The road to recovery: Sensing public opinion towards reopening measures with social media data in post-lockdown cities

Abstract

The COVID-19 pandemic has resulted in cities implementing lockdown measures, causing unprecedented disruption (e.g. school/shop/office closures) to urban life often extending over months. With the spread of COVID-19 now being relatively contained, many cities have started to ease their lockdown restrictions by phases. Following the phased recovery strategy proposed by the UK government following the first national lockdown, this paper utilises Greater London as its case study, selecting three main reopening measures (i.e., schools, shops and hospitality reopening). This paper applies sentiment analysis and topic modelling to explore public opinions expressed via Twitter. Our findings reveal that public attention towards the reopening measures reached a peak before the date of policy implementation. The attitudes expressed in discussing reopening measures changed from negative to positive. Regarding the discussed topics related to reopening measures, we find that citizens are more sensitive to early-stage reopening than later ones. This study provides a time-sensitive approach for local authorities and city managers to rapidly sense public opinion using real-time social media data. Governments and policymakers can make use of the framework of sensing public opinion presented herein and utilise it in leading their post-lockdown cities into an adaptive, inclusive and smart recovery.

Keywords: COVID-19; Public opinion; Recovery measures; Social media data; Twitter; Urban management.

Fig. 1

Daily trends of COVID-19 confirmed cases and implementation of reopening measures in London (Data Source: Public Health England).

Fig. 2

Research framework of sensing public opinion regarding reopening measures from social media data.

Fig. 3

The proportion of reopening discussions under COVID-19 discussions in London (Data source: Twitter Streaming API).

Fig. 4

Daily trends of three reopening measures related tweets in London (Data source: Twitter Streaming API). Note: blue line – schools reopening, orange line – shops reopening, red line – hospitality reopening. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5

Overview of the sentiment of tweets in London (Data source: Twitter Streaming API). Note: blue line – COVID_19 tweets, red line –reopening tweets, black line – random tweets. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 6

Daily trends of discussions on three reopening measures in London (Data source: Twitter Streaming API). Note: blue line – schools reopening, orange line – shops reopening, red line – hospitality reopening. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 7

Four topics and terms frequency for school reopening discussion found in tweets. On the left panel, each circle refers to a topic; circle sizes denote the importance of the topic over the corpus (i.e., the larger the circle, the more pertinent the discussion). The distance between circles indicates the similarity between topics. On the right panel, a horizontal bar chart lists the top 30 most relevant terms with the overall term frequency.

Fig. 8

Five topics and terms frequency for shop reopening discussion found in tweets. On the left panel, each circle refers to a topic; circle size denotes the importance of the topic over the corpus. The distance between circles indicates the similarity between topics. On the right panel, a horizontal bar chart lists the top 30 most relevant terms with the overall term frequency.

Fig. 9

Three topics and terms frequency for hospitality reopening discussion found in tweets. On the left panel, each circle refers to a topic; circle size denotes the importance of the topic over the corpus. The distance between circles indicates the similarity between topics. On the right panel, a horizontal bar chart lists the top 30 most relevant terms with the overall term frequency.