Statistic-Based Sentiment Analysis of Social Media Data

This paper presents one of the practices of text/opinion mining of web data which can help to provide assistance to prepare reports for Customer Relationship Management (CRM). For convenience we use Twitter tweets data for sentiment analysis. These sentiment values are further treated with statistical evaluations like z-tests and chi-squared test. As social media data are considered as normally distributed, the test like ANOVA test, t-test (for small sample-size) and hypotheses test can be used. Python programming language is used for the task as it has several libraries and packages for Natural Language Processing, statistics, data visualization while supporting the features of general-purpose programming language. This paper also dictates the process of fetching, storing, cleaning, language - translating, sentiment & statistical evaluating, simulating & building theoretical model for hypothesis testing of the data.