Crowd Sourcing Design Preferences from Social Media using Natural Language Processing: Insights on Green Infrastructure Design
Abstract
Human modifications to natural landscapes, such as urbanization and loss of tree canopy, are increasingly recognized as contributors to problems such as increased urban heat, flooding, and associated human health impacts. Designing human landscapes that better mimic natural landscapes, such as through green stormwater infrastructure that uses vegetation to absorb water and reduce heat, can reduce these impacts. However, gaining public acceptance of more natural landscapes, which are often described as "weedy" or mosquito breeding, requires a better understanding of human landscape preferences. Existing research on landscape preferences typically uses resource-intensive surveys and questionnaires, where study participants score or evaluate photos or, more recently, through image-based machine learning approaches (Rai, et al., 2019). Since almost 70% of US adults use social media to connect with friends and families, or to follow news and topics of interest (Pew Research, 2015) research is needed to identify whether social media postings provide useful information about preferences for landscape settings. In this work, we label text comments from Flickr, Instagram, and Twitter to train a lexicon-based sentiment classification model that predicts human sentiments (preferences) about green infrastructure landscapes. The results show a 77% correlation between text-based sentiments from social media and image based green infrastructure preferences. Considering the variability of human preferences, these results show significant promise for the use of text comments in social media to identify landscape design preferences.
References: Pew Research Center surveys (2015) Rai, A., Minsker, B., Sullivan, W. & Band, L. (2019), A novel computational green infrastructure design framework for hydrologic and human benefits. Environmental Modelling & Software, 118, 252-261. https://doi.org/10.1016/j.envsoft.2019.03.016.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2020
- Bibcode:
- 2020AGUFMIN0140006R
- Keywords:
-
- 1920 Emerging informatics technologies;
- INFORMATICS;
- 1938 Knowledge representation and knowledge bases;
- INFORMATICS;
- 1954 Natural language processing;
- INFORMATICS;
- 1958 Ontologies;
- INFORMATICS