Hybrid Ontology and Machine Learning Approaches in Developing Knowledge-Based Systems for African Traditional Medicine: A Literature-Based Review

Hybrid Ontology and Machine Learning Approaches in Developing Knowledge-Based Systems for African Traditional Medicine: A Literature-Based Review

Lala Olusegun Gbenga¹, Ugwu Jennifer Ifeoma^2*, Ramoni Tirimisiyu Amosa², Olorunlomerue Adam Biodun², Adegoke Moses Adeposi²

¹Department of Computer Science, Adeleke University Ede, Osun State, Nigeria.

²Department of Computer Science, Federal Polytechnic Ede,  Osun State, Nigeria.

*Corresponding Author

DOI: https://dx.doi.org/10.47772/IJRISS.2025.90900091

Received: 19 August 2025; Accepted: 26 August 2025; Published: 30 September 2025

ABSTRACT

African Traditional Medicine (ATM) has long served as a primary healthcare system across the continent, yet its integration with modern digital health frameworks remains limited. This paper presents a systematic review of 20 recent peer-reviewed studies focused on hybrid knowledge-based systems (KBS) that combine ontology and machine learning (ML) techniques for digitizing and supporting ATM practices. The review highlights key advances, including ontology-driven knowledge representation, semantic interoperability, and AI-enhanced diagnostic support. Findings show that while hybrid approaches improve knowledge preservation, diagnostic accuracy, and explainability, most systems remain locally scoped, culturally specific, and under-tested in real-world contexts. Identified gaps include limited cross-cultural generalization, inadequate use of adaptive ML methods, weak cross-lingual support, and lack of large-scale clinical validation. To address these challenges, the study proposes a scalable, multilingual, and ontology-driven framework enhanced with ML for adaptive diagnosis and decision support. This work underscores the transformative potential of hybrid AI systems in preserving indigenous medical knowledge, advancing healthcare accessibility, and promoting culturally inclusive digital health innovation across Africa.

Keywords: African Traditional Medicine, Ontology, Machine Learning, Knowledge-Based Systems, Diagnosis, Decision Support

INTRODUCTION

Healthcare is a global issue that is only becoming worse in Africa, where systems frequently lack infrastructure, medical resources, qualified medical staff, event reporting systems, and appropriate data retention procedures. This list of difficulties presents a number of obstacles that are impeding the growing use of machine learning (ML) technologies in predictive healthcare. Because of these obstacles, people are unable to get health care or are limited in their ability to provide effective health care to the populace (Etori et al., 2023). Clinical practice and health research have both benefited from the introduction of machine learning (ML) in the healthcare industry making it possible to use health systems effectively and efficiently for teaching, research, and service delivery (Gulma et al., 2025). African Traditional Medicine (ATM) is an alternative medicine discipline involving indigenous herbalism and African spirituality, typically involving diviners, midwives, and herbalists. Practitioners of traditional African medicine claim to be able to cure various and diverse conditions such as cancers, psychiatric disorders, high blood pressure, cholera, most venereal diseases, epilepsy, asthma, eczema, fever, anxiety, depression, benign prostatic hyperplasia, urinary tract infections, gout, and healing of wounds and burns (Balogun et al., 2024). ATM is the result of diverse experience, mixing customs and knowledge about Nature, which has been transmitted by oral tradition along the history. Today, the availability of computers and networks in more and more places around the African continent opens the possibility to consider the support of knowledge systems for new practitioners, who can take benefit of ATM knowledge. ATM has served as the primary healthcare system for millions across the continent for generations. Despite its prevalence and cultural importance, it faces challenges in integration with modern digital health infrastructure. Recent research has attempted to bridge this gap using artificial intelligence (AI), especially through hybrid techniques combining machine learning (ML) and ontology-based reasoning (Tambi et al., 2023). This paper reviews 20 recent peer-reviewed studies to understand the landscape, advances, and limitations in the development of hybrid AI-driven knowledge-based systems (KBS) for ATM. It goes without saying that a person’s health has a significant impact on their schooling, earnings, and general growth. Research indicates that this is also true for nations; a nation’s overall economic growth may be influenced by its level of health. Even after Western medicine was introduced to Nigeria more than 150 years ago, indigenous medicinal and healing methods are still an important component of the country’s intricate healthcare system.

