Framework to Create Sentiment Lexicons For Extremism Detection in Text Documents

Hate speech or extremism detection among the text posts of social media is a major problem nowadays. Hate speech or extremism in text documents can be detected using machine learning methods or using sentiment lexicons. Machine learning methods require already classified data for training. In this paper we have proposed a framework to create sentiment lexicons for any language. In these lexicons terms will be added manually. The sentiment lexicons created by this framework may contain n-grams also where
$\mathrm{n} > =1$
. As the proposed framework is used for creating sentiment lexicons for extremism or hate speech detection, it is better to include only negative terms in sentiment lexicons.