Bio/Ecosystem Informatics

Transcription

1 Bio/Ecosystem Informatics Renée J. Miller University of Toronto DB research problem: managing data semantics R. J. Miller University of Toronto 1

2 Managing Data Semantics Semantics modeled by Schemas (structure and constraints) Creating semantics Data design Managing semantics Understanding and reconciling different choices made in modeling data semantics R. J. Miller University of Toronto 2

3 Managing Data Semantics Problem Active area of DB research Not specific to bio/ecosystem informatics Solutions Tailored to data characteristics Our solution is focused on curated data R. J. Miller University of Toronto 3

4 Clio: Schema & View Management University of Toronto Renée J. Miller Periklis Andritsos, Ariel Fuxman, Tasos Kemesetsidis, Yannis Velegrakis IBM Almaden Laura Haas Ron Fagin, Mauricio Hernández Howard Ho, Phokion Kolaitis Lucian Popa, Ling-Ling Yan R. J. Miller University of Toronto 4

5 Clio: Schema & View Management Manage datasets that may differ in: Data models Structures (table or nesting structure) Semantics (constraints) Content Example problems Schema Mapping Schema Translation (Wrapper Generation) Schema Evolution Schema Integration R. J. Miller University of Toronto 5

6 Schema Mapping Source schema S schema mapping Target schema T conforms to conforms to data translation program (queries) Clio facilitates: Creation, management, debugging of schema mappings Uses Data exchange: create a (materialized) target instance Data integration: translate queries on (virtual) target into queries on source(s) R. J. Miller University of Toronto 6

7 Data Publishing & Exchange WWW XML schema Business partner WebService WSDL Life Sciences Company Relational Schema Microarray data in DB2 SwissProt PubMed GenBank R. J. Miller University of Toronto 7

8 Illustration: Clio Schema Mapping R. J. Miller University of Toronto 8

9 R. J. Miller University of Toronto 9

10 Illustration: Clio Schema Mapping Support Nested Structures Element correspondences Human friendly Automatic discovery Preserve data meaning Discover data associations Use constraints & schema Create New Target Values Produce Correct Grouping And produce XQuery R. J. Miller University of Toronto 10

11 Why Should DB Researchers Care about Bio/Ecosystems? Great Data Rich, complex semantics Exercises many facets of our models Publicly available!!! Domain curators are expert DBAs Provide important feedback R. J. Miller University of Toronto 11

12 How Should DB Researchers Work with Bio/Ecosystem Scientists? Our Model Took existing, general DM problem Explored systems and foundational issues Prototyped tool Team of developers work closely with domain scientists Provide feedback Arms-length collaboration!!! R. J. Miller University of Toronto 12

13 Overview Goal: interoperability between independent data sources Schema Mapping Data Translation (Data Exchange Problem) Challenges Schemas can be arbitrarily different Still, data must not lose its meaning during translation Maximum advantage of semantics embedded in schemas & data Used in compilation Facilitate user specification of any additional semantics As by-product user learns if semantics incorrect/incomplete Performed manually: complex user queries, programs, etc. Output: correct data translation program R. J. Miller University of Toronto 13