HARNESSING MESH ONTOLOGY FOR SEMANTIC CLASSIFICATION IN MEDICAL TEXTS

S.NARMATHA; Dr.V.MANIRAJ

Authors

S.NARMATHA RESEARCH SCHOLAR, A.V.V.M SRI PUSHPAM COLLEGE ,POONDI, THANJAVUR-613503
Author
Dr.V.MANIRAJ RESEARCH SUPER VISOR, A.V.V.M SRI PUSHPAM COLLEGE ,POONDI, THANJAVUR-613503
Author

DOI:

Keywords:

Medical Natural Language Processing; Semantic Classification; Domain Ontologies; Medical Subject Headings (MeSH); Concept-Based Representation; Document Classification; Hypernym Integration; Ohsumed Dataset; Text Mining; Knowledge-Based Features

Abstract

Abstract:
This research investigates the difficulties associated with employing domain-specific ontologies for the classification of online textual content. The study specifically examines the use of the Medical Subject Headings (MeSH) ontology to enhance the classification of medical documents. A novel content representation model is introduced based on the insights obtained from this investigation. The proposed method is evaluated against conventional stem-based text representations using two well-known data mining algorithms, namely C4.5 and K-Nearest Neighbors (KNN). Experimental results demonstrate that the proposed approach consistently outperforms traditional methods. By integrating semantic concepts and hierarchical relationships, such as hypernyms, from domain ontologies, the quality of feature representations for document classification is significantly improved. Validation on the Ohsumed biomedical dataset shows that the proposed model achieves up to a 30% increase in classification accuracy compared to standard stem-based techniques.