A novel framework for assessing the criticality of retrieved information

Abstract

Data created by microblogging platforms provide an exceptional opportunity to mine valuable insights; however, their application in critical information retrieval is still at its inflection point. Taking advantage of Deep Learning (DL) and Natural Language Processing (NLP) techniques, this paper proposes a novel framework for retrieving critical information from Twitter to manage emergencies effectively. The proposed framework classifies the tweets into relevant and irrelevant classes using Bidirectional Encoder Representations from Transformers (BERT). Subsequently, relevant tweets are clustered using a k-means algorithm based on textual semantic similarity obtained using Universal Sentence Encoder (USE). Finally, the critical value of tweets is computed to segregate the relevant information that may assist the management teams to plan and organize their operations efficiently. The proposed work was tested on a real-world dataset of Uttarakhand Floods that occurred in February 2021. The critical information retrieved may be deployed to quickly manage disastrous situations and take the appropriate measures in time.