Technical Synthesis and Deployment Framework for a Pashto-Centric Pharmaceutical Information System Utilizing Gemma-4 Multimodal Architectures and Local Knowledge Bases
Abstract
Abstract We present three novel computational contributions for Pashto—a low-resource language with approximately 50 million speakers and no prior computational stemmer. First, PashtoMedStem: a rule-based morphological stemmer for Pashto medical text that reduces inflected pharmaceutical terms to their base forms through suffix-stripping rules with protected-stem constraints. Second, PashtoMedNorm: a script normalization pipeline that canonicalizes Unicode variants, strips diacritics, and tokenizes Pashto text for downstream NLP tasks. Third, MARS (Morphology-Aware Retrieval Scoring): a retrieval algorithm that combines stemmed Pashto field matching, cross-script transliteration, and morphological variant bonuses with traditional lexical scoring. These algorithms are developed and evaluated within Pharmaciest, a mobile-first pharmaceutical information system that integrates an 11,825-record locally embedded medicine knowledge base with a Retrieval Augmented Generation (RAG) pipeline built on the Gemma-4 family of large language models. The system employs Google Gemma-4 26B—a Mixture-of-Experts architecture with 3.8B active parameters—as its primary generation model, accessed free of charge through the OpenRouter API gateway. Evaluation on a benchmark of 50 Pashto medical queries across three categories (inflected terms, transliterated drug names, dialect variants) shows that MARS achieves substantial retrieval recall where the baseline English keyword search returns zero…
Read the paper
Related studies
- Knowledge Management Perspective of Generative Artificial Intelligence
- Navigating Generative Artificial Intelligence Promises and Perils for Knowledge and Creative Work
- Characterizing the evolving research on enterprise content management
- Large language models and multimodal foundation models for precision oncology
- From Generative AI to Extended Reality: Multidisciplinary Perspectives on the Challenges, Opportunities, and Future of Educational Computing
- Generative Artificial Intelligence: A New Engine for Advancing Environmental Science and Engineering
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.