Lancaster sliced regressive keyword extraction based semantic analytics on social media documents

Abstract

Semantic analytics is one of the new issues materialized in Natural Language Processing (NLP) with the emergence of social networks. Semantic analytics on social media documents refers to the procedure of employing NLP techniques for analyzing deeper sense and context of text on social media platforms. Making use of amount of information being now available, research and industry have attempted materials and mechanisms to analyze sentiments automatically in social networks.It just goes beyond keyword exploration to understand the associations between words, phrases and concepts within a social media post, recognizing for a more refined clarification of user sentiment and purpose. While the extensive greater part of these days researchare completely concentrating on enhancing the algorithms employed for sentiment evaluation, the present one emphasizes the advantages of employing a semantic based method for representing the analysis’ results, the emotions and social media specific concepts. In this work a method called, Lancaster Tokenized Sliced Inverse Regressive Keyword Extraction (LT-SIRKE) for performing efficient semantic analysis on social media documents is introduced. LT-SIRKE technique is divide as query pre-processing as well as keyword extraction. Initially in LT-SIRKE method, the user inputs their query into the user window. Afterward, the query is sent to the system for efficient pre-processing. In query pre-processing phase, Stochastic Gradient Descent Keras-based tokenization, Lancaster-based stemming and Zipf’s Law-based stop word removal process is carried out. After preprocessing, keywords are extracted using Bayesian Averaging and Sliced Inverse Regression-based Keyword Extraction to facilitate efficient information access. Experimental assessment is performed with various metrics namely precision, recall, accuracy, keyword extraction time and error with number of user requested queries.