Detection of Natural Calamities from Assamese Posts in Social Media

Abstract

Abstract Social media users and online news portals are rising exponentially in the Northeastern state of India, where every small instance of daily life are posted on social media platforms such as Facebook, Twitter by the users in their native language. Every social media user post about their daily experiences on such kind of social media platform which gives explicit information happening in a particular place. From those posts, extracting of information has been tried regarding natural calamities, a detection system which detects real world happening of natural calamities in a particular place. The objective of this paper is to detect such events from Assamese text posted on social media. Suppose an earthquake has happened, the online news portals, social media users started reporting about the happening in various platforms, just by observing the post an earthquake can be detected easily. A powerful statistical model Conditional Random Field (CRF) is used to detect those natural calamities as events from the posts which are being posted using the Assamese language. This model has the objective to capture the real-world happenings while a goal have been achieved by introducing an event extraction rule. The CRF model is trained with a large dataset NCED20 which is develop manually. The model is trained on a set of features and selecting those features is a significant step in the learning process. In this paper, an algorithm to capture events from the social media post has been proposed.