Sentiment Analysis Disaster 2019 Website


February 28, 2019


For the third year, IBM and the U.S. Chamber of Commerce Foundation are collaborating to encourage companies in their corporate social responsibility (CSR) efforts by quantifying the brand value of CSR social media communications. For the first time, the analysis concentrates on the impact of social media conversations around disaster recovery and response, and leverages IBM’s Watson capabilities to analyze millions of social media snippets.IBM’s Watson™ technology is helping illustrate for companies the full value of CSR engagements in community disaster response.

You can listen to the recording of the webinar here

Report Highlights

  • Compared to any other type of activity, local and national disasters and tragedies had the biggest impact on the level of social media conversation about companies in the second half of 2017.
  • Corporate social responsibility (CSR) tends to be a positive high point in conversations on social media. Readers tend to react positively to these stories, compared to other topics.
  • Approximately 25% of neutral sentiment of a company can be shifted to positive by increasing the promotion of CSR project activities on social media to roughly 3% of all company media activities.
  • Most chatter, for community engagement generally and disaster response specifically, occurs on discussion boards where many users are anonymous.

The U.S. Chamber of Commerce Foundation (USCCF) works to support companies to lead important work in their communities. Over the past three years, we have collaborated with IBM to leverage their social media analysis to determine the brand impact of a company’s CSR work. The goal has always been to encourage companies to engage in CSR efforts and show the value they can obtain by doing it in the public eye.

Disaster response and recovery work is one of the important areas of work that companies engage in. When a disaster strikes, many companies get involved through philanthropic efforts, in-kind immediate support, and long-term recovery planning.

Many companies engage in a significant amount of disaster resilience and response work, but few companies publicize it. Companies want to avoid the perception of benefiting from suffering, so they often conduct the work in private or with no communications strategy. However, this attitude, particularly in the disaster response space, can be counterproductive and stall recovery.

Conducting work quietly stops companies from receiving constructive insights or criticisms on what they are doing, leaving out an important feedback loop from stakeholders. Also, many companies wait to benchmark their own efforts after a disaster, by using tools like the USCCF Corporate Aid Tracker. If companies are quiet about their disaster recovery and resilience work, it does not encourage other companies to carry out their own.

IBM and USCCF want to inspire more companies to share more of their work in the CSR space. By analyzing effective CSR communication strategies using social media analysis, particularly around disaster response programming, companies can encourage greater recovery efforts and improve their brand value and reputation.

Knowing how the public perceives these efforts is an important piece of information for companies as they frame the structure of their disaster resilience and recovery programs. Social media analytics is one of the best ways for companies to capture that data.

Social media analytics gathers data from social media to help companies make more informed business decisions. Natural language processing (NLP) provides a powerful method to analyze, interpret, understand, and influence what people discuss about a company on social media.

This year we included an additional focus to the nearly 800,000 snippets we collected. One of the enhancements to the IBM tools includes a greater ability to place sentiment in context. Instead of relying solely on the sentiment of a statement or paragraph, Watson™ NLU scores the sentiment of a statement (“entity”) in the context of a paragraph and in the context of the entire webpage (document). That scale ranges from -1 (most negative) to 1 (most positive).

This process allows us to analyze differentiations between a topic and the context in which it is presented. For instance, when someone reads an online article about a hurricane, the negative emotions brought on by the topic of the disaster tend to overpower any mention of corporate engagement on the ground. This is quantified in the chart below, which takes the average entity and document score across each CSR type.

This shows that even though the sentiment scores for disaster are low, there is a significant difference between an online conversation about a disaster and a company’s work to address it, much more so than any other type of CSR intervention.

Disaster response and recovery efforts are one of the most common types of CSR activity. We can now offer greater insights and actionable analyses to companies who want to share their disaster response work effectively. Our hope is that this encouragement of companies to do more, and talk about it effectively, will incentivize the entire private sector to engage more completely in their communities, particularly in times of crisis.

Download the full report.