A Dissimilarity based Hybrid Approach Along with Novel Score Exploration for Event Detection in Social Streams

The Event detection scheme will helps to detect an event from large dataset and it is also needed preserve the privacy of reviewers because of the social and security issues. A novel recommendation system has been proposed in which depth pre-processing is implemented. In depth pre-processing, the short texts or internet slang acronyms or abbreviations are processed. These short texts are considered to be a mandatory factor for deciding the polarity of user reviews. In the further processing of the user reviews, rather than using the bag-of-words for dictionary compilation, an improved reinforced dictionary formation technique is implemented. A novel feature extraction technique Novel score Exploration (NSE) has been proposed based on considering the positivity, negativity and neutrality along with the a novel user reviews exploration method which will be processed by segregating the user reviews as a sentences, from which the sentence score, position score, modality score and other features are extracted. These additional feature are effective in improving the classification accuracy.