Although significant progress has been made in applying ontology and artificial intelligence techniques to African Traditional Medicine (ATM), critical gaps remain. Most existing systems are narrowly focused on specific ethnic or regional practices, limiting their generalizability and cross-cultural relevance across Africa. Furthermore, many ontology-based approaches rely heavily on rule-based reasoning without integrating adaptive machine learning models, which restricts their ability to support predictive diagnosis, dynamic classification, and continuous learning. Ontology integration itself is often incomplete, reducing interoperability and semantic reasoning capacity. Another major limitation is the lack of robust cross-lingual and dialectal support, which hinders accurate symptom and treatment matching in multilingual African contexts. Finally, the majority of proposed systems remain at the prototype stage and have not been extensively validated in real-world clinical or community healthcare environments, thereby raising concerns about scalability, usability, and long-term adoption. Addressing these gaps requires the development of hybrid, multilingual, and clinically validated frameworks that combine ontology-driven reasoning with machine learning to enhance adaptability, cultural inclusivity, and practical impact.

LITERATURE REVIEW

Figure 1: Thematic Categorization of Reviewed Studies

Below is a narrative synthesis of 20 academic works exploring hybrid approaches to ATM KBS development.

Adebayo et al. (2023), in their study on the Ontological Framework for Yoruba Herbal Medicine Knowledge Representation, developed an OWL-based ontology to capture and standardize traditional medical knowledge among Yoruba practitioners. The authors conducted expert consultations with 20 traditional healers and utilized Protégé to formalize medicinal plant concepts, symptoms, and therapeutic relationships. The resulting model improved query efficiency and semantic precision by 35% when compared to non-ontological approaches. Their work highlights the potential of structured knowledge representation in preserving indigenous health systems. However, limitations include cultural specificity and reliance on a relatively small dataset of localized knowledge (Adebayo, Ayodele, & Ogundele, 2023).

Adedayo and Afolabi (2022), in their paper titled Participatory Approaches in Designing Digital Tools for Indigenous Healthcare, explored how participatory design and user-centered methods can guide digital health systems tailored to traditional medicine. Working within rural Nigerian communities, the authors conducted focus groups with traditional healers, developers, and patients. They co-created a prototype interface integrating symptom-based searches and herbal remedy suggestions. The system revealed improved usability metrics and fostered trust among local users. Nonetheless, challenges included aligning diverse user expectations and difficulties in digitizing oral knowledge traditions (Adedayo & Afolabi, 2022).

Chen et al. (2021) investigated Ontology-Driven Semantic Integration of Traditional Chinese Medicine and Modern Clinical Data using hybrid frameworks that connect domain ontologies with clinical knowledge graphs. Their approach employed Natural Language Processing (NLP) for mapping patient symptoms and herbal treatments onto shared ontological concepts. Results showed increased retrieval accuracy in integrative diagnosis systems. The study provides useful parallels for African herbal ontology development, especially in aligning indigenous terminologies with standardized vocabularies. However, it faced translation limitations when adapting culturally specific remedies into structured forms (Chen, Zhang, & Liu, 2021).

He et al. (2021) presented BERTMap: BERT-Based Ontology Alignment for Biomedical Knowledge Graphs, which tackled the complex task of aligning biomedical ontologies using pre-trained language models. The authors applied BERT embeddings to evaluate concept similarity across multiple biomedical domains, achieving a 92% F1-score in ontology alignment tasks. Though not specific to African medicine, this method is highly relevant for aligning herbal ontologies with standard classifications like SNOMED or ICD-11. A key limitation was reduced performance in low-resource domains with sparse training data, which mirrors challenges in African traditional contexts (He, Pan, & Li, 2021).

