Dataset Card for HuSST

Dataset Summary

This is the dataset card for the Hungarian version of the Stanford Sentiment Treebank. This dataset which is also part of the Hungarian Language Understanding Evaluation Benchmark Kit HuLU. The corpus was created by translating and re-annotating the original SST (Roemmele et al., 2011).

Supported Tasks and Leaderboards

'sentiment classification'

'sentiment scoring'

Language

The BCP-47 code for Hungarian, the only represented language in this dataset, is hu-HU.

Dataset Structure

Data Instances

For each instance, there is an id, a sentence and a sentiment label.

An example:

{
"Sent_id": "dev_0",
"Sent": "Nos, a Jason elment Manhattanbe és a Pokolba kapcsán, azt hiszem, az elkerülhetetlen folytatások ötletlistájáról kihúzhatunk egy űrállomást 2455-ben (hé, ne lődd   le a poént).",
"Label": "neutral"
 }

Data Fields

• Sent_id: unique id of the instances;

• Sent: the sentence, translation of an instance of the SST dataset;

• Label: "negative", "neutral", or "positive".

Data Splits

HuSST has 3 splits: train, validation and test.

Dataset split	Number of instances in the split
train	9344
validation	1168
test	1168

The test data is distributed without the labels. To evaluate your model, please contact us, or check HuLU's website for an automatic evaluation (this feature is under construction at the moment).

Dataset Creation

Source Data

Initial Data Collection and Normalization

The data is a translation of the content of the SST dataset (only the whole sentences were used). Each sentence was translated by a human translator. Each translation was manually checked and further refined by another annotator.

Annotations

Annotation process

The translated sentences were annotated by three human annotators with one of the following labels: negative, neutral and positive. Each sentence was then curated by a fourth annotator (the 'curator'). The final label is the decision of the curator based on the three labels of the annotators.

Who are the annotators?

The translators were native Hungarian speakers with English proficiency. The annotators were university students with some linguistic background.

Additional Information

Licensing Information

Citation Information

If you use this resource or any part of its documentation, please refer to:

Ligeti-Nagy, N., Ferenczi, G., Héja, E., Jelencsik-Mátyus, K., Laki, L. J., Vadász, N., Yang, Z. Gy. and Vadász, T. (2022) HuLU: magyar nyelvű benchmark adatbázis kiépítése a neurális nyelvmodellek kiértékelése céljából [HuLU: Hungarian benchmark dataset to evaluate neural language models]. XVIII. Magyar Számítógépes Nyelvészeti Konferencia. pp. 431–446.

@inproceedings{ligetinagy2022hulu,
  title={HuLU: magyar nyelvű benchmark adatbázis kiépítése a neurális nyelvmodellek kiértékelése céljából},
  author={Ligeti-Nagy, N. and Ferenczi, G. and Héja, E. and Jelencsik-Mátyus, K. and Laki, L. J. and Vadász, N. and Yang, Z. Gy. and Vadász, T.},
  booktitle={XVIII. Magyar Számítógépes Nyelvészeti Konferencia},
  year={2022},
  pages = {431--446}
}

and to:

Socher et al. (2013), Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank. In: Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing. 1631--1642.

@inproceedings{socher-etal-2013-recursive,
    title = "Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank",
    author = "Socher, Richard  and
      Perelygin, Alex  and
      Wu, Jean  and
      Chuang, Jason  and
      Manning, Christopher D.  and
      Ng, Andrew  and
      Potts, Christopher",
    booktitle = "Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing",
    month = oct,
    year = "2013",
    address = "Seattle, Washington, USA",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/D13-1170",
    pages = "1631--1642",
}

Contributions

Thanks to lnnoemi for adding this dataset.