Deep Learning Based Event Detection Using Multilingual Tweets During Disaster

Concerning the calamity, users typically provide data via social media in their native language, which is difficult to process and understand for others who are unfamiliar with that language. The proposed approach is intended to determine the location during a disaster using multilingual tweets. The study analysis with the tweets, extracted using # 2015 Chennai floods and # heavy rain in Tamil language. Multilingual tweets are trained using an auto-encoder model and then disaster-related tweets are grouped using the K-means clustering method. Integration of natural language processing and machine learning techniques helps to extract the location from the tweets. The trained model results in producing the encoded representation of tweets and the number of clusters is determined by the elbow method using K-mean clustering. IndicNER Fine-tuned model with 93% accuracy is used to identify the location entities of the Tamil Tweets extracted. DisMan application is designed to visualize the disaster location in the geographical map and it helps in quick response with the rescue team to identify the disaster-affected area.