OncoPatient-MT: A Pilot Dataset for Medical Non-English (Hindi) to English AI Translation (PDMNAIT)

Introduction: Effective communication between patients and healthcare providers is central to oncology care, yet much of the available computational research on clinical language has focused on English-only resources or structured biomedical text (e.g., guidelines, abstracts). This creates a gap for patients whose primary language is not English, particularly in low- and middle-income countries where patient narratives are often expressed in native languages. The OncoVoice-MT (Pilot Dataset) addresses this gap by providing a small but unique collection of parallel Non-English (Hindi) ↔ English transcripts of cancer patient narratives. Unlike existing biomedical translation corpora, which focus on clinical documents or scientific writing, this dataset captures the patient’s own voice, offering insights into lived experiences, symptom descriptions, emotional expressions, and treatment challenges. Dataset: Thirty transcripts were collected from cancer patients who shared their experiences in Hindi. Narratives were obtained in the context of routine care/structured interviews, with patient consent and under institutional ethics clearance. Each transcript was de-identified to remove personal identifiers (names, addresses, dates, hospital IDs). Original-language transcripts: plain text (.txt) All patients provided written informed consent for anonymized data sharing.