Thursday 20 March 2025
As the world grapples with the challenges of the COVID-19 pandemic, corporations have been scrambling to demonstrate their commitment to corporate social responsibility (CSR). In an effort to gauge the effectiveness of these efforts, researchers have turned to natural language processing (NLP) techniques to analyze and summarize vast amounts of text data.
A recent study published in a leading academic journal has shed new light on this endeavor. The authors employed two prominent NLP methods – latent dirichlet allocation (LDA) and deep distributed representation (DDR) – to extract meaningful insights from corporate press releases, blogs, and other publicly available documents. Their goal was to identify the key themes and topics that underpin CSR initiatives during times of crisis.
The researchers began by collecting a large dataset of text documents related to CSR efforts made by top companies in various sectors during the early stages of the pandemic. They then applied LDA, a popular topic modeling technique, to uncover latent topics within the corpus of text data. LDA works by identifying patterns and relationships between words, allowing it to distill complex documents into meaningful themes.
The study’s findings revealed that CSR initiatives focused on employee welfare, company operations, and social responsibility were the most prominent during this period. Interestingly, the authors discovered that these themes became increasingly prominent as the pandemic progressed, suggesting a growing recognition of the importance of CSR in times of crisis.
Building upon these insights, the researchers next employed DDR, a technique that uses neural networks to represent words and phrases in high-dimensional vector space. This allowed them to identify key document sentences that best captured the essence of CSR initiatives.
The results were striking: by analyzing the top-ranked sentences extracted using DDR, the authors could summarize the abstract of corporate initiatives with remarkable accuracy. Moreover, they found that altering a single parameter – known as alpha value – could significantly alter the cut-off score for sentence selection, yielding desired lengths and levels of detail in the summary.
These findings have significant implications for policymakers, regulators, and consumers alike. By leveraging NLP techniques to analyze and summarize CSR initiatives, stakeholders can gain valuable insights into corporate behavior during times of crisis. This information can be used to inform decision-making, promote transparency, and hold companies accountable for their actions.
The study’s authors also highlight the potential applications of their approach in other domains. For instance, they suggest that similar techniques could be applied to analyze and summarize scientific literature, news articles, or even social media posts.
Cite this article: “Uncovering Corporate Social Responsibility during Times of Crisis: A Natural Language Processing Approach”, The Science Archive, 2025.
Covid-19 Pandemic, Corporate Social Responsibility, Natural Language Processing, Latent Dirichlet Allocation, Deep Distributed Representation, Topic Modeling, Employee Welfare, Company Operations, Social Responsibility, Crisis Management.
