GeoDirectory in Limerick City Council

1
GeoDirectory in Limerick City Council
Jason Murphy Gresham Hotel 22nd November 2007
Over 40 of the buildings in the state have a
non-unique address
(ComReg, Report on Postcodes, 2005)
2
Content

• Context
• What we have learned
• What we have achieved to date
• The Possibilities Considerations
• Questions

The sheer diversity of databases has hampered
data integration (Higgs and Martin, 1995)
3
National Context

• Multiple Functions in Local Authority
• Disparate Databases (and Apps)
• Possibilities for Address DB Standardisation
  (semantic)
• Client, Property, Area-Centric Options
• 1998 GeoDirectory Standardised and Current.
• Tools of the Local Authority/ LGCSB/ wider
  industry partners
• RESEARCH Identify a Benchmark

4
Why Consider the Topic?

• In the context of BLG improved data quality can
  improve customer interactions, reducing costs
  and streamlining the decision making process
  (Redman, 1996)
• Raise the topic of integration for debate and
  further consideration.
• Barr (2002) refers to the property-centric
  database as the No-Go-Zone Particularly
  relevant in terms of GeoDirectory and Irish LAs.

5
Outline of Research

• 30 Stratified random sample of the Register of
  Electors
• 40,000 Records, 173 EDs
• An initial geo-coding run against GeoDirectory
• Use of standard LA tools.
• A Data grooming exercise
• removing inconsistencies and errors (Galdardas
  et al, 2001)
• Second geo-coding run
• Analysis the matched and mismatched.

6
Practicalities of the Research

• The aim was not in itself to achieve the perfect
  match rate but rather to gain a better
  understanding of the dynamic of integration.
• This approach represents one of many possible
  ways of analysing the dynamic and quantifying the
  result
• Not definitive
• Not without limitations

7
STUDY AREA
8
Address-Match Algorithm Mismatch Assessment

• Purpose To find a best match and to identify
  missing Thoroughfare, Secondary Locality
  elements. (T, S L)
• All records given a score based on the number
  of delimited string elements.
• Each and every record from the sample compared to
  every record from GeoDirectory and best match
  identified.
• The related T, S L records were passed back to
  the sample address to confirm their existence.

9
GeoAddress Sample

• Register 7 Hilary Close Abbeyfeale Limerick
  (Target Address)
• dedwrd_id 1
• dedwrd_name Abbeyfeale
• BUILDING_ID 80845731
• ADDRESS_POINT_ID 0
• GEOADDRESS 7 HILLARY CLOSE ABBEYFEALE LIMERICK
  (Source)
• Mscore 4
  (No. Matching address elements)
• RegScore 5
  (Count value for Target Address elements)
• GeoScore 5
  (Count value for Source Address elements)
• MTH HILLARY CLOSE (Missing "Thoroughfare"
  Elements)
• MSTH (Mising "Secondary" Elements)
• MLOC (Mising "Locality" Elements)

10
Actual Findings

• Match Rates Before Grooming (6.6)
• City 37 County 5.2 (Limitations)
• Match Rates after Basic Grooming (17)
• 10 Point Increase.
• Not 40 but Closer to 20 Non-Unique In this
  study
• Further Grooming based on knowledge of local
  semantics 38
• Also using a Bounding File (ED)
• GAMMA Soundex Matching 53, Interactive 74

?
11
What does this mean

• Semantics understand your data
• Look-up tables
• Address matching Algorithm
• Interactive GeoCoding
• Point of Entry -- Address and/or Map

For Any Address Matching Exercise one Table has
to set the standard.
12
Primary Reasons for a Failure to GeoCode

• Non-Unique Addresses
• Incomplete Address (missing elements)
• Various spellings descriptions within LA
  Database
• Inclusion of Unit Numbers Apts etc
• Abbreviations, punctuation, fàda (à).
• Localised spelling versus GeoDirectory
• Additional Noise in the Address file
• Ambiguous Addresses
• Impossible Addresses (1 Dead Letter, Dublin 6E,
  Ireland)

When it comes to cleaning or editing data, there
is little formal guidance (Leahy,2004)
13
RESIDENTS PARKING PERMITS
Thoroughfare
14
Address Model
15
(No Transcript)
16
(No Transcript)
17
(No Transcript)
18
(No Transcript)
19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
ADDRESS DATABASE OPTION
Address Matching
24
(No Transcript)
25
Joined-Up
Library Services/ Membership
HOME
Waste Permists
Fire Safety Certificates
Waiver Schemes
Other Grant Schemes
Commencement Notices
Contributions
Customer Relationship Management
HEG
Residents Parking Permits
Register of Electors
Litter Fines
Rates Commercial and Other
Derelict Sites
Rental Accommodation Schemes
Parking Fines
Road Openings
26
Considerations

• Non-Unique - Cater for In Software.
• Independent Applications (HOME/RAS etc)
• Non-Existent Addresses Feedback to GeoDirectory
• Alias
• Data Protection