Exploring Differences in the Sentiment Analysis Tools Using Twitter Data Concerning Autism Awareness

There are multiple studies analyzing people\u27s conversations through social media data sources such as Facebook and Twitter (Ghiassi et al., 2015, Abbasi et al., 2014, Marquez et al., 2013, Rodrigues et al., 2016). Opinion mining, or sentiment analysis, is commonly applied to social media conversations to discover how people feel and their level of affect toward a specific topic. Multiple methods exist to perform sentiment analysis which can result in different outcomes. The goal of this research is to investigate two methods (sci-kit learn and vader) for calculating the sentiment expressed in user Twitter feeds and to report on the similarities and differences of each method. We plan to examine the strengths and weakness of each method, prescribing how they may be best applied for future research objectives. Lastly, we demonstrate the use of sentiment analysis as it applies to Twitter conversations in relation to autism awareness, an important topic in public health. To conduct this research, we look primarily at exploring the differences in tools used in sentiment analysis as demonstrated in prior studies using social media data (Ghiassi et al., 2015, Abbasi et al., 2014, Marquez et al., 2013, Yang et al., 2015). To narrow the scope of this research, we chose to investigate the technique of applying sentiment analysis using sci-kit learn and vader, both open source language libraries in python, the software development platform used to build our application. Python is versatile in analyzing text data and computational linguistics, by applying natural language processing, allowing us to extract substantial information from text data and perform sentiment analysis (Batrinca and Treleaven, 2014). Furthermore, we apply sentiment analysis from each method using Twitter data from users discussing topics related to autism awareness. The benefits from this research is twofold. First we are able to demonstrate two different techniques usable for sentiment analysis with Twitter data. Second, using the method of sentiment analysis, we can illustrate some of the main themes and opinions related to public health and autism. We anticipate the outcomes of this research will benefit scholars that are looking to use sentiment analysis by illustrating available methodologies and produce results using those methods. Lastly, we believe by demonstrating the use of sentiment analysis with Twitter data associated with public health, we can demonstrate the benefits of sentiment analysis exploring public views of a very important topic