Article Preview
TopIntroduction
In this time of technological development, the social media have become the main way of human communication. The basic comfort of the people irrespective of the distance or magnitude of the audience is only allowed in this platform of social media. Recent improvement in technology depicts that most of the data in social media appears with noise. One of the studies regarding social network data (Ko & Seo, 2000) has revealed that approximately 80% of the data currently available in social media is fully unstructured. As a result of this, it is more sophisticated for social platforms to analyze these data and obtain viable information from the data without delay. In order to carry out this work, two approaches called sentiment analysis and opinion mining have become the world's most important techniques for gaining the most feasible data insights. Social media sentiment analysis can deal with many problems to be solved and helps to provide many indications including public opinion, advertisement, healthcare, and public satisfaction. Discovering hidden pattern from those complex and complicated datasets can improve overall prediction and classification in social media. For this goals to be atained, this research work is focused in analyzing the feelings of many posts and comments on social media data so that immediate action or legal reaction can be conducted. Although, the use of mining techniques to analyzing sentiment from unstructured social network data has become truly challenging. Currently, there are many methods available to enhance the ability of sentiment analysis including feature selection, data integration, data cleaning, and crowdsourcing. Realizing the worth of sentiment analysis in social media platform, this task has become a very potential research area for social media platforms. One of the main reason behind this situation can be demonstrated as free and at any time users authorization and evaluation of the account they have, and all the activities carried out by each and every user of social media are stored for further analysis and manipulation, in order to increase user experience and user satisfaction with these platforms. Several recent works on social media platforms have shown that screening of terrorist activities and the perception of users are now a mandatory task for these platforms. Sentiment analysis is a kind of Natural Language Processing (NLP)where all the texts are processed, understood and sentiments are anticipated to predict suggestions to the users for their better experience with that specific platform. Thousands of works have already conducted by researchers in the field of product reviews, political party, tweet updates, brands of products, social media analysis, etc. (Agarwal et al., 2011). All these systems express opinions in two ways: one is positive and other one is negative. But, here, we focused on multi-class classification for this contribution and also included neutral sentiment. Therefore, from the perspective of the proposed classification, there are two classes: negative and positive. We have mainly concentrated on social network posts, comments, and public opinion, and collected data from Facebook, tweeter, and Instagram. However, the main focus was on Facebook data because this platform is one of the most growing platforms nowadays. Moreover, for status update, Facebook allows a total of 5000 characters for users to save huge data in their own database, and the total number of Facebook users is higher than the one any other platform. It would therefore be easy to collect samples from them, and most of them will be based on real life and challenges. According to a survey article (Mohammadi et al., 2018), a total of 510000 comments are posted on Facebook every 60 seconds and 293000 posts are updated. Realizing the ways how users express their emotions and opinions, the overall analysis can be improved. Therefore, it is mandatory for all the social network platforms to collect the required data and analyze them further to take steps to add additional features. In addition, the user growth rate of twitter from 2010 to 2018 (Bruns & Weller, 2014) was shown in a study, Figure 1.
Figure 1. Graphical representation of Twitter’s user growth rate