It includes 229 texts from the Ukrainian Brown corpus with 217,381 tokens and 6,751 annotated Named Entities.

Data description

The corpus of annotated data can be found in the data folder. It includes:

• 229 texts
• 217,381 tokens
• 6,751 Named Entities:
• PERS - 4,060
• LOC - 1,442
• ORG - 649
• MISC - 600

The source of the data is the open Brown corpus of Ukrainian texts. For each processed text, there are two files:

• the one with extension tok.txt includes the tokenized version of the text (tokenization was done with the following rules)
• the one with extension tok.ann includes NER annotations for this text in Brat Standoff Format (each line of the file includes 3 notations divided by a tab: the number of annotation, the start and end indices (with a space in between) in the tokenized text, and the named entity.

Annotation was performed by two annotators for each text with the following rules. The annotators’ disagreement was resolved by a third annotator.

For training and validation models, we recommend using the standard division into dev and test sets. 

Licensing

This data is available for usage under the licence "Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License."