Comparative Study on Multi- Modal Topic Modeling with Dense block Detection

---------------------------------------------------------------------***-------------------------------------------------------------------Abstract— There has been incredible growth of events over the internet in recent years. Google has become the giant source of knowledge for any event which has happened or happening over the internet. Some networking sites such as facebook, micro blogging sites such as twitter are evolved with time and became the highly used sites over the internet. Various E-commerce websites such as Amazon, Ebay, Flipkart etc are the widely used sites for online shopping. Above mentioned sites generates large amount of text data. In association with text data some images are also uploaded over the internet on these sites. The images associated with particular topic plays vital role for understanding the semantic relationship between the textual content and visual content which are related to images. To model this huge amount of data having both textual and visual contents multi-modal topic model is suggested in this paper. While dealing with multi-modality, study of semantic relationship between the images and text data is crucial part. This model also helps to study semantic relationship between them effectively. Topics which are trending, popular over the world can be seen on Social sites as well as micro blogging sites. In online shopping sites fake reviews, advertises, spam spreading information is posted. So to study this suspicious behavior dense suspicious block is also detected in this model. In this paper some and methods are presented and compared. The focus is to effectively model the textual and visual contents in multi-modal dataset having dense block for examining lockstep behaviors.