README.md

metadata

annotations_creators:
  - Barcelona Supercomputing Center
language_creators:
  - Racó Català
language:
  - ca
license: cc-by-4.0
multilinguality:
  - monolingual
pretty_name: CaSERa
task_categories:
  - text-classification
task_ids: []

Dataset Card for CaSERa, the Catalan Stance and Emotions Dataset from Racó Català

Table of Contents

• Table of Contents
• Dataset Description
  • Dataset Summary
  • Supported Tasks and Leaderboards
  • Languages
• Dataset Structure
  • Data Instances
  • Data Fields
  • Data Splits
• Dataset Creation
  • Curation Rationale
  • Annotations
  • Personal and Sensitive Information
• Considerations for Using the Data
  • Social Impact of Dataset
  • Discussion of Biases
  • Other Known Limitations
• Additional Information
  • Dataset Curators
  • Licensing Information
  • Citation Information
  • Contributions

Dataset Description

• Homepage Projecte AINA
• Repository HuggingFace
• Point of Contact: Language Technologies Unit

Dataset Summary

The CaSERa dataset is a Catalan corpus from the forum Racó Català annotated with Emotions and Dynamic Stance. The dataset contains 15.782 unique sentences grouped in 10.745 pairs of sentences, paired as parent messages and replies to these messages.

Supported Tasks and Leaderboards

This dataset can be used to train models for emotion detection and dynamic stance detection tasks.

Languages

The dataset is in Catalan (ca-ES).

Dataset Structure

Each instance in the dataset is a pair of parent-reply messages annotated with the relation between the two messages (the dynamic stance). For each message there is an individual id and the emotions identified in the message.

Data Instances

{
"id_conversation": "782135", 
"id_reply": "782135_2_2", 
  "parent_text": "Alguns petits apunts que s'haurien de tenir en compte en la creació d'aquesta hipotètica grada jove: -Renúncia total i explícita a la violència, aquest per mi és el punt clau i bàsic que la directiva tindria en compte, sense renúncia no hi ha grup. -Edat de la gent: indispensable establir un mínim i un màxim d'edat per adquirir entrades de la grada jove, s'ha d'acabar amb els freaks de 40-50 anys amb esperit jove i ganes d\'animar (sona discriminatori, però crec que és òbvi que s'ha de limitar l'edat) -Seguretat bàsica: Agradi o no, s'hauria d'incrementar moltíssim la seguretat privada (és un dels punts que no agrada a la directiva). Els indesitjables de sempre no tolerarien cap grada jove sense ells. -Per tant doncs, absteniu-vos i oblideu-vos tots els hooligans potencials (n'hi ha a grapats per tot Catalunya) de crear una grada hooligan amb skins, punkis i resta d'estètiques "característiques", també seria un punt clau (confirmat per un directiu!) En definitiva, ara per ara veig força difícil i inviable la creació d'una grada jove nombrosa i contundent, s'hauria d'optar per opcions més "descafeïnades" rotllo Sang Culé o Dracs.", 
  "reply_text": "Doncs si es creés una grada jove ja et dic jo que sompliria d'skins, i els primers en entrari serien els Boixos que han expulsat del seu lloc. Tot i aìxò que hi hagin skins no vol dir que hi hagi violència, no crec que es peguessin amb els del mateix equip. Si es creés un grada jove l'ambient al camp nou seria brutal, si 200 Boixos feien bivrar el camp nou, imaginat a 4000 perones o la gent que hi capigués a la grada jove.", 
    "dynamic_stance": "Elaborate", 
  "parent_emotion": ["fear", "distrust", "anticipation"], 
  "reply_emotion": ["anger", "sadness", "fear", "distrust", "anticipation"]
}

Data Splits

The dataset does not contain splits.

Dataset Creation

Curation Rationale

We created this corpus to contribute to the development of language models in Catalan, a low-resource language.

Source Data

The data was collected using the messages of the forum Racó Català by the Barcelona Supercomputing Center.

Initial Data Collection and Normalization

The data was collected selecting random messages from 13 of the thematic sections in Racó Català that had at least 15 tokens and at most 300. Then, we kept the messages that had at least one replying message with the same length requirement. We got a maximum of 3 replying messages per parent message.

Who are the source language producers?

The source language producers are users of Racó Català.

Annotations

• Emotions are annotated in a multi-label fashion. The labels can be: Anger, Anticipation, Disgust, Fear, Joy, Sadness, Surprise, Distrust, and No emotion.

• Dynamic stance is annotated per pair. The labels can be: Agree, Disagree, Elaborate, Query, Neutral, Unrelated, NA.

Annotation process

• For emotions there were 3 annotators. The gold labels are an aggregation of all the labels annotated by the 3. The IAA calculated with Fleiss' Kappa per label was, on average, 38.73.

• For dynamic stance there were 4 annotators. If at least 3 of the annotators disagreed, a fifth annotator chose the gold label. The overall Fleiss' Kappa between the 4 annotators was 57.63, and the average Fleiss' Kappa of the annotators with the gold labels is 85.98.

Who are the annotators?

All the annotators are native speakers of Catalan.

Personal and Sensitive Information

The data was annonymised to remove user names and emails, which were changed to random Catalan names. The mentions to the chat itself have also been changed.

Considerations for Using the Data

Social Impact of Dataset

We hope this corpus contributes to the development of language models in Catalan, a low-resource language.

Discussion of Biases

We are aware that, since the data comes from a public forum, this will contain biases, hate speech and toxic content. We have not applied any steps to reduce their impact.

Other Known Limitations

Additional Information

Dataset Curators

Language Technologies Unit (LangTech) at the Barcelona Supercomputing Center.

This work/research has been promoted and financed by the Government of Catalonia through the Aina project.

Licensing Information

Creative Commons Attribution 4.0 International.

Citation Information

@inproceedings{figueras-etal-2023-dynamic,
    title = "Dynamic Stance: Modeling Discussions by Labeling the Interactions",
    author = "Figueras, Blanca  and
      Baucells, Irene  and
      Caselli, Tommaso",
    editor = "Bouamor, Houda  and
      Pino, Juan  and
      Bali, Kalika",
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
    month = dec,
    year = "2023",
    address = "Singapore",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.findings-emnlp.432",
    doi = "10.18653/v1/2023.findings-emnlp.432",
    pages = "6503--6515",
}

@inproceedings{gonzalez-agirre-etal-2024-building-data,
    title = "Building a Data Infrastructure for a Mid-Resource Language: The Case of {C}atalan",
    author = "Gonzalez-Agirre, Aitor  and
      Marimon, Montserrat  and
      Rodriguez-Penagos, Carlos  and
      Aula-Blasco, Javier  and
      Baucells, Irene  and
      Armentano-Oller, Carme  and
      Palomar-Giner, Jorge  and
      Kulebi, Baybars  and
      Villegas, Marta",
    editor = "Calzolari, Nicoletta  and
      Kan, Min-Yen  and
      Hoste, Veronique  and
      Lenci, Alessandro  and
      Sakti, Sakriani  and
      Xue, Nianwen",
    booktitle = "Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)",
    month = may,
    year = "2024",
    address = "Torino, Italia",
    publisher = "ELRA and ICCL",
    url = "https://aclanthology.org/2024.lrec-main.231",
    pages = "2556--2566",
}

Contact Information

For further information, please send an email to langtech@bsc.es.