Ontology Based Machine Learning Approach to Automatic Labelling for Research Papers on Wildlife of Sri Lanka

Abstract

Sri Lanka, being a global biodiversity hotspot, places great emphasis on biodiversity from an ecological perspective, socio-economic, and cultural factors. However, the wildlife of Sri Lanka is critically threatened due to several factors. Mainly human activities and needs supersede conservation measures. Lack of knowledge and technical support also hinder wildlife management activities. Findings of wildlife research studies could be incorporated into data-driven conservation and management decisions but the current contribution is not satisfactory. This research shows a novel data mining approach for finding hidden keywords and automatic labeling of past research work in this domain. We used the Latent Dirichlet Allocation (LDA) algorithms to model topics and identify the major keywords. Using the output of Topic Modelling an ontology model was also developed to represent the relationships between each keyword. Using the ontology instances we classified the research papers using Artificial Neural Network (ANN) to predict the labels for research papers in the wildlife domain. These approaches can be used for guiding future research endeavors, with the recognition of research gaps and by classifying the subjects related to a publication by the non-professional related fields. The experimental results demonstrated a 83% of accuracy for the proposed method.