A Twitter Framework to assess the Effectiveness of Indian Government Campaign

Abstract

Many government and private agencies introduce social, governmental, educational, commercial and public-health based campaigns to refine various sections of the society. Traditionally, citizen feedback and on-field surveys are performed to assess the effectiveness of such campaigns. However, there is limited availability of the social media tools that assess the impact of different government campaigns automatically. In this research work, a framework has been proposed which uses the social media data i.e. Twitter data pertaining to an Indian Cleanliness Campaign Swachh Bharat Abhiyan (SBA) to perform the effectiveness assessment. This research work has been performed in two parts. First, Twitter data has been processed to predict the perceptions of citizens using Word Embeddings -based Tweet Pooling and Subjectivity score -based Sentiment Analysis and second, the performance of the cities has been predicted using the demographic features based predictor models. The experimentation shows a 0.77 correlation between the proposed framework and the government surveys while predicting the citizens’ perspectives and 80(+/-15)% accuracy while predicting the performance of cities using Random Forest Regression. Furthermore, the cities have been clustered using Twitter and the demographic data to find out the interesting patterns and behaviors. This research work provides better insights to the potential of social media data in the interventional studies of the developing countries such as India.

Sentiment Analysis of National Eligibility-Cum Entrance Test on Twitter Data Using Machine Learning Techniques

“…Finally, the recital of the city was predicted using predictive models generated from demographic features. Experiments confirm a correlation of 0.77 as well as using Random Forest Regression; cities can be predicted with a degree of accuracy between 80 and 85% [2]. Nadia Felix F et al presented an overview of semi-supervised methods for classification of tweets.…”