An efficient sentimental analysis using hybrid deep learning and optimization technique for Twitter using parts of speech (POS) tagging

摘要

The topic sentiment analysis is like a buzz word among researchers with the advancements in business and social network analysis. Sentiment analysis is the process of recognizing, grouping and classifying the sentiments or opinions conveyed over the social networks creating an immense measure of emotions with rich information as tweets, announcements, blog entries and more. Sentiment analysis considered to be an exceptionally valuable technique in artificial intelligence and is widely used for opinion mining and parts of speech (POS) tagging. Twitter is one among the social network with large number users expressing their thoughts or opinions in a precise and simple way. Analysis of Twitter data is complex compared to other social network data with the existence of slang words and incorrect spellings in a short sentence format. Twitter only permits a maximum of 280 characters per tweet. There were multiple approach such as knowledge based and Deep learning based approach for sentiment analysis using text data. POS is considered as one the required tools in natural language processing (NLP) and Deep learning applications. In this paper, we analyze the tweets of the individual person using hybrid deep learning (HDL) techniques. The proposed system preprocesses the input data before applying HDL techniques. Sentiment analysis in this research is applied using the five-point scale classification as highly negative, negative, neutral, positive and highly positive. The proposed work results in better accuracy and takes less time with a greater number of tweets in comparison with other extensively used models like Random forest, Naive Bayes, and decision tree classifiers. By analyzing various classifiers results in terms of accuracy and precision, ANN achieved 92% accuracy and 91.3% precision, its quite improved results than the other classifiers.