September 20, 2013 By Matthew McClellan and Stephen Goldsmith, Data-Smart City Solutions
The Internet is full of public opinions, both expert and novice. Yelp, Facebook, Twitter, Tumblr, and other social media outlets curate public opinion to inform other patrons in the decision-making process. Many of these communications take the form of reviews, ratings, and recommendations – including those on local government services. Each of these opinions is a potential unstructured data point. Sentiment analysis, also called opinion mining, seeks to make sense of this broad landscape of informal opinions.
Living Cities and the Ash Center recently hosted a convening that brought together mayors’ chiefs of staff, city innovation and information officers, and local government experts from across the country to explore the promise of 311 to change local government through the power of civic engagement. The group identified sentiment analysis as one approach to dramatically change governments’ responsiveness and the use of civic participation in policy development and implementation.
Sentiment analysis software automatically extracts opinions from a piece of text. Review aggregators like Rotten Tomatoes and Metacritic determine whether reviews are positive or negative and, in some cases, assign a score based on tone. Scores can then be aggregated to give a general sense of critical and public opinion of the movie, restaurant, game, or other product. In typical reviews, feedback on restaurant quality may be classified in terms of food quality, portion size, atmosphere, staff friendliness, and more. Some government services that resemble traditional customer service interactions, like a visit to get a driver’s license at the registry of motor vehicles, or other permitting applications, can be evaluated along similar lines.
But not all government services feel like familiar customer service relationships, and feedback on them might not look like a product review. In some cases, residents only notice the service when it is late or lacking, like when trash and recycling are left uncollected or their streets go unplowed. Others evoke overwhelmingly positive responses, like fire and emergency medical services.
Washington D.C. is the first municipal government to adopt a public-facing sentiment analysis initiative, Grade.DC.gov, and their Fire and EMS (FEMS) have received uniformly positive monthly grades. Naturally, feedback from those whose loved ones, homes, and neighborhoods have been saved by FEMS has been overwhelmingly positive. It is important to remember that resident reviews of government services are not the only, objective measure of agency quality, but a valuable reflection of public perception.
Smartphone ownership has jumped 20% since 2011, with a majority of Americans now carrying devices connected to the Internet.
Of course, government services are already reviewed on comment cards or surveys. Sentiment analysis does more than simply confirm or shatter previously held assumptions about public opinion. This new approach expands the number and types of people involved in giving feedback.
Phone calls are still the primary way many residents contact their city governments, but it is no longer the only way. According to Susan Crawford’s case study on the Mayor’s Office of New Urban Mechanics in Boston, about 60 percent of resident communications with City Hall go through the Mayor’s Hotline, with the remaining 40 percent mostly split through the City Hall website and the CitizensConnect smartphone app.
Social media comments may not yet provide a comprehensive sample of public opinion, but they do capture the voices of groups who have not always been engaged by traditional feedback forums. According to Pew Research, 15 percent of American adult Internet users also use Twitter. Urban residents use Twitter at a higher rate, with young adults (26 percent) and African-Americans (28 percent) Tweeting substantially more. While home broadband adoption has stalled in the last three years, smartphone ownership has jumped 20 percent since 2011, with a majority of Americans now carrying devices connected to the Internet.
While Tweets and other social media represent a small minority of communications in Boston, the strong engagement with CitizensConnect shows that residents are quick to embrace their smartphones as effective means of communicating with their government. Most cities do not yet have an application as robust as CitizensConnect, and they might find similar feedback by mining the social media landscape.
By mining social media, governments can harness feedback from residents even when the residents don’t reach out directly to the government. One challenge, then, is for cities to improve the quality and volume of feedback by encouraging their residents to think of social media as channels to communicate with and about government. Finally, cities need to fulfill the promise of listening to these reviews by responding to constructive feedback and improving their services.
This article originally appeared on Data-Smart City Solutions.