Database Reference
In-Depth Information
4.
Mystery Number Three:
Memory Structure
Because the relations between proplets within a proposition and between
propositions are coded solely by means of proplet-internal attribute values
(addresses), proplets are inherently
order-free
. Thus proplets can be stored
in accordance with any kind of database schema - without losing their inter-
proplet relations, but gaining the use of the storage and retrieval mechanism of
the database selected. Given that there are several different database schemata
available, the question of the third mystery is: which one is suited best for
modeling cognition, including the cycle of natural language communication?
4.1 Database Schema of a Word Bank
When faced with the choice of a computer memory, the most basic alternative
is between a
coordinate-
addressable and a
content-
addressable approach (cf.
Chisvin and Duckworth 1992 for an overview). Though peppered with patents,
the content-addressable approach is less widely used than the coordinate-
addressable approach. A content-addressable memory is suited best for the
super-fast retrieval of content which is written once and never changed.
A coordinate-addressable memory, e.g., an RDBMS, resembles a modern
public library in which a topic can be stored wherever there is space (random
access) and retrieved using a separate index (inverted file) relating a primary
key (e.g., author, title, year) to its location of storage (e.g., 1365). A content-
addressable memory, in contrast, is like a private library in which topics with
certain properties are grouped together on certain shelves, ready to be browsed
without the help of a separate index. For example, at Oxford University the
2 500 volumes of Sir Thomas Bodley's library from the year 1598 are still
organized according to the century and the country of their origin.
In an initial response to a content-addressable approach, mainstream data-
base scientists pointed out that it can be simulated by the coordinate-address-
able approach (Fischer 2002), using well-established relational databases. The
issue here, however, is whether or not the formal intuitions of the content-
addressable approach can be refined naturally into a model of cognition.
Search WWH ::
Custom Search