RONEC, at version 2.0, holds 12330 sentences with over 0.5M tokens, annotated with 15 classes, to a total of 80.283 distinctly annotated entities.

It is more than twice the size of the previous version. Quality has been significantly increased as verything has been annotated from scratch to confrom to a single standard. For the older version 1, please see the ronec_v1 folder.

Update - 16 Nov 2021 : RONEC v2 is available on HuggingFace's datasets library. To use, simply pip install datasets and run the following:

from datasets import load_dataset
ronec = load_dataset("ronec")

the ronec dict contains the 'train', 'validation' and 'test' splits. Access elements like ronec["train"][1000] - please see the simple format below. You can also explore the dataset here.

The corpus has the following classes and distribution in the train/valid/test splits:

The data is available as a train/valid/test split in data, as json files. Each file is a list of instances, where an instance is a dictionary that contains the following keys:

{
  "id": 10454,
  "tokens": ["Pentru", "a", "vizita", "locația", "care", "va", "fi", "pusă", "la", "dispoziția", "reprezentanților", "consiliilor", "județene", ",", "o", "delegație", "a", "U.N.C.J.R.", ",", "din", "care", "a", "făcut", "parte", "și", "dl", "Constantin", "Ostaficiuc", ",", "președintele", "C.J.T.", ",", "a", "fost", "prezentă", "la", "Bruxelles", ",", "între", "1-3", "martie", "."], 
  "ner_tags": ["O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-PERSON", "O", "O", "O", "O", "O", "O", "B-ORG", "O", "O", "O", "O", "O", "O", "O", "B-PERSON", "I-PERSON", "I-PERSON", "I-PERSON", "I-PERSON", "B-ORG", "O", "O", "O", "O", "O", "B-GPE", "O", "B-PERIOD", "I-PERIOD", "I-PERIOD", "O"], 
  "ner_ids": [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 3, 0, 0, 0, 0, 0, 0, 0, 1, 2, 2, 2, 2, 3, 0, 0, 0, 0, 0, 5, 0, 19, 20, 20, 0], 
  "space_after": [true, true, true, true, true, true, true, true, true, true, true, true, false, true, true, true, true, false, true, true, true, true, true, true, true, true, true, false, true, true, false, true, true, true, true, true, false, true, true, true, false, false]
}

The tokens are the words of the sentence. The ner_tags are the string tags assigned to each token, following the BIO2 format. For example, the span "între", "1-3", "martie" has three tokens, but is a single class PERIOD, marked as "B-PERIOD", "I-PERIOD", "I-PERIOD". The ner_ids are the integer encoding of each tag, to be compatible with the standard and to be quickly used for model training. Note that each B-starting tag is odd, and each I-starting tag is even. The space_after is used to help if there is a need to detokenize the dataset. A true value means that there is a space after the token on that respective position.

• Stefan Daniel Dumitrescu

Big thanks to termene.ro for carefully annotating the full expanded dataset. RONEC v2 would not have seen the light of day without them!

Please consider citing the following paper as a thank you to the authors of the RONEC, even if it describes v1 of the corpus and you are using v2:

Dumitrescu, Stefan Daniel, and Andrei-Marius Avram. "Introducing RONEC--the Romanian Named Entity Corpus." arXiv preprint arXiv:1909.01247 (2019).

or in .bibtex format:

@article{dumitrescu2019introducing,
  title={Introducing RONEC--the Romanian Named Entity Corpus},
  author={Dumitrescu, Stefan Daniel and Avram, Andrei-Marius},
  journal={arXiv preprint arXiv:1909.01247},
  year={2019}
}