Week 13 notes completed
This commit is contained in:
Binary file not shown.
|
After Width: | Height: | Size: 177 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 208 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 88 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 97 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 93 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 25 KiB |
@ -0,0 +1,93 @@
|
||||
# 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:
|
||||

|
||||

|
||||
Reference in New Issue
Block a user