Mapping ideas from cyberspace to realspace: visualizing the spatial context of keywords from web page search results

Figures & data

Table 1. Three major geolocation methods for mapping web content and social media.

Methods	Accuracy/spatial resolution	Spatial data availability	Implementation requirement
IP geolocation	US: Zip code level (92119) or city level (10–100 km). International: city level (Taipei, New York, etc.) or country level	All web pages have IP addresses. We can label 90% of web pages with geotags using IP geocoding methods	WHOIS databases (public), commercial or free IP geolocation databases or web services (IPPAGE.org, MaxMind.com)
Mobile device tracking (GPS, Wi-Fi, or cellular signals)	• GPS: 8 meters (average median error) • Wi-Fi: 74 meters • Cellular: 600 meters	Public: Twitter has only 1% of tweets with geolocation enabled. Private: mobile phone companies have user location data, but they are confidential	Smart phones or mobile devices with assisted GPS functions or Wi-Fi enabled
Geographic context analysis (gazetteers, geographic names)	Spatial footprint resolutions range from 1 to 1000 km, to 3000 km	10% web pages have US zip code information; 20% have geographic identifiers (Himmelstein 2005)	Challenging. Requires a comprehensive framework with geographic name ontology, gazetteers, spatial reasoning tools

Figure 1. The SWARMS framework.

Figure 2. The SWARMS prototype interface for the keyword search of ‘Jerry Sanders’ and the output of the top 978 web pages from the Yahoo search engine.

Figure 3. An example of geocoded web information databases (top) and visualization maps (using ‘Jerry Sanders’ as the keyword search in Yahoo).

Figure 4. Creating a web information landscape (using kernel density method based upon the modified web page ranks) for the ‘Jerry Sanders’ search result web pages (red dots). GIS parameters: 3 map unit threshold (radius), 0.5 map unit output resolution (1 map unit equals 1 decimal degree).

Figure 5. The background web information landscape (by using 56,000 web pages for kernel density map).

Figure 6. The differential web information landscape of ‘Jerry Sanders’ versus Background Information Landscape (with Yahoo Search Results).

Figure 7. The Differential Web Information Landscapes of three city mayors (McGinn, Emanuel, and Adams).

Figure 8. The global distribution patterns of the keyword search ‘Osama bin Laden’ in three different languages (English, Chinese (simplified), and Arabic).

Himmelstein, M. 2005. “Local Search: The Internet is the Yellow Pages.” Computer, 38 (2): 26–34. doi: 10.1109/MC.2005.65.

 Web of Science ®Google Scholar