leondz / emerging_entities_17 Goto Github PK

Is there any metadata associated with the raw text (e.g. tweet ID, timestamp)? I could only find the raw text data.

In line 1674 (1-based) of the file emerging.dev.conll, the label I-creative-work follows the label I-person.
I have no idea what this sentence is talking about and thus no idea what the correct label sequence would actually be, but I think for this corpus, I labels of two different types should not follow each other.

The training and dev files have file extension .conll so one assumes the correct test file to use would also be the one with extension .conll
However, it looks as if the file emerging.test.conll really contains the up to three different labels from human annotators while the emerging.test.annotated file contains the actual consolidated test labels?

BTW, there are still some problems in the test file, e.g. in line 726 where "KEINE BREMSEN" is annotated as person but creative work (one of the two labels in .conll) seems correct.
Or line 1489 where "Bellerin v walker" is annotated as "B-person O B-person" instead of "B-person I-person I-person"