Integration of Spatial Information Sources Based on Source Description Framework

1
Integration of Spatial Information Sources Based
on Source Description Framework

• Yoshiharu Ishikawa, Gihyong Ryu, and Hiroyuki
  Kitagawa
• University of Tsukuba

2
Background

• Spatial information sources emerging new
  information sources on the Internet
• information sources that provide region- or
  location-oriented information
• they support mobile users with GPSs and hand-held
  devices
• Need for the technology to integrate spatial
  information sources
• description of spatial information sources by
  taking their contents into consideration
• efficient and effective query planning and
  processing

• Popular approach for information integration
• well-known wrapper-mediator approach
• Wrapper
• encapsulates the detail of each information
  source
• provides abstract uniform view of the source
• Mediator
• selects appropriate information sources for a
  given query
• query planning and processing

3
Our Objective and Approaches

• Objective development of a spatial information
  integration framework for location-aware
  information services
• to provide useful location-oriented information
  service to mobile users
• Our approach (1) development of a description
  method to represent spatial information sources
• based on the source description framework
• describes the contents and the service of the
  source

• Our approach (2) development of query planning
  and processing methods that effectively utilize
  source descriptions
• considers the heterogeneity of the underlying
  information sources
• effective use of the query processing power of
  each information source

4
Motivating Example (1)

• Information Source A provides restaurant info
  for a specific area
• contains information of restaurants within the
  rectangle area rectA
• given name or address, it returns the matched
  restaurants

• Query received by the mediator show top-20
  nearest restaurants such that
• within 1000 meters from the current position
• the evaluation score is more than or equal to 2.5
  stars

rectA
6
7
1000m
2
5
1
4
3
5
Motivating Example (2)

• Information Source B supports spatial conditions
  to query restaurant info
• returns restaurants within the specified circle
  area (results are ordered by their distances)
• receives additional condition on restaurant
  category

• Information Source C provides restaurant
  evaluation scores
• given restaurant name, it returns the evaluation
  score

select from Source-C where name Manchu
category Chinese
name
score
3.0
Manchu
4
1
2
3
2
1
3
5
6
Source Description Framework (1)

• Source Description Framework a formal framework
  to specify meta information for an information
  source
• proposed in Information Manifold Levy et al 96
• A source description consists of
• Contents Description describes the contents of
  the source in terms of the global schema
• Capability Description describes the types of
  queries which the source can support

• Our approach (1) incorporates the notion of
  spatial data types into the source description
  framework then represents
• spatial information
• spatial queries
• Our approach (2)
• allows the specification of top-N query
  capability in capability descriptions

7
Source Description Framework (2)

• Data model
• a global schema is written in the relational data
  model enhanced with spatial data types
• a global schema specifies a virtual database
  each information source is (partially) mapped
  into the schema

• Query language monoid comprehension
  FegarasMaier95
• a declarative query language
• an extension of list comprehension (used in
  functional programming) to multiple collection
  types (e.g., bag, set)
• basic form ME Q1, Q2, ..., Qn
• M the collection type of the evaluation result
  of the form
• E allowable expression
• Qi generator (with the form v ? V ) or filter
  (predicate)

relation Restaurant relation Evaluation
name string name string category
string score real address string
location point
8
Examples of Source Description

• Source description for information source A
• it provides information of restaurants within the
  region rectA
• it can receive name or address as query conditions

• Source description for information source B
• it receives the query point and the allowable
  maximal distance
• it returns ordered results
• it can receive category as an additional
  filtering condition

contents description
capability description
Source B contents SB ? setr r ?
Restaurant input/output ltq point, m realgt ?
sorteddx x ? SB, d ? dist(x.location, q), d
? m filters ltc stringgt ? category c
sorted result based on distance values
9
Query Processing (1)

• Step 1 an access target description query is
  generated
• AQ specifies the required information to process
  query Q
• Step 2 subqueries are extracted QR and QS

• Example query retrieve the name and address of
  restaurants such that
• within 1000 meters from mypos (the current
  position of the user)
• their evaluation scores are larger than or equal
  to 2.5 stars
• within the nearest top-20

AQ setrs r ? Restaurant, s ? Scoring,
r.name s.name, s.score ? 2.5,
dist(r.location, mypos) ? 1000
Q head20( sorteddltname r.name, address
r.addressgt r ? Restaurant, s ? Evaluation,
r.name s.name, s.score ? 2.5, d ?
dist(r.location, mypos), d ? 1000)
QR setr r ? Restaurant,
dist(r.location, mypos) ? 1000 QS sets s
?Evaluation, s.score ? 2.5 AQ setrs r ? QR
, s ? QS , r.name s.name
10
Query Processing (2)

• Step 3 target information sources are determined
  for each subquery
• For example, source A may contain required
  information for QR and becomes the target
  information source if SA ? QR ? ? (SA has
  appeared in the source description for source A).
  
• This condition is equivalent to

• Step 4 a query plan is generated for each
  combination of information sources for example,
• Step 5 the final integration plan is generated
  based on the subqueries over the information
  sources

PA, C setxy x ? IterC(score ? 2.5),
y ? IterA(name x.name),
dist(y.location, mypos) ? 1000
the condition in(r.location, rectA) ?
(dist(r.location, mypos) ? 1000) is satisfiable