Font Script Identification Based on N-gram Text Categorization

Abstract

In this paper, we propose a method for identifying font scripts of Myanmar Language. Because of the unavailability of nationwide standardized encoding scheme in Myanmar font scripts, knowledge written in Myanmar language are scattered across internet pages. Font scripts Identifier are essential to merge those scattered knowledge into one for NLP application such as text categorization, information retrieval and text summarization. Our proposed method use N-gram based text categorization. A piece of text for 11 font scripts is taken for training. TF-IDF (Term Frequency-Inverse Document Frequency) weights of character N-grams for each font script are computed and stored as a profile for that particular font script. When a new text document is given to testify, TF-IDF weight is computed for that font script and cosine similarity is measured between the test and trained profiles. The highest similarity scored of the font script is taken as a result. 100% accuracy is obtained for testing of 11different font scripts by applying TF-IDF approach. Therefore, this method works well for Myanmar font script identification.