hipe-eval / hipe-2022-data Goto Github PK

Encountered issue while NEL data processing in file HIPE-2022-v2.1-newseye-dev-de.tsv:

lines 32928-32931

Haa¬	B-PER	_	O	_	_	B-ORG	Q1405350	_	NoSpaceAfter|EndOfLine
senstein	I-PER	_	O	_	_	I-ORG	Q56322697	_	_
&	O	_	O	_	_	I-ORG	Q56322697	_	_
Vogler	O	_	O	_	_	I-ORG	Q56322697	_	_

The Qid covers the correct entity of type ORG. "Haasenstein & Vogler" is considered to be nested entity, that has a person type included. From the definition of nested entities, the smaller entity should be the one nested, thus:

Haa¬	B-ORG	_	O	_	_	B-PER	Q1405350	_	NoSpaceAfter|EndOfLine
senstein	I-ORG	_	O	_	_	I-PER	Q56322697	_	_
&	I-ORG	_	O	_	_	O	Q56322697	_	_
Vogler	I-ORG	_	O	_	_	O	Q56322697	_	_

Later on, in the same file (lines 33541-33543), the correct annotation is used:

Haasenstein	B-ORG	_	O	_	_	B-PER	Q56322697	_	_
&	I-ORG	_	O	_	_	O	Q56322697	_	_
Vogler	I-ORG	_	O	_	_	O	Q56322697	_	_

Hi,

Many thanks for the repository. I am looking for the wikidata dump file you used for the annotation. Is it the same as in HIPE 2020 (https://files.ifi.uzh.ch/cl/siclemat/impresso/clef-hipe-2020/)?

Wondered if there is a link to wikidata file (or entity catalogue) with version specified?
Is the same wikidata dump version used for HIPE2020 and topRes19th?

Best regards,
A

Hi,

during review of adding HIPE-2022 dataset into Flair, we just found that some of the listed entites do not exist in the actual dataset.

These entities are: ALIEN, OTHER, FICTION.

Could you please clarify what happened to these entites? Will they be added later (or will they appear in the final test dataset).

Many thanks,

Stefan

Hello!

I've noticed a possible missing entity type in COARSE-METO in HIPE-2022-v2.1-hipe2020-train-fr.tsv, where M. Théodore Reinach should (possibly) be a pers.ind (line 2,141-2,150):

M	O	O	O	O	B-comp.title	O	_	_	NoSpaceAfter
.	O	O	O	O	I-comp.title	O	_	_	_
Théodore	O	O	O	O	B-comp.name	O	_	_	_
Reinach	O	O	O	O	I-comp.name	O	_	_	NoSpaceAfter
,	O	O	O	O	O	O	_	_	_
député	O	O	O	O	B-comp.function	O	_	_	_
radical	O	O	O	O	I-comp.function	O	_	_	_
de	O	O	O	O	I-comp.function	O	_	_	_
la	O	O	O	O	I-comp.function	O	_	_	EndOfLine
Savoie	B-loc	O	B-loc.adm.reg	O	I-comp.function	O	Q12745	_	NoSpaceAfter

Due to several evaluation processes on my side, I'll be checking more in depth other annotated files also, and open an issue for each (if any).

Masked test files are sometimes called ...-test-allmasked-... (e.g. in ajmc) and sometimes ...-test_allmasked-... (notice the underscore). This should be harmonized.

Hi,

I've just written some testcases for reading the v2.0 version of the corpus in Flair, and it seems that there are some issues for AJMC:

• HIPE-2022-v2.0-ajmc-train-de.tsv: In line 16.537 the token ἄνδοα starts with a leading whitespace (very minor issue). Leading spaces also appear in other AJMC splits.
• HIPE-2022-v2.0-ajmc-train-en.tsv: Two "empty" tokens are in the dataset at line 5.157 and 5.645. Those tokens should be removed or replaces with a non-whitespace.
• HIPE-2022-v2.0-ajmc-dev-en.tsv: Line 5.660 unfortunately has two tokens (separated with whitespace): περάνας sa

Would be awesome if this could be fixed in the next release(s), I'm going to catch these issues in Flair for now :)