EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

INPERIA-Interview: A Synthetic Spanish Dataset of Penitentiary Pseudo-Interviews for Natural Language Processing and Risk Factor Analysis

Enrique Lobo Rubio, Victoria Pachón Álvarez, Jacinto Mata Vázquez

Data in Brief · 2026-11 · 113404

DOI: 10.1016/j.dib.2026.113404 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Data in Brief, 2026-11. Verfasst von Enrique Lobo Rubio und 2 weiteren Autor:innen. Die Arbeit nennt 10 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Research on natural language processing methods for penitentiary decision-support is limited by the scarcity of publicly available interview data, largely because such data involve sensitive legal, institutional and personal information. To address this gap, we present INPERIA-Interview, a synthetic Spanish text dataset composed of penitentiary pseudo-interviews and designed to support NLP research on the analysis of interview answers related to risk factors for temporary release permits. The dataset contains 55 fictional pseudo-interviews and 550 individual answers organised around ten factors inspired by the Spanish Tabla de Valoración del Riesgo (TVR): foreign status, drug dependence, criminal professionalisation, recidivism, breach of sentence, Article 10 of the Spanish General Penitentiary Organic Law, absence of previous prison permits, cohabitation deficiencies, distance from the release destination and internal pressures. The data were generated through structured prompting with ChatGPT and Gemini Pro, using entirely fictional inmate profiles, TVR-inspired questions and simulated contextual information as input. No real penitentiary interviews, prison records, judicial files, clinical data, administrative documents or identifiable information from inmates were entered into these tools. Generated responses were subsequently reviewed manually to improve coherence with the fictional profiles, relevance to the corresponding questions and linguistic naturalness, and to confirm that no real or identifiable information was included. The dataset is provided in tabular format and includes predefined training and test partitions to facilitate reproducible exploratory experiments. Its long-format version allows each answer to be treated as an independent text instance while preserving its link to the original pseudo-interview, question identifier and assessed factor. A deliberately limited subset of 33 response-level records includes preliminary reference levels for three representative factors: foreign status, criminal professionalisation and distance from the release destination. These labels are intended only for exploratory evaluation and should not be interpreted as expert-validated ground truth. Unlike most open Spanish NLP resources, INPERIA-Interview focuses on coherent multi-answer interview sequences in a sensitive institutional domain rather than isolated texts from social media or general-purpose corpora. The resource can be reused for exploratory NLP tasks, prompt engineering experiments, preliminary comparison of language models, design of annotation protocols, error analysis and teaching activities. It is publicly available through Zenodo with a persistent DOI. The resource is intended for research, education, prompt engineering and methodological development, not for operational decision-making.

Zitieren

APA

Rubio, E. L., Álvarez, V. P., & Vázquez, J. M. (2026). INPERIA-Interview: A Synthetic Spanish Dataset of Penitentiary Pseudo-Interviews for Natural Language Processing and Risk Factor Analysis. Data in Brief, 113404. https://doi.org/10.1016/j.dib.2026.113404

Vancouver

Rubio EL, Álvarez VP, Vázquez JM. INPERIA-Interview: A Synthetic Spanish Dataset of Penitentiary Pseudo-Interviews for Natural Language Processing and Risk Factor Analysis. Data in Brief. 2026:113404. doi:10.1016/j.dib.2026.113404

BibTeX
@article{Rubio2026,
  author = {Rubio, Enrique Lobo and Álvarez, Victoria Pachón and Vázquez, Jacinto Mata},
  title = {INPERIA-Interview: A Synthetic Spanish Dataset of Penitentiary Pseudo-Interviews for Natural Language Processing and Risk Factor Analysis},
  journal = {Data in Brief},
  year = {2026},
  pages = {113404},
  doi = {10.1016/j.dib.2026.113404}
}
Kostenlos starten →