Leveraging aspect-based sentiment analysis for trend and polarity identification on immigration issues in Singapore

• Main Author: Fang, Yuan
• Other Authors: Na Jin Cheon
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Social sciences::Communication
• Online Access: https://hdl.handle.net/10356/165118
• Institution: Nanyang Technological University

Singapore is an immigration country, its open immigration policy which allows for an influx of workers from different skill levels, undesirably brings anxiety about job competitiveness and national identity to the local native citizens. Hence, understanding the public’s attitudes toward immigration has significant meaning to policymakers to stabilize society while accelerating the economy with immigrants. Social science researchers usually design surveys or interviews to collect data and then analyse it, but this process takes a few months even longer to execute and sample size is constrained. On the other side, with the rapid development of the social network, public awareness and participation in popular or heated social topics have been largely improved, the advancement results in a tremendous number of opinionated texts being generated online every minute. These digital data provide unique values to social research, yet exceed the language processing capability of human beings. Thus, using computational methods to analyse, process, and reveal people’s sentiments hidden behind texts is becoming an emerging research area in the intersection area of computer science and social science. Extensive research of sentiment analysis at document-level and sentence-level existed, however, coarse-grained sentiment analysis doesn’t capture the multi-dimensional issues associated with immigration. In order to overcome these gaps, a fine-grained aspect-based sentiment analysis (ABSA) leveraging a BERT-based deep learning model is proposed. Drawing on a data set of 23, 244 comments and natural language processing, the results discovered that immigration-related comments increased along with the COVID-19 cases during year 2020 – 2022; overall public opinions tended to be negative, especially on the aspects of “foreign talent” and “foreign domestic worker”; the pandemic outbreak has had negative social consequences. These revealed trend and polarity of public opinions can vi be seen as social feedback on policies, which could be leveraged by the Government to monitor the changes of public sentiments over time and develop the nuanced adjustments to migration and integration policies. Furthermore, the discovery captured the emerging scenarios and new insights which were not studied in previous surveys. For example, impact of “foreign talent” on social cohesion has been studied extensively by social scholars, but research on the impact of “foreign domestic worker” is considered far lesser. These new insights evidenced that the research interests of computer science and social science should be continuously encouraged to connect, and the intersection could be a powerful tool for understanding and improving the social world around us.