This document discusses geocoding in geographic information retrieval systems. It begins by defining geocoding as assigning geographic coordinates to recognized locations. It then discusses how geocoding can involve addresses, objects, people, or animals/habitats. The document also notes that geocoding is sometimes called georeferencing or geotagging. It provides examples of how search engines like Google and Bing georeference the location of Knoxville, Tennessee. Other topics covered include gazetteers, ontologies for geographic information retrieval, and the need for specialized gazetteers and ontologies to enable spatial queries in domain-specific applications.

1. Geocoding in Geographic
Information Retrieval Systems
UT-Knoxville Geography Research Symposium
02/22/2014
Tanner Jessel
School of Information Sciences

2. What is Geocoding?
➲ Geography + Code = Geocode
 Place + Space
 Assigning geographic coordinates to a recognized
location.
 Location + Spatial Coordinates

3. What is Geocoding?
Means different things to different
practitioners
 Addresses
 Objects
 People
 Animals / Habitats

4. What is Geocoding?
 Sometimes referred to as “Georeferencing” or
“Geotagging.”
 1955 – 2008 Google “ngram” illustrates trend
 Interdisciplinary

5. Geocode (Blue) Georeference (Red)

6. Interdisciplinary

7. Georeferencing
Knoxville, Tennessee – what do you think of?

8. Georeferencing in IR
Knoxville, Tennessee – what Google thinks

9. Georeferencing in IR
Knoxville, Tennessee – what Bing thinks

10. Georeferencing in IR
Knoxville, Tennessee – what Bing thinks you
think

11. Georeferencing in IR
➲ How do we answer this question?
 What do we mean by “Edinburgh”
 What do we consider “near”
Query: what castles are near Edinburgh?

12. Georeferencing in IR
➲ Which castle?
 Scotland?
 Jamaica?

13. Georeferencing in IR
➲ What is “near?”
 1 km?
 3 km?
 20 km?

14. Georeferencing in IR
➲ Match text to known “footprint,” or space an
object occupies
 Castle footprint is known
 City footprint is known
 Country footprint is known

15. Geoparsing
➲ Evaluate the geographic context of an item
➲ “Castles near Edinburgh”
 A physical place
 Implied location
 Edinburgh
 Scotland
 United Kingdom
 Europe

16. Geoparsing
➲ Possible castle names
 Craigmillar Castle
 Edinburgh Castle
 Lauriston Castle
 Tantallon Castle
 Chricton Castle
 Blackness Castle
 Dirleton Castle

17. Gazetteer
➲ Links a name to a footprint (point or area)

18. Gazetteer
➲ Links a name to a footprint (point or area)

19. Gazetteer
➲ Geographical Index or Dictionary

20. Ontology for Geo IR
➲ Pragmatic Ontology: how entities can be
grouped in a hierarchy
➲ Nation
➲ State
➲ County
➲ City
➲ Neighborhood

21. Ontology for Geo IR
➲ Semantic Ontology
➲ words have different meanings to different
groups
 “Skin Temperature” of Atmospheric Scientists
 “Surface Temperature” of Marine Scientists
 Both are same data point, but mean different things to
different communities of scientists
➲ Technology can handle these differences
➲ “Semantic” mediation
➲ Applies to spatial domain

22. Real-World Ontology

23. Ontology to Computer

24. Metadata: Machine Friendly
➲ Place
➲ Time
➲ Geographic Coordinates
➲ Geographic Footprint (Area coverage)
➲ Relationships
 Parent
 Child
 Similarities

25. Geographical Information Retrieval
System
➲ Spatially Aware
➲ Can handle spatial queries
 Castles
 Within 1 KM of
 Edinburg (Scotland)

26. Geographical Information Retrieval
System

27. Geographic Information Retrieval
System Application

28. Geocoding Toolkit
➲ Research effort reviewed 53 citations
 Extracted tools mentioned across citations
➲ Categorized tools:
Geographic IR Tools Disambiguation Tools
Georeferencing Tools Error Calculation Tools
Ontology Tools Algorithmic and Programming
Tools
Geographic IR Systems Visualization Tools
Database Tools Thesauri Tools
Extraction Tools

29. Need?
➲ Tools for automation of geocoding /
georeferencing

30. Next Steps
➲ Specialized gazetteers for subject areas are
needed
 These reliably link place to space
 They enable spatial queries
➲ Specialized ontologies for domain-specific
applications are needed
 These establish relationships between objects in the real and
virtual worlds
 They enable spatial queries

31. Questions?