Ontology Construction and Knowledge Graph for Cross Domain Unstructured Text

Abstract

Ontology construction takes a lot of effort and time. Semantic web extract accurate knowledge from large databases. In this paper, an ontology construction process is proposed for cross domain data. The amazon and flip kart reviews are taken to construct ontology for unstructured text data . The data is pre-processed to clean and remove noise. The combined approach of cosine similarity and TF-IDF has been used to find similarity. Further, K means clustering is applied to identify topics. The hierarchical clustering is implemented to represent ontology. The accuracy, precision and recall are calculated by applying different classifier algorithms like Decision Tree Classifier, Gaussian NB, Random Forest Classifier, Support vector classifier and, K Neighbors Classifier. Support vector classifiers show excellent results comparative to other classifier algorithms. Support vector classifier performance shows accuracy - 0.70%, precision- 0.83% , recall- 0.70% and F1-score - 0.73%.