Database Reference
In-Depth Information
its core and (ii) its prn value, e.g., know 625 . Most cognitive operations, how-
ever, require internal access based on addresses (pointers; cf. Sect. 4.4).
Compared to the classic 1969 CODASYL network database, a Word Bank is
highly constrained. First, the member proplets belonging to an owner proplet
are listed in the temporal order of their arrival. Second, the members in a token
line must share the owner's core value (no multiple owners). Third, the only
connections between proplets across token lines are the semantic relations of
functor-argument and coordination. Fourth, like the relations between owners
and members, the semantic connections are 1:n relations: one functor - sev-
eral possible arguments; one first conjunct - several possible successors; one
original - several possible coreferent address proplets.
A Word Bank is a kind of navigational database because it supports the nav-
igation from one proplet to the next, using the semantic relations between
proplets (3.3.2) and along token lines (4.2.2) like a railroad system, with the
algorithm of LA-grammar (3.4.2) as the locomotive. However, while the nav-
igational databases of the past (Bachman 1973) and the present (XPath, Kay
2004) are intended to be driven by external human users, the system presented
here is located inside an artificial cognitive agent, serving as the container and
structural foundation of autonomous control (Chap. 5).
4.2 Retrieving Answers to Questions
Because the proplets derived in the hear mode (3.3.1) have a core and a prn
value, they are suitable for (i) storage in a Word Bank (4.1.1). For the same
reason, stored proplets support the operation of (ii) navigating from a given
proplet to a successor proplet across token lines (3.3.2) in one of the two basic
kinds of think mode. Moreover, because there is a speak mode which