Live Streaming Data Analysis using Distributed Stochastic Bi-LSTM Model

Abstract

In the real world, currently there are more than millions bytes of social media data being created. With the help of modern computational capacity in addition to advanced research in the field of artificial intelligence, a great improvement is shown in natural language processing using deep learning techniques. These AI techniques including machine learning algorithms and deep learning algorithms make it possible to analyze the given tons of data. In addition to it, it also helps us to have a deep dive knowledge in people's mindset towards the current trend in media application. There is another problem of cyber bullying wherein many people start sharing and posting anonymous and fake information to make a significant impact by making riots just as fake heroes. By continuously monitoring these types of online riots will provide a clean and peaceful environment which in turn supports the government in a better way. This article focuses on performing sentiment analysis over a given real time social media data by categorizing it. The effective analysis of live streaming from social media applications is done with word2vec embedding method in natural language processing and trained using bi-directional LSTM model. The performance is compared with the LSTM model, and traditional machine learning techniques like naïve bayes and linear regression model and Bi-LSTM model outperforms the other techniques with 80%-85% accuracy with live streaming unconventional social media data.