Using Machine Learning and Twitter Data to Evaluate the Public Opinion of Environmental Policy
The views expressed are those of the author and do not necessarily reflect the views of ASPA as an organization.
By April Heyward
January 12, 2022
Public administration has the ongoing challenge of solving innumerous diverging problems simultaneously, while attempting to balance the expectations of diverse stakeholders. For the last 24 months, public administration has been combating the COVID-19 pandemic that has recently evolved with the Omicron variant. COVID-19 continues to dominate due to its complexity. Simultaneously there are other complex problems, such as climate change—a global multifaceted environmental issue with diverse stakeholders and expectations. The 2021 United Nations Climate Change Conference, also referred to as COP26, was held in Fall 2021 to primarily focus on combating climate change and to formally prescribe a path forward on addressing climate change. One of the outcomes of COP26 is the Glasgow Climate Pact. April Heyward is researching environmental policy and interventions and employing machine learning and Twitter data to evaluate the public opinion of climate change, extreme weather, climate action, COP26, and the Paris Agreement. Public administration can employ computational methods such as machine learning to evaluate public opinion and policy interventions which serves as a nod to computational social science.
Environmental policy is an intervention employed primarily by governments and public entities globally to address environmental issues (e.g., climate change, global warming, droughts and rising sea levels). This policy area is very complex and requires global collaborations, partnerships and participation. Environmental policy emerged in the late twentieth century and has undergone at least three phases of development. Initially, governments employed action as an emergency response and then increased actions to protect nature resources. Climate change’s continuing evolution sparks public discourse between stakeholders who are diametrically opposed with competing interests. The impact of environmental policy extends to other policy areas such as health policy, economic policy and social policy. There are challenges with employing environmental policy to include availability and commitment of resources; regulation development and enforcement; garnering support and participation from key stakeholders; and developing collaborations and partnerships. Environmental issues can cross national borders and cannot be solved in silos.
Machine learning is a fascinating subdiscipline of artificial intelligence. Its vastness facilitates natural language processing to evaluate public policy and public opinions, which is a value add to public administration. Heyward’s primary research question is: What is the public opinion of climate change and the environmental policies that address climate change? To date, Heyward collected close to 1 million environmentally related tweets via primary data collection to answer the primary research question. Heyward started extracting environmentally related Twitter data via RStudio (an Integrated Development Environment for the R Programming language) with the rtweet package developed by Michael Kearney to study public opinions in October 2021. Retweets were excluded from data extraction. Datasets are created with each extraction and saved as .CSV files. Heyward’s data collection contains 97 datasets to date on varying environmental areas and a sample of the data collected will be presented. Prior to data analysis, datasets were merged by topic in RStudio in a matter of minutes with the execution of a few lines of R code. The text of tweets was cleaned in RStudio by removing punctuations, http links, stopwords, digits, other symbols and capital letters, which are transformed to lowercase letters for preparation of analysis. Employing varying lexicons—dictionaries that ascribes words to sentiments and emotions—can extend understanding of data analysis. The Bing Lexicon developed by Bing Lui and NRC Lexicon developed by Saif Mohammad and Peter Turney were employed in the analysis of climate action tweets, Paris Agreement tweets and COP26 tweets.
There are numerous methods for evaluating tweets to include the analysis of sources of tweets, positive and negative emotions found in tweets and words that contribute to positive and negative emotions. Climate action tweets were analyzed, and the data identified 135 sources of tweets which depicted how climate action tweets were posted. Examples of the sources of tweets included Twitter Web App, Twitter for iPhone and TweetDeck.
See Figure 1 for the Top Tweet Sources of Climate Action No Hashtag Tweets Donut Chart.
See Figure 2 for Words Contributed to Positive and Negative Emotions Found in Climate Action No Hashtag Tweets (Bing Lexicon).
The top words contributing to positive emotions were appeal, support, sustainable and clean. The top words contributing to negative emotions were emergency, crisis, urgent and failed. The Paris Agreement is an international treaty on climate change that was negotiated by countries around the globe.
See Figure 3 for Emotions Analysis of October 20-November 11, 2021 Paris Agreement No Hashtag Tweets.
The sample data indicated significantly more positive emotions about the Paris Agreement than negative emotions. Trust, anticipation and joy were the top positive emotions and fear, sadness and anger were the top negative emotions. COP26 was held in Fall 2021.
See Figure 4 for Words Contributed to Positive and Negative Emotions Found in COP26 Hashtag Tweets (NRC Lexicon).
The sample data indicated more positive emotions about COP26 than negative emotions. The top words contributing to positive emotions were action, time, united and green and the top words contributing to negative emotions were change, crisis, demand and unknown. Machine learning is a method for public administration to adapt and employ from the computer science discipline.
Author: April Heyward is an Author for PA Times, Public Sector Practitioner, 4th Year Doctor of Public Administration Student at Valdosta State University, and a R Programmer. She can be reached at [email protected] and followed on Twitter @ heyward_april. All opinions and views are her own and does not reflect the views and opinions of her affiliations.
(No Ratings Yet)
Loading...
