An insight of script text extraction performance using machine learning techniques

Abstract

With the evolution of huge amount of ancient and modern text available in digital format, it is ascertain to mine for researchers, government, tourist and travelers visiting all over the world. However, it is very challenging and costly. Further, it takes a lot of effort and time for script text mining. Therefore, the study investigates various techniques for script text mining viz supervised and unsupervised techniques. Firstly, the study presents a survey for various kinds of techniques adopted by the users for extraction of text from image. It also delivers information about gaps involved in the various approaches. Furthermore, it incorporate the quantitative comparisons based among the study of various approaches and techniques for text extraction as well as script level comparison. The result inferred on the basis of the script comparison indicates that, the accuracy level of ancient script was found to be 5% lesser than modern script. Again, furthermore comparison has been done on standalone and hybrid machine (Combination of CNN and KNN) / deep learning techniques. The accuracy has been found to be lower(4%) in case of standalone techniques.