Temporal Analysis of Public Sentiment From Opinions Shared in News Contents

Abstract

With the advanced technology and state-of-the-art models, natural language processing and sentiment analysis have been a major area for research. In this research, we have included the implementation of temporal analysis along with sentiment analysis models. The work is based on news article data from a Bengali news channel. Viewers share their opinions below the comment section of the corresponding news item. These news and comments have been a great source of information and research. In this work, we first proposed a dataset containing 7,62,678 public comments and replies from 16,016 video news published from 2017 to 2023 from a renowned Bengali news YouTube channel. The dataset is then processed and data labels are annotated using a machine-learning model after repetitive testing. Data temporal analysis have been done later to identify how the data pattern changes over time for news data. The temporal study has been segmented as year-oriented and topic-oriented analysis. The work reveals that public sentiment shifts based on events that trigger people’s opinions. Based on some events, temporal pattern changes from 2017 to 2023 both for the news topic and public sentiment. The work identifies that people are more likely to comment on negative news articles compared to others. Some frequent users are found whose sentiment polarity is mostly positive.