Thinking aloud or screaming inside: exploration of sentiment around work

Journal article

Background: Millions of workers suffer from work-related ill health every year. The loss of working days often accounts for poor well-being due to discomfort and stress caused by the workplace. The ongoing pandemic and post-pandemic shift in socio- economic and work culture can continue to contribute to work-related adverse sentiments. Critically investigating the state-of-the-art technologies, this paper identifies the research gaps in recognising workers’ need for well-being support and we aspire to understand how such evidence can be collected to transform the workforce and workplace.
Objectives: Building upon recent advances in sentiment analysis, this paper aims to closely examine the potential of social media as a tool to assess workers’ emotions towards the workplace.
Methods: This paper collects a large Twitter dataset comprising both pandemics as well pre-pandemic tweets, facilitated through a Human-in-the-loop approach, in combination with unsupervised learning and meta-heuristic optimisation algorithms. The raw data pre-processed through natural language processing techniques are assessed by a generative statistical model and lexicon-assisted rule-based model, mapping lexical features to emotion intensities. This paper also assigns human annotation and performs work-related sentiment analysis.
Results: A mixed method approach, including topic modelling using Latent Dirichlet Allocation, identified the top topics from the corpus to understand how Twitter users engage with discussions on work-related sentiments. Assorted aspects have been portrayed through overlapped clusters and low inter-topic distances. Further analysis comprising VADER, suggests a smaller number of negative polarities, however, among diverse subjects. In contrast, the human-annotated dataset created for this paper contains more negative sentiments. In this paper, sentimental juxtaposition revealed through the labelled dataset is supported by the n-gram analysis as well.
Conclusion: The developed dataset demonstrates that work-related sentiments are projected on social media, which offers an opportunity to better support the workers. The infrastructure of the workplace, nature of the work, the culture within the industry and the particular organisation, employers, colleagues, person-specific habits, and upbringing all play their part in the health and wellbeing of any working adults that contribute to the productivity of the organisation. Therefore, understanding the origin and influence of the complex underlying factors, both qualitatively and quantitatively, can inform the next generation workplace to drive positive change by relying on empirically grounded evidence. Therefore, this paper outlines a comprehensive approach to capture deeper insights into work-related health.

Authors: Hoque Tania, M; Hossain, R; Jahanara, N; Andreev, I; Clifton, D

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: JMIR Publications
• Journal: JMIR Formative Research
• Volume: 6
• Issue: 9
• Article number: e30113
• Publication date: 2022-09-30
• Acceptance date: 2022-08-10
• DOI: 10.2196/30113
• ISSN: 2561-326X
• Language: English
• Keywords: artificial intelligence; FFR; natural language processing; sentiment analysis; occupational health; bayesian inference; machine learning; work-related mental health