Developing a System for Detection of Offensive Comments Against Pak Army

Abstract

By the upsurge in the use of internet, huge amount of online data is accessible. Hence, there is a necessity to extract relevant data. In this project, we aim to develop a corpus which comprises of general opinion/comments given by the users on Twitter about Pak Army. For assurance that the annotation phase is error free, we will manually inspect all the comments for annotation. Therefore, we will develop 3 different annotators that will check its progress. Subsequently, we aim to apply tokenization and remove redundant words that do not affect the comments at all, then we apply N-gram model for characters and word cases. In character case 3-gram to 10-gram model will be used whereas in word case 1-gram to 3-gram model will be applied for comparison of each case. Machine learning algorithms will be used for analysis with the N-gram technique. Later, we will develop a graphical user interface (GUI) which have an input field where user can paste/type any piece of text, then we’ll send that text to our model which then return that either this comment is against Pak Army or not. Result will display in a popup modal or as a flash message below input field.