Acomparto: A RAG-Based Spanish Chatbot for postpartum care
Oxford Open Digital Health · 13.11.2026
DOI: 10.1093/oodh/oqag010 ↗Kurz gefasst
Fachartikel in der Zeitschrift Oxford Open Digital Health, online erschienen am 13.11.2026. Verfasst von Andrea Torres-Calderón und 4 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Nowadays, large language models (LLMs) have become fundamental for the development of conversational agents, which are increasingly adopted across sectors ranging from healthcare and education to customer-facing and organizational support services. When creating a chatbot, several considerations must be taken into account, such as personal data handling, communication strategies, and hallucination mitigation. Moreover, the application domain entails additional considerations regarding the source and quality of the information used. This paper presents a pipeline for integrating a locally hosted LLM into an artificial intelligence assistant, Acomparto, which specializes in evidence-based recommendations for postpartum care. From an implementation perspective, this work examines the practical deployment and integration of the proposed system in real-world maternal health contexts. Our pipeline incorporates a locally running model to mitigate the risk of third-party application programming interfaces using user data in other contexts. The proposed approach involves collecting, preprocessing, and refining textual data to enhance a pre-trained model using a retrieval augmented generation approach. Acomparto addresses one crucial topic: breastfeeding as a starting point. To evaluate our LLM chatbot, we combine standard natural language processing performance measures with expert domain assessments provided by medical and non-healthcare professionals. The results indicate that the pipeline improves the quality and relevance of LLM chatbot responses, particularly for users with limited access to conventional healthcare services . While not intended to replace healthcare professionals, Acomparto provides accessible and trustworthy support through a web-based platform. Additionally, key insights into LLM adaptation to non-English data are discussed, highlighting both the challenges and practical strategies to reduce invalid or misleading outputs.
Zitieren
APA
Torres-Calderón, A., Castillo, E., Choperena, D., Campos-Rivera, P. A., & Sosa-Hernández, V. A. (2026). Acomparto: A RAG-Based Spanish Chatbot for postpartum care. Oxford Open Digital Health. https://doi.org/10.1093/oodh/oqag010
Vancouver
Torres-Calderón A, Castillo E, Choperena D, Campos-Rivera PA, Sosa-Hernández VA. Acomparto: A RAG-Based Spanish Chatbot for postpartum care. Oxford Open Digital Health. 2026. doi:10.1093/oodh/oqag010
BibTeX
@article{TorresCaldern2026,
author = {Torres-Calderón, Andrea and Castillo, Esteban and Choperena, Daniel and Campos-Rivera, Paola Abril and Sosa-Hernández, Victor Adrian},
title = {Acomparto: A RAG-Based Spanish Chatbot for postpartum care},
journal = {Oxford Open Digital Health},
year = {2026},
doi = {10.1093/oodh/oqag010}
}
EUVIMED