Emotion Classification and Crowd Source Sensing; A Lexicon Based Approach

Abstract

In today's world, social media provides a valuable platform for conveying expressions, thoughts, point-of-views, and communication between people, from diverse walks of life. There are currently approximately 2.62 billion active users' social networks, and this is expected to exceed 3 billion users by 2021. Social networks used to share ideas and information, allowing interaction across communities, organizations, and so forth. Recent studies have found that the typical individual uses these platforms between 2 and 3 h a day. This creates a vast and rich source of data that can play a critical role in decision-making for companies, political campaigns, and administrative management and welfare. Twitter is one of the important players in the social network arena. Every scale of companies, celebrities, different types of organizations, and leaders use Twitter as an instrument for communicating and engaging with their followers. In this paper, we build upon the idea that Twitter data can be analyzed for crowd source sensing and decision-making. In this paper, a new framework is presented that uses Twitter data and performs crowd source sensing. For the proposed framework, real-time data are obtained and then analyzed for emotion classification using a lexicon-based approach. Previous work has found that weather, understandably, has an impact on mood, and we consider these effects on crowd mood. For the experiments, weather data are collected through an application-programming-interface in R and the impact of weather on human sentiments is analyzed. Visualizations of the data are presented and their usefulness for policy/decision makers in different applications is discussed.

Keywords

• Big data
• crowd-sourced sensing
• lexicon-based approach
• Twitter
• social networks