UNDERSTANDING THE PERCEPTION OF MICROPLASTICS USING TWITTER DATA: A SOCIAL MEDIA DATA-BASED SENTIMENT ANALYSIS APPROACH

Abstract

Firms have emphasised on the triple bottom line as a means of doing well by doing good. However, in many respects, doing well would require drawing upon synthetic resources that are potentially environmentally damaging, not now perhaps, but definitely in the future. In this regard, the pervasive application of microplastics while helping to re-engineer many production and supply chain processes and thus save manufacturing and operating costs, poses an insidious threat to bio-diversity and environmental sustainability. This paper will attempt to investigate through Twitter and the user generated content therein, the understanding and awareness of microplastics from the perspective of the consumers who purchase and consume such products. We posit that field of microplastics has been relatively slow in studying big data for sentiments and efficacy of information channels. Therefore, this study aims to contributes to the community by proposing a novel, analytical framework (Twitter Analytics) for analysing microplastics and highlighting the current use of Twitter in microplastics context and further developing insights into the potential role of Twitter for microplastic practice and research. We apply VADER sentiment analysis to decode some of the nuances expressed by consumers towards the enigma of microplastics in sustainability. In this study, we have proposed a microplastics specific sentiment dataset which consists of tweets collected from 22 march 2007 to 8 July 2019. Our dataset consist of 59,757 tweets manually labelled by annotators based on the gold standard and the microplastics-specific sentiment lexicon includes a list of 2,500 words labelled according to its polarity. In this study, we have conducted an extensive experiment to evaluate the performance of three algorithms (Support Vector Machine, Multinomial Na‹ve Bayes and Random Forest) to recognizing sentiment appearing in microplastics conversations on social media therefore utilising different features on our proposed dataset. Also, in this study, we investigated the correlation between the sentiment of user generated content and social events using two different correlation methods(Pearson correlation analysis and Spearman rank correlation analysis).For future work, we plan to investigate the impact of advanced features such as multimedia sentiment analysis and geo-tagged data sentiment analysis.