Optimization of Microblog Representation Using Deep Learning Approach

Abstract

Microblogging, as a new form of online communication in which users talk about their daily lives, to gather real-time news, opinion about people or share information by short posts, has become one of the most popular social networking services today, e.g. to stay in touch with friends . Finding proper representations of microblog texts is a challenging issue. The overview of microblog included how to extract information from microblog, also discuss about Microblogging and Twitter. This paper included existing research on Optimization of Microblog Representation by using existing techniques and methods as well as microblogging services.Current approaches, for microblog reduction are single indexing. In this work, we have proposed double indexing method for microblog reduction. We have used our own generated dataset for microblog reduction. We have collected tweets from real time twitter for both single and double indexing technique. Also, we have use LDA with Semantic Similarity algorithm. We compared our results with single indexing. Experimental results shows that double indexing gives better performance than single indexing. This may happen because we will show both original and reduced tweets, so it will be known to user which tweet are removed and relevant part shows as output. We have also compared our results with the state-of-art. The proposed double indexing gives better performance than existing 1-ROCA of improved online SVM technique.