Mining of people’s opinions towards government’s decisions/activities in the twitter posts using Sentiment Analysis

Abstract

In modern world social media is widely used by people around the world and now it has become the most popular platform to express people’s opinions on popular day today topics. This study was carried out by using one of the most popular social media data with some of the trending computer technologies like machine learning, LSTM and word embeddings. One of the challenges that government face is, they cannot get an overall idea about how the people in the country react to the decisions and actions they take, whether the people accept those or not and these cannot be analyzed frequently. As a result, government loses votes in elections if the majority of the people were unhappy about the decisions and actions they made. In order to overcome above problem this study suggests an application to analyze how people react to different types of decisions and actions the government takes, and this can be done frequently, even in daily basis without any additional cost. To use this application developed, the government should have an official twitter page. There the government can tweet their day to day decisions and activities. When the people put comments, replies and likes to those tweets, this application can be used to get an idea about how the majority of the people react to those decisions and actions. The government can use this application in two different ways. 1. They can get an overall idea about how majority of the people reacted to an action/decision they took by analyzing the comments, replies and likes people put to that particular post they have published describing the action/decision they took. 2. They can analyze how the people reacted to the actions/decisions they took during a time period (this month, this week) and can understand whether people are happy about the government or not by analyzing the comments and likes people put to the posts published during that period. Here replies won’t be taken into consideration due to limitations in the twitter API and this will be discussed further in following chapters. To evaluate the model built during this study, metrics like accuracy, precision, recall and f1- score were calculated.