An interesting question when choosing sentences from the COWLH2 corpus is: what to do with the different annotators. Perhaps for now, we could just choose the annotations which both annotators put in (to reduce noise). We are not data hungry in this case.
Annotator1 has annotated much more, and much more consistently it seems (at least in all most of the mismatches, the mismatch is due to a mistake by Annotator2). So perhaps just use Annotator1?
An interesting question when choosing sentences from the COWLH2 corpus is: what to do with the different annotators. Perhaps for now, we could just choose the annotations which both annotators put in (to reduce noise). We are not data hungry in this case.