2.3 KiB
Semantic Databases
The meaning that we have in common can be shared more effectively and we can communicate structures and information between databases
Semantic databases - What does a table actually tell us
Sharing a Relational database
- Just data
- CSV
- A table in a RDB, the data in the table but also the column specification
- The CSV file is the data but not the relation
- Full relation (data and data types)
- Database dump (.sql)
Example
SF Movie Locations
- Actor1, Actor2, Actor3
- Personal name
- An actor
- Year
- Year is a positive integer
- Year is a calendar year
- Location
- Text
- Geographical location
- In San Francisco
Layers of meaning
- Data type
- string, integer, float
- Data domain
- place, person, date
- Data semantics
- Person acted in film
The more meaning we can specify, the more we're able to spot when someone else is specifying the same meaning.
Shared meaning
- Share data
- e.g. a .csv file
- Share difinitions
- Data entry documentation
- share syntax
- Validate new data. Agreements how we'll encode our information so new entries are accepted
- Share machine-readable semantics
Deductive database
First order logic:
ancestor(c, a) <- parent(c, a)
ancestor(c, a) <- parent(c, b) AND ancestor(b, a)
sibling(a, b) <- parent(c, a) AND parent(c, b)
In this example:
<-ancestor('Anakin', 'Kylo')
It won't look for explicit '___ is ancestor of ___' but it will look for the relation.
Sharing meaning in the real world
Common semantics
- Shared documents(s)
- Formal specifications
- Human-readable definitions
We always need realiable ways of sharing:
- Information
- Structures and semantics in computer-readable form
- Structures and semantics in a human-readable form
With these three, we can start to communicate and share meaning.
For example, STEM a music notation, tries to convey the last two (computer and human-readable)
A specification document in human-readable form:
XML - Documents with semantics
eXtensible Markup Language
Example of text that has been marked up:
Atributes
There always needs to be a parent node betcause documents are trees:
