Pleasant Urban Experiences: Sentiment Analysis of Geo-Coded Social Media Data

Dr Huang, Jianxiang   (Principal Investigator (PI))

Dr Li Lishuai   (Co-Investigator)

5

2016-07-01

39000

Pleasant Urban Experiences: Sentiment Analysis of Geo-Coded Social Media Data

High-Density City, Housing Estates, Sentiment Analysis, Social Media

Building and Construction,Environmental

N/A

Seed Fund for Basic Research for Resubmission of GRF/ECS Proposals

2016

Completed

The abstract for the proposal to be submitted on the 2017/2018 exercise is included below: The objectives of the proposed study lies three-fold: 1) to quantify the impact of urban form on occupant mood & behaviours; 2) to examine classic theories on urban design & place-making; 3) to formulate new theories of place-making in the digital age. The study addresses the question of what kind of cities are conducive to pleasant urban experiences? Classic urban planning and design literature covered extensively this topic yet scientific evidences are rare. The rise of social media offered new opportunities to study wellbeing in the built environment. Posts with geographical coordinates, time stamp, and often user profiles, contains rich information about local communities. The social impact of the built environment, such as transit infrastructure, urban renewal, and gentrification, can be monitored continuously. However, there are no studies to our knowledge that used social media data to study the built environment. Previous studies uses phone call and text messaging activities as proxies (De Nadai et al., 2016) without knowledge of the content of communication. Political studies rely on sentiment analysis (source), which loses important dimensions other than happiness / depressed. Keywords search are used to predict influenza(Nagar et al., 2014); the richness of social media data are yet to be explored further. We hypothesize that classic theories on place-making can be re-examined in the digital age using geo-coded social media data (behavior and mood) as well as GIS information. Methodological-wide, we will conduct a quasi-experiment looking at social media characteristics before and after major interventions. Instead of using OLS regression models, we will apply Principle Component Analysis to deal with the multi-dimensionality of the social media dataset. The work will deliver a monitoring tool for urban planning and management. The outcome will enrich the growing body of literature on urban informatics. Hong Kong has the highest penetration of smart phone and social media use. The city has undergone recent development and redevelopment in the digital age, allowing us to examine the before-after social media characteristics of major changes in the built environment.