LANDSLIDE EVENT DETECTION USING TWITTER DATA

Abstract

The rise of mobile-posted user-generated content on social media has led to the phenomena known as "citizen sensing". Inspite of the reality that English is presently the accepted international language, people regularly provide local updates in addition to English covering events like disasters. Attempting to combine citizen reports from different languages is challenging. This study describes the solutions that handle this problem in order to enable citizen sensing of the reported landslide incidents around the world. The cornerstone for multilingual help is the first unified cross-lingual library of word vectors for expressing texts in numerous languages. The part of social media in catastrophe reaction and administration is growing. Such a channel could be essential for alerting people to situations, identifying urgent needs, and guiding responses in accordance with expectations. The "native" and "translated" techniques based on monolingual word vectors are inferior than the classification model based on the suggested cross-lingual word vectors. It is also not necessary to write a special instruction manual in a local language or translate it into English.