Topic Modelling and Opinion Analysis On Climate Change Twitter Data Using LDA And BERT Model.

Abstract

Nowadays, Climate change is an important environmental factor that affects every living thing on the earth. It is very essential to study the public perceptions regarding the disaster events frequently happening due to climate change. In today's digital era individuals are using social network platforms namely Twitter, Facebook, and Weibo now and then to express their views about any events. In this paper, the climate change Twitter data set was considered for analyzing the topics and the opinions discussed by the public regarding climate change. The Latent Dirichlet Allocation(LDA) method was used to list out the various topics present in the data set and the Bidirectional Encoder Representation from Transformers(BERT uncased) is an efficient deep learning technique used to classify the sentiments present in the data set. Here the sentiments were labelled as pro news, support, neutral and anti. The performance of the proposed topic modelling and sentiment classification model was compared using the precision, recall, and accuracy measures. The BERT uncased model with has shown the best results such as precision of 91.35%, recall of 89.65%, and accuracy of 93.50% compared to other methods.