Using Machine Learning and Twitter Data to Evaluate the Public Opinion of Environmental Policy
The views expressed are those of the author and do not necessarily reflect the views of ASPA as an organization.
By April Heyward
January 12, 2022
Public administration has the ongoing challenge of solving innumerous diverging problems simultaneously, while attempting to balance the expectations of diverse stakeholders. For the last 24 months, public administration has been combating the COVID-19 pandemic that has recently evolved with the Omicron variant. COVID-19 continues to dominate due to its complexity. Simultaneously there are other complex problems, such as climate change—a global multifaceted environmental issue with diverse stakeholders and expectations. The 2021 United Nations Climate Change Conference, also referred to as COP26, was held in Fall 2021 to primarily focus on combating climate change and to formally prescribe a path forward on addressing climate change. One of the outcomes of COP26 is the Glasgow Climate Pact. April Heyward is researching environmental policy and interventions and employing machine learning and Twitter data to evaluate the public opinion of climate change, extreme weather, climate action, COP26, and the Paris Agreement. Public administration can employ computational methods such as machine learning to evaluate public opinion and policy interventions which serves as a nod to computational social science.
Environmental policy is an intervention employed primarily by governments and public entities globally to address environmental issues (e.g., climate change, global warming, droughts and rising sea levels). This policy area is very complex and requires global collaborations, partnerships and participation. Environmental policy emerged in the late twentieth century and has undergone at least three phases of development. Initially, governments employed action as an emergency response and then increased actions to protect nature resources. Climate change’s continuing evolution sparks public discourse between stakeholders who are diametrically opposed with competing interests. The impact of environmental policy extends to other policy areas such as health policy, economic policy and social policy. There are challenges with employing environmental policy to include availability and commitment of resources; regulation development and enforcement; garnering support and participation from key stakeholders; and developing collaborations and partnerships. Environmental issues can cross national borders and cannot be solved in silos.
Machine learning is a fascinating subdiscipline of artificial intelligence. Its vastness facilitates natural language processing to evaluate public policy and public opinions, which is a value add to public administration. Heyward’s primary research question is: What is the public opinion of climate change and the environmental policies that address climate change? To date, Heyward collected close to 1 million environmentally related tweets via primary data collection to answer the primary research question. Heyward started extracting environmentally related Twitter data via RStudio (an Integrated Development Environment for the R Programming language) with the rtweet package developed by Michael Kearney to study public opinions in October 2021. Retweets were excluded from data extraction. Datasets are created with each extraction and saved as .CSV files. Heyward’s data collection contains 97 datasets to date on varying environmental areas and a sample of the data collected will be presented. Prior to data analysis, datasets were merged by topic in RStudio in a matter of minutes with the execution of a few lines of R code. The text of tweets was cleaned in RStudio by removing punctuations, http links, stopwords, digits, other symbols and capital letters, which are transformed to lowercase letters for preparation of analysis. Employing varying lexicons—dictionaries that ascribes words to sentiments and emotions—can extend understanding of data analysis. The Bing Lexicon developed by Bing Lui and NRC Lexicon developed by Saif Mohammad and Peter Turney were employed in the analysis of climate action tweets, Paris Agreement tweets and COP26 tweets.
There are numerous methods for evaluating tweets to include the analysis of sources of tweets, positive and negative emotions found in tweets and words that contribute to positive and negative emotions. Climate action tweets were analyzed, and the data identified 135 sources of tweets which depicted how climate action tweets were posted. Examples of the sources of tweets included Twitter Web App, Twitter for iPhone and TweetDeck.
See Figure 1 for the Top Tweet Sources of Climate Action No Hashtag Tweets Donut Chart.
See Figure 2 for Words Contributed to Positive and Negative Emotions Found in Climate Action No Hashtag Tweets (Bing Lexicon).
The top words contributing to positive emotions were appeal, support, sustainable and clean. The top words contributing to negative emotions were emergency, crisis, urgent and failed. The Paris Agreement is an international treaty on climate change that was negotiated by countries around the globe.
See Figure 3 for Emotions Analysis of October 20-November 11, 2021 Paris Agreement No Hashtag Tweets.
The sample data indicated significantly more positive emotions about the Paris Agreement than negative emotions. Trust, anticipation and joy were the top positive emotions and fear, sadness and anger were the top negative emotions. COP26 was held in Fall 2021.
See Figure 4 for Words Contributed to Positive and Negative Emotions Found in COP26 Hashtag Tweets (NRC Lexicon).
The sample data indicated more positive emotions about COP26 than negative emotions. The top words contributing to positive emotions were action, time, united and green and the top words contributing to negative emotions were change, crisis, demand and unknown. Machine learning is a method for public administration to adapt and employ from the computer science discipline.
Author: April Heyward is an Author for PA Times, Public Sector Practitioner, 4th Year Doctor of Public Administration Student at Valdosta State University, and a R Programmer. She can be reached at [email protected] and followed on Twitter @ heyward_april. All opinions and views are her own and does not reflect the views and opinions of her affiliations.
Follow Us!