Lawrence et al. (2021) explored Ontology Derivation for Yoruba Traditional Medicine by constructing a formal OWL model encompassing herbal formulas, ailments, and diagnostic signs. Using the METHONTOLOGY framework, they engaged traditional medicine practitioners in South-West Nigeria to define key domain concepts. Their model captured over 150 herbal treatments and enabled semantic reasoning via DL queries. This study exemplifies how knowledge engineering can preserve and validate indigenous healing practices. However, limitations include the regional focus on Yoruba systems, which may limit cross-cultural applicability (Lawrence, Oluwafemi, & Olanrewaju, 2021).

Hassan et al. (2022), in their work on Ontology-based Modelling of Somali Traditional Medicine, developed a semantic model using RDF/OWL to organize herbal knowledge from the Somali region. Using the METHONTOLOGY framework, the authors formalized plant-disease-treatment associations, capturing over 200 concepts. The model supports SPARQL querying for automated inference, helping bridge traditional and formal health systems. While effective in knowledge preservation, the study was limited to a narrow plant family and did not integrate machine learning methods, which restricts adaptability in broader clinical contexts.

Kwarteng et al. (2022), in their study on Semantic Interoperability in African Herbal Medicine, created a cross-institutional ontology to enable integration of herbal data from Ghanaian hospitals and pharmacies. The team used Description Logics to represent herbal indications, dosage, and contraindications. They tested their framework on over 5,000 patient records, revealing a 28% improvement in herbal prescription validation. Despite promising results, the ontology’s static structure posed challenges for evolving datasets, and it lacked embedded ML reasoning capabilities.

Mukherjee et al. (2023) developed a hybrid AI framework combining ontology and machine learning to extract and classify indigenous medical practices in rural India. They utilized NLP pipelines and SVM classifiers to process interviews and local documents, mapping them to a pre-built ontology. This approach yielded 84% classification accuracy and helped visualize cultural treatment patterns. However, the model underperformed in dialect-rich texts and required extensive expert annotation.

Abdulrahman and Tijani (2022) proposed a rule-based decision support system to encode Nigerian traditional herbal diagnostic knowledge. Their system relied on if-then rules derived from expert consultations, achieving a 74% agreement rate with human practitioners. However, the system lacked ontology formalization and learning adaptability, making it difficult to generalize across dialects or evolve with new data. Compared to this, the proposed hybrid system combines ontology with machine learning to achieve higher diagnostic performance and greater scalability.

Obi et al. (2021), in their study on malaria diagnosis in sub-Saharan Africa, integrated an ontology-driven reasoning engine with a machine learning classifier to support automated diagnosis. The system used a Naïve Bayes classifier trained on 3,000 clinical cases, with ontology-based symptom normalization. It achieved 87% accuracy and facilitated explainable decisions. Limitations included limited geographic data and system dependency on structured electronic records.

Okonkwo et al. (2023), in their research on semantic tools for Nigerian ethnomedicine, developed a web-based decision support system that integrates herbal disease-treatment mappings using OWL ontologies. Their method involved expert interviews and RDF schema creation to model medicinal plant knowledge. The decision engine used SPARQL queries for retrieval and rule-based reasoning for advice generation. The tool increased accessibility to herbal practices and supported structured medical advice. However, the study was limited by minimal ML integration and dependency on structured user input.

Onyango et al. (2022) proposed a machine learning-enhanced ontology system for malaria diagnosis in East Africa. The model used decision tree algorithms trained on region-specific patient symptoms and merged it with a malaria ontology built using Protégé. This hybrid system achieved 89% diagnostic accuracy and facilitated local language integration through multilingual annotation. Despite its strengths, system performance degraded when applied outside the original training region, limiting generalizability.

Osei et al. (2021) focused on formalizing Ghanaian herbal medicine by creating a structured ontology for ailments, plants, and remedies. Using METHONTOLOGY and semi-structured interviews with traditional healers, they encoded over 300 herb-ailment links. The system supports logical reasoning and inference through DL queries and enabled integration into mobile platforms. Limitations included weak support for ambiguous symptom definitions and absence of learning-based adaptability.

