Lessons Learned from Event Detection from Arabic Tweets: The Case of Jordan Flash Floods near Dead Sea

Event detection is essential for decision makers to understand the events surrounding their real world. Social media microblogging platforms play a significant role in our life. One of these platforms is Twitter, which has an extreme high exchange rate and accordingly has become a valuable and relevant source for many political and social events. Event detection from social media attracted the attention of researchers in different natural languages. Extracting and detecting events from Arabic tweets is still under investigation. In this paper, we have used the Python Natural Language Toolkit (NLTK) library to develop two classifiers for filtering and detecting extracted events from Arabic tweets. The first classifier filters the collected tweets using two passes. The first pass identifies the hashtags while the second pass does a shallow analysis on the tweets content. The second classifier analyzes the text extracted from the tweets. As a case study, we present the tragic events of the Jordan flash floods near the Dead Sea. The model successfully filtered all the collected tweets and picked the ones describing the incidents within that region. Analyzed data revealed important information to learn from this lesson for the future. The solution can be generalized and adapted to other problems.