Identification and classification of relations for Indian languages using machine learning approaches for developing a domain specific ontology

Information extraction and classification using Natural Language Processing techniques of layered architecture such as pre-processing task, processing of semantic analysis etc., helps in implementing further deeper evaluation techniques for the accuracy of natural language based electronic database. This paper explores relational information extraction of multilingual IndoWordNet database matching with domain specific terms. Further, extracted information are processed through conventional statistical methods, Normalized Web Distance (NWD) similarity method and two other machine learning evaluation techniques such as Support Vector Machine (SVM), Neural Network (NN) to compare with their accuracy. Results of machine leaning based techniques outperform with significant improved accuracy over conventional methods. The objective of using these techniques along with semantic web technology is to initiate a proof of concept for ontology generation by identification and classification of extracted relational information from IndoWordNet. This paper also highlights domain specific challenges and issues in developing relational model of ontology.