🎲 Analyzing Gambling Addictions: A Spanish Corpus for Understanding Pathological Behavior

This repository accompanies the paper:
"Analyzing Gambling Addictions: A Spanish Corpus for Understanding Pathological Behavior"
📍 Accepted at Findings of EMNLP 2025.

📂 Dataset

The main contribution of this work is a Spanish sentence retrieval dataset focused on symptoms associated with pathological gambling.

• Corpus: resources/dataset/corpus.jsonl
• Queries & Qrels: also available in the same directory. Since 4 assessors were employed, the majority agreement is considered when at least two authors consider the sentence relevant.
• All files follow the BEIR compatible format, enabling easy use with standard baselines (see Section 4 of the paper).

Additionally, a subfolder with pools is provided, containing the material used by both human annotators and LLMs for dataset labeling.

---

⚙️ Code

The src folder is structured as follows:

• train/ → Training scripts for our domain-adapted ludoBETO model.
• labelling/ → Statistics and analysis of human vs. automatic label generation. "David Gallego" user is the trained psychologist.

---

🤖 Model

We introduce ludoBETO, a BETO-based model adapted to the pathological gambling domain.
This model is publicly available on HuggingFace for further research and fine-tuning.

🔧 In our paper, we also implemented a cross-encoder using the SimCSE strategy with custom parameters over ludoBETO.

---

📖 Citation

If you use this resource, please cite:

@inproceedings{couto-etal-2025,
    title = "Analyzing Gambling Addictions: A Spanish Corpus for Understanding Pathological Behavior",
    author = "Couto-Pintos, Manuel and
              Fernández-Pichel, Marcos and
               Aragón, Mario Ezra and
              Losada, David E.",
    booktitle = "Findings of the 2025 Conference on Empirical Methods in Natural Language Processing (EMNLP)"
}