Development and Evaluation of a Multilingual AI-Driven Dialogue System for Cervical Cancer Awareness and Risk Assessment (Preprint)

BACKGROUND Cervical cancer is the fourth most common cancer among women globally, with an estimated 660,000 new cases and 350,000 deaths in 2022. It imposes a disproportionate burden on underserved populations in low- and middle-income countries, where language barriers, cultural stigma, and limited access to healthcare hinder awareness and uptake of preventive measures. OBJECTIVE This study aimed to design, develop, and evaluate a multilingual AI-driven dialogue system for cervical cancer awareness, personalized risk assessment, and healthcare facility navigation, targeting underserved populations through a culturally adaptive and privacy-preserving digital health platform. METHODS A design science research framework guided the development of a three-tier, cross-platform system comprising (1) a Random Forest risk prediction model trained on the UCI Cervical Cancer Risk Factors dataset (n=858), (2) a retrieval-augmented generation (RAG) chatbot powered by GPT-4-turbo, and (3) a geospatial healthcare facility recommendation module. The system was implemented in Python, Node.js, and Flutter (Google LLC, Mountain View, CA, USA) and deployed across Android, iOS, and web platforms with multilingual support for English, Spanish, Yoruba, and Swahili. Model performance was evaluated using accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve (AUC-ROC). Chatbot quality was assessed using retrieval precision, hallucination rate, and expert clinician review. Usability was evaluated using the System Usability Scale (SUS). RESULTS The Random Forest classifier achieved an accuracy of 94.2% (95% CI 91.4%–96.8%), recall of 93.5%, F1-score of 92.6%, and AUC-ROC of 0.97 (95% CI 0.95–0.99). The RAG chatbot achieved retrieval precision@3 of 0.88 and a 71.4% reduction in hallucination rate compared with a non-RAG baseline. Expert clinician review rated factual correctness at 4.3/5. The geospatial module achieved a specialty relevance precision of 0.87. Backend API latency averaged 312 ms (SD 47 ms) under a simulated load of 50 concurrent users. Preliminary usability testing yielded a mean SUS score of 78.3, indicating good usability. CONCLUSIONS This study demonstrates the technical feasibility of integrating machine-learning-based risk stratification, RAG-grounded conversational AI, and geospatial healthcare navigation within a unified multilingual digital health platform. The system shows promise for improving cervical cancer awareness, risk literacy, and screening access in underserved populations. Future work will focus on large-scale community validation, bias auditing, and integration with public health screening programmes.