SentiMP-En / README.md
emcamara's picture
Update README.md
5a3c3b0 verified
|
raw
history blame
3.4 kB
metadata
license: cc-by-sa-4.0
task_categories:
  - text-classification
language:
  - es
  - en
  - el
tags:
  - code
size_categories:
  - 1K<n<10K

SentiMP Dataset

The SentiMP Dataset is a multilingual sentiment analysis dataset based on tweets written by members of parliament in Greece, Spain and United Kingdom in 2021. It has been developed collaboratively by the Andalusian Research Institute in Data Science and Computational Intelligence (DaSCI) research group from the University of Granada, the SINAI research group from the University of Jaén and the Cardiff NLP research group from the University of Cardiff.

DaSCI UJAEN Cardiff

Dataset details

The dataset containst 1500 tweets in three different languages: Greek (500 tweets), Spanish (500 tweets) and English (500 tweets). For each tweet we provide the following information:

  • full_text: Which containts the content of the tweet.
  • fold: Proposed partitions {0,1,2,3,4} in 5 folds for 5 fold cross-validation.
  • label_i : Annotator's i label (i in {1,2,3} for English and Greek and i in {1,2,3,4,5} for Spanish). It takes values in {-1,0,1}.
  • majority_vote: The result after applying the majority vote strategy to the annotators' partial labelling. When there is a tie we use the label "TIE". It takes values in {-1,0,1,TIE}.
  • tie_break: We use this column to break ties in cases where there is a tie. Therefore, it is only completed when TIE appears in the majority_vote column. It takes values in {-1,0,1}.
  • gold_label: It represents the final label. It is a combination between the majority_vote abd the tie_break columns. It takes values in {-1,0,1}.

Downloads

You can find these files in the following repositories:

Citation

If you use this dataset, please cite:

Contact

Nuria Rodríguez Barroso - rbnuria@ugr.es

Acknowledgements

This dataset has partially supported by the R&D&I grant PID2020-116118GA-I00 funded by MCIN/ AEI/10.13039/501100011033.

Shield: CC BY-SA 4.0

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

CC BY-SA 4.0