we decided to split the sentence into two statements.

How did you annotate your data?

We tried to describe our process in detail in our Annotation Guidelines

Why are there empty tokens? Can we delete them?

The data from the Hurraki homepage is quite noisy. We decided to keep the noise to reflect upon noisy real-world data. These empty tokens are not annotated in the statement spans. So, you can, of course, remove them in your (pre-)processing. However, make sure that the span annotations still use the original numbering.