Malleable Schemas: A Preliminary Report

1
Malleable SchemasA Preliminary Report

• Xin (Luna) Dong and Alon Halevy
• _at_ WebDB 2005
• June 17, 2005

2
Domain Modeling in Data Integration is Hard
Unstructured data sources
?
Clean Schema
Structured data sources
3
Real-Estate Website I

• Single Family Home
• Condo/Townhouse/Co-op
• Multi-Family Home
• Mfd/Mobile Home
• Land
• Farms
• Rentals

4
Real-Estate Website II

• Condo/Townhouse
• Foreclosures
• Lots/Land
• Mobile Home
• Multi-Family Dwelling
• Other
• Single Family Homes
• Timeshares

5
Real-Estate Website III

• Acreage/Semi-Rural
• Alpine
• Apartment
• Apt/Unit/Flat
• Block of Units
• Duplex/Semi-detached
• Flat
• House
• Retirement Living
• Serviced Apartment
• Studio
• Terrace
• Townhouse
• Unit
• Villa
• Warehouse

6
Real-Estate Websites

• Single Family Home
• Condo/Townhouse/Co-op
• Multi-Family Home
• Mfd/Mobile Home
• Land
• Farms
• Rentals

• Condo/Townhouse
• Foreclosures
• Lots/Land
• Mobile Home
• Multi-Family Dwelling
• Other
• Single Family Homes
• Timeshares

• Acreage/Semi-Rural
• Alpine
• Apartment
• Apt/Unit/Flat
• Block of Units
• Duplex/Semi-detached
• Flat
• House
• Retirement Living
• Serviced Apartment
• Studio
• Terrace
• Townhouse
• Unit
• Villa
• Warehouse

7
Real-Estate Websites

• Single Family Home
• Condo/Townhouse/Co-op
• Multi-Family Home
• Mfd/Mobile Home
• Land
• Farms
• Rentals

• Condo/Townhouse
• Foreclosures
• Lots/Land
• Mobile Home
• Multi-Family Dwelling
• Other
• Single Family Homes
• Timeshares

• Acreage/Semi-Rural
• Alpine
• Apartment
• Apt/Unit/Flat
• Block of Units
• Duplex/Semi-detached
• Flat
• House
• Retirement Living
• Serviced Apartment
• Studio
• Terrace
• Townhouse
• Unit
• Villa
• Warehouse

8
Real-Estate Websites

• Single Family Home
• Condo/Townhouse/Co-op
• Multi-Family Home
• Mfd/Mobile Home
• Land
• Farms
• Rentals

• Condo/Townhouse
• Foreclosures
• Lots/Land
• Mobile Home
• Multi-Family Dwelling
• Other
• Single Family Homes
• Timeshares

• Acreage/Semi-Rural
• Alpine
• Apartment
• Apt/Unit/Flat
• Block of Units
• Duplex/Semi-detached
• Flat
• House
• Retirement Living
• Serviced Apartment
• Studio
• Terrace
• Townhouse
• Unit
• Villa
• Warehouse

9
Real-Estate Websites

• Single Family Home
• Condo/Townhouse/Co-op
• Multi-Family Home
• Mfd/Mobile Home
• Land
• Farms
• Rentals

• Condo/Townhouse
• Foreclosures
• Lots/Land
• Mobile Home
• Multi-Family Dwelling
• Other
• Single Family Homes
• Timeshares

• Acreage/Semi-Rural
• Alpine
• Apartment
• Apt/Unit/Flat
• Block of Units
• Duplex/Semi-detached
• Flat
• House
• Retirement Living
• Serviced Apartment
• Studio
• Terrace
• Townhouse
• Unit
• Villa
• Warehouse

10
Malleable Schemas

• Key idea capture the important aspects of the
  domain model without committing to a strict schema

Unstructured data sources
Clean Schema
Structured data sources
11
Malleable Schemas are Needed in Many Cases

• The structure is inherently evolving
• The domain is extremely complicated
• Domain structure is not completely known
• The borders between structured and unstructured
  parts of data are fuzzy
• Modeling every detail is too overwhelming and
  impractical

12
Main Ideas in Malleable Schemas (I)

• Malleable elements in Schema Definition
• Use keywords or phrases as element names
• Allow regular expressions as element names
• Examples
• ShortTermRental, VacationRental RealEstate
• rentalClerk, agent contactPerson

13
Main Ideas in Malleable Schemas (II)

• Imprecise Reference in Query
• SELECT city, price, leaseAgent
• FROM ShortTermRental
• WHERE location Tahoe

14
Malleable SchemasA Preliminary Report

• Xin (Luna) Dong and Alon Halevy
• _at_ WebDB 2005
• http//www.cs.washington.edu/homes/lunadong