Finding Spatial-Textual Clusters in COVID Tweets

Finding Semantic similarity in text is a vital concept in the fields of information mining, text-based profiling. There have been many approaches to improve information retrieval by mining the semantics of the text. With the pandemic situation prevailing all over the world, we come across many useful posts about the COVID infection that is being tweeted by medical practitioners and people in the health care sector. While we come across such tweets, we also have tweets related to the vaccines, medical facilities, change in economic conditions due to pandemic, etc. But there is no methodology to efficiently study the tweet data and retrieve useful information out of them. Also, we need to utilize the geographical information that comes with each tweet. Though there have been many studies conducted on sentiment analysis, statistical analysis related to twitter data, there has not been much research on finding out the geographical distribution of COVID related tweets combined with query-based textual similarity of COVID related tweets. In this paper, we try to study the semantics of geo-tagged twitter data related to COVID and segregate the tweets based on their geographical location and according to the content of tweets. We use an improved version of Density-Based Spatial Clustering for clustering the tweets according to geo-spatial information. Then, we apply cosine similarity techniques to do the textural clustering and evaluate the performance of proposed model. The proposed model is able to cluster tweets using the spatial coordinates and classify the tweets based on the textual similarity measure.