Sentiment Analysis for COVID-19 National Vaccination Policy
Muneeb and Chandler suggest that the ability to assess constituent sentiment, within specific geographies or demographic communities, can be critical to executing vaccine distribution initiatives. For example, in preparing a distribution strategy, healthcare providers and local government health agencies require insight into general feelings about receiving a vaccination, including within specific demographic or geographic communities. Then, as the programs roll out, sentiment analysis can help to track changes in beliefs and behaviors, as well as the success of the vaccination initiative. Despite a wealth of methods for collecting data, many policymakers have been unable to access and harness data during the pandemic. Researchers and policymakers should start laying the groundwork now for emergencies of the future, developing data-sharing agreements and privacy-protection protocols in advance to improve reaction time for such deployments in the future.
Introduction
Given healthcare decisions are personal – one of the largest hurdles faced by governments in emerging economies such as Pakistan is getting individuals to agree to receive the COVID-19 vaccine. Traditionally, health departments at the state level had the option to conduct a sentiment analysis for each specific program using polls and surveys before a vaccination campaign began, and then revisit that intelligence only a few times afterwards. However, with the scale of the vaccine distribution program and the rise of digital feedback tools, healthcare providers and local government health agencies can benefit from far more immediate and deeper understandings of constituent concerns.
That is because constituents today engage with government organizations across a wide variety of communications platforms, including call centers, websites, mobile apps, social media, search engines, and news feeds. One way to do this is by conducting a sentiment analysis which makes the use of natural language processing (NLP), text analysis, computational linguistics, and biometrics to systematically identify, extract, quantify, and study affective and subjective states of people and other personal information from a wide variety of communication platforms. For computer scientists and researchers, big data are valuable assets for understanding people’s sentiments regarding current events, especially those related to the pandemic. Therefore, analyzing these sentiments can help yield remarkable findings.
Our Sentiment Analysis Research
Our research conducted back in February 2021 consisted of conducting a sentiment analysis – which first began by implementing a methodology known as the “domain analysis.” This methodological approach make use of various machine learning techniques and augmented intelligence platforms to create an ecosystem that can analyze massive amounts of data from OSNIT, alternative and internal data to find patterns between products, events, markets, trends, and people, before deciding what to prioritize.
Open Source Intelligence (OSINT) can be any data or intelligence that is freely available. Some examples include news articles, social media, economic, governmental & market data or open-sourced corporate data such as Google Search Trends. An excellent example how the process works is visible in our research in the context of Pakistan where we first began the process by using advanced natural language processing (NLP) and network analytics to cluster OSINT documents that mention the “COVID-19 Pakistan” domain. The network map illustrated below consists of a wide number of stories which we group into clusters based on their magnitude.
Out of the initial 3,225 stories visible in our Network Map, we further decided to refine our search results by identifying a total of 1,718 stories visible in Figure 2. We choose the top story clusters on the basis of their relevance in the context of our research. Finally, Figure 3 shows the end result and the added value of conducting a sentiment analysis. The analysis allows government agencies to use sentiment analysis to identify prevailing opinions of the population (positive, neutral, negative). They can also assess the magnitude of sentiments, and get updates as perception changes.
This allows agencies to track these changes over time, and see changes related to the topics being discussed. When we conducted our research back in February, the cluster which had shown that the most favorable sentiment within the COVID 19 domain – that is the cluster which people felt most positive about – was the trade deal between China and Pakistan which allowed the latter to import vaccinations from the former (70% Positive Sentiment). However, our results show that this development did not translate into immediate optimism regarding a mass-scale Vaccine Drive (51% Positive Sentiment). This was likely the result of people being doubtful that a wide mass-scale vaccination drive would begin immediately given a trade deal was struck just a few days prior to our conducting the sentiment analysis.
Conclusion
A sentiment analysis is just one tool governments have at their disposal in fighting the spread of COVID-19 and may prove to be one of the most useful in helping state and local administrative units with their vaccine distribution efforts at scale. While our research only shows a rather preliminary and basic conceptualization of how the sentiment analysis works – there is the potential to use the sentiment analysis in a more focused and targeted manner by policymakers in Pakistan.
The ability to assess constituent sentiment, within specific geographies or demographic communities, can be critical to executing vaccine distribution initiatives. For example, in preparing a distribution strategy, healthcare providers and local government health agencies require insight into general feelings about receiving a vaccination, including within specific demographic or geographic communities. Then, as the programs roll out, sentiment analysis can help to track changes in beliefs and behaviors, as well as the success of the vaccination initiative.
Despite a wealth of methods for collecting data, many policymakers have been unable to access and harness data during the pandemic. Researchers and policymakers should start laying the groundwork now for emergencies of the future, developing data-sharing agreements and privacy-protection protocols in advance to improve reaction time for such deployments in the future.
About the authors:
Muneeb Sikandar works at International Group for Artificial Intelligence
Chandler T Wilson works with Datavest Partners