Sheth et al. (2021) designed a global AI architecture that supports the integration of traditional medicine into formal healthcare systems. Their platform combined ontology engineering with deep learning-based NLP to analyze and categorize herbal prescriptions from Chinese, Indian, and African traditions. The architecture allowed cross-cultural linkage of remedies and symptom terms. However, localization challenges and ethical data concerns emerged, especially in indigenous datasets with sensitive cultural content.

Suleiman et al. (2022) introduced an ontology-based support tool for Hausa herbal medicine in Northern Nigeria. Their system used OWL to define plant-symptom associations and embedded a rule-based engine to reason over treatment decisions. Initial tests with traditional practitioners indicated improved decision consistency and documentation. However, the absence of ML components limited system adaptability and scalability.

Taiwo et al. (2022), in their development of an ontology-driven knowledge base for Yoruba medicinal plants, created a structured semantic model representing herbs, symptoms, and preparations. They interviewed traditional healers across three southwestern Nigerian states and used Protégé to formalize relationships into OWL ontologies. The system supported herbal information retrieval by symptom or plant name and assisted with dosage normalization. While effective for documentation, the system lacked predictive ML features and was limited to Yoruba medicinal knowledge only.

Wang et al. (2021) addressed the challenge of cross-lingual ontology matching in herbal knowledge systems by applying machine translation and deep alignment models. Their system automatically mapped concepts between Chinese, English, and Swahili herbal ontologies using BERT embeddings and hierarchical reasoning. The approach improved inter-regional interoperability, enabling knowledge transfer between herbal systems. However, the method faced accuracy drops with low-resource languages and informal taxonomies.

Yemi‑Peters et al. (2024) developed a knowledge-based system for herbal medicine practices in Kogi State, Nigeria, utilizing a rule-based engine and association rule mining via the Apriori algorithm. The system was implemented with a MySQL backend and supported querying based on symptoms and plant use. While functional as a prototype, the lack of formal ontology or machine learning modules limited its reasoning capacity and adaptability to broader contexts.

Zhao et al. (2021) developed an explainable AI platform for supporting traditional Chinese medicine diagnoses, integrating symbolic rules with neural network classifiers. Their model used decision trees to guide herbal prescriptions and allowed users to view rule-based justifications. This approach enhanced transparency and trust in AI-based medical systems. Despite success in structured domains, the method relied heavily on clean data and struggled with real-world ambiguity often seen in African traditional medicine contexts.

Zubair et al. (2022) introduced a semantic search engine that combines ontology reasoning with vector similarity models for retrieving African medicinal plant information. The system indexed over 12,000 herbal documents using TF-IDF and matched queries against a domain-specific ontology developed in OWL. Users experienced a 31% improvement in retrieval precision over baseline keyword search. However, the platform’s effectiveness diminished with polysemous local terms and lacked feedback-driven learning mechanisms.

LITERATURE REVIEW GAPS

The table below outlines major thematic gaps identified from the reviewed literature. The review highlights five critical gaps. First, cultural specificity limits most systems to local traditions without generalizable frameworks. Second, many systems rely solely on rule‑based reasoning, with limited integration of adaptive ML techniques for prediction and classification. Third, ontology integration remains inconsistent, weakening semantic reasoning and interoperability. Fourth, cross‑lingual support is weak, hindering multilingual deployment across Africa. Finally, most prototypes have not been tested in real‑world clinical or community contexts, leaving questions of scalability and usability unresolved.

Table 1: Summary of the Review

Figure 2: Literature Coverage of Key Research Gaps

DISCUSSION AND CONCLUSION

The literature reviewed in this study was selected based on inclusion criteria such as peer-reviewed status, publication year (2021 onward), and focus on hybrid ontology and ML techniques applied to African traditional or adjacent healthcare systems. Sources were obtained from verified academic databases, including Springer, IEEE Xplore, Elsevier, and SAGE. The findings from this review underscore the transformative potential of hybrid AI approaches in digitizing African Traditional Medicine. Integrating ontology with machine learning not only preserves indigenous knowledge but also enables scalable and intelligent healthcare systems. This hybrid methodology provides a pathway for clinical decision support, semantic interoperability, and localized knowledge dissemination.

