RDF Access to Relational Databases

1
(No Transcript)
2
RDF Access to Relational Databases Ashok
Malhotra Oracle Corporation
3
Extracting Information from Relational Databases

• There is very valuable data stored in Relational
  Databases
• We should be able to do a much better job of
  extracting information from this data
• There are two basic problems
• Related data is often fragmented among many
  isolated databases
• Relational databases have very little semantic
  information
• Semantic information oftem stored on bits of
  paper or in the minds of the developer
• Sometimes encoded in table and column names

4
Thus, we propose

• Step1 Collect all the relevant Relational
  Databases.
• Step 2 Create an ontology or semantic cover
  encompassing the information in the databases
• This is a difficult process
• More art than science

5
Relationships between databases

• These can take many different forms
• Analogous data in one database might be
  represented in a different number of tables in
  another database. 
• Analogous data in one database might be
  represented in a different number of columns in
  another database.
• Analogous data in one database might be
  represented in a different number of rows in
  another database

6
Relationships between databases

• Analogous data in one database might be
  represented using different values in another
  database (e.g., one database uses "m" and "f" for
  "male" and "female", while the other might use
  "0" and "1", respectively), or the same values in
  the various databases might mean different things
  (e.g., one database uses "1" and "0" for "male"
  and "female", while the other might use "0" and
  "1", respectively), or the values in the various
  databases might be incompatible in ways that make
  the mapping very difficult (e.g., one database
  uses "0" and "1" for "male" and "female",
  respectively, which the other database uses
  values from "0" through "9" to represent
  medically-significant variations of human sexes,
  such as types of hermaphrodites).

7
Thus, we propose

• Step 3
• Write a SQL query to extract data relevant to the
  class from each participating database
• Write an integrating query to integrate the data
  obtained from the individual databases to create
  the data for the class
• It is now possible to write queries on the
  semantic cover using SPARQL

8
Thus, we propose

• Step 4
• Translate SPARQL queries into underlying SQL
  queries using the queries created for each
  class