Topic Modelling and Clustering of Disaster-Related Tweets using Bilingual Latent Dirichlet Allocation and Incremental Clustering Algorithm with Support Vector Machines for Need Assessment

The occurrence of various types of disasters are thoroughly described in social media websites like Twitter which can be a useful source of data. This research aimed to solve the problem of bilinguality of data which caused a complexity that yielded inaccurate results during clustering. The researchers were able to develop a system which adds a bilingual topic model, Bilingual Latent Dirichlet, to an existing system. This research was able to compare in terms of precision, recall, accuracy, and area under the curve metrics, the clustering accuracy of the incremental clustering module with and without the bilingual topic model, while also classifying the needs that the Tweet will require.