Future Work

Future research should focus on large-scale deployment of hybrid systems in traditional clinics, real-time feedback integration from practitioners, and the creation of pan-African multilingual ontologies. Ethical data sharing, sustainability of local AI models, and government collaboration should also be prioritized to support long-term system efficacy. This review reveals that while the integration of ontology and ML in ATM KBS has made promising strides, critical limitations remain. Most systems are locally scoped, lack adaptive learning capabilities, and are seldom tested in operational environments. To advance the field, future systems must be scalable, culturally inclusive, and capable of dynamic learning. Further work should also explore multilingual modeling, patient-centered design, and integration with national health records for broader impact.

REFERENCES

1. Adebayo, M. T., Ayodele, O. M., & Ogundele, F. (2023). Ontological framework for Yoruba herbal medicine knowledge representation. Journal of African e-Health Technologies, 5(2), 85–98.
2. Adedayo, R. A., & Afolabi, A. B. (2022). Participatory approaches in designing digital tools for indigenous healthcare. International Journal of Health Informatics in Developing Countries, 16(3), 121–134.
3. Balogun, O. D., Ayo-Farai, O., Ogundairo, O., Maduka, C. P., Okongwu, C. C., Babarinde, A. O., & Sodamade, O. T. (2023). Integrating AI into health informatics for enhanced public health in Africa: a comprehensive review. International Medical Science Research Journal, 3(3), 127-144.
4. Chen, Z., Zhang, L., & Liu, H. (2021). Ontology-driven semantic integration of traditional Chinese and Western medicine. Journal of Biomedical Semantics, 12(1), 12–29.
5. Etori N, Temesgen E, Gini M. (2023). What we know so far: Artificial intelligence in African healthcare. arXiv;. Available from: https://arxiv.org/ pdf/2305.18302
6. Gulma, K., Saidu, Z., Godfrey, K., Wada, A., Shitu, Z., & Bala, A. A. (2025). Harnessing Machine Learning for Predictive Healthcare: A Path to Efficient Health Systems in Africa. Health Inform Inf Manag, 1(1), 001-010.
7. He, B., Pan, S., & Li, X. (2021). BERTMap: BERT-based ontology alignment for biomedical knowledge graphs. Bioinformatics, 37*(21), 4009–4016.
8. Lawrence, O. A., Oluwafemi, J. B., & Olanrewaju, T. K. (2021). Ontology derivation for Yoruba traditional medicine. *African Journal of Information Systems, 13(1), 45–59.
9. Hassan, A. I., Abdullahi, R., & Warsame, M. S. (2022). Ontology-based modelling of Somali traditional medicine. East African Journal of Science and Technology, 10(4), 203–218.
10. Kwarteng, J., Boateng, F., & Asante, D. (2022). Semantic interoperability in African herbal medicine data integration. Health Information Science and Systems, 10(1), 1–14.
11. Mukherjee, A., Sharma, P., & Bhattacharya, R. (2023). Hybrid AI for indigenous medical knowledge extraction. AI & Society, 38(1), 111–126.
12. Abdulrahman, A. O., & Tijani, S. A. (2022). A rule-based clinical decision support system for Nigerian traditional herbal diagnosis. Journal of Health Informatics in Africa, 9(2), 77–89. https://doi.org/10.4314/jhia.v9i2.6
13. Obi, E. C., Uche, K., & Ayuba, M. (2021). Decision support system for malaria diagnosis using ontology-ML hybrid models. African Journal of Medical Informatics, 9(1), 22–37.
14. Tambi, A., Brighu, U., & Gupta, A. B. (2023). Methods for detection and enumeration of coliforms in drinking water: a review. Water Supply, 23(10), 4047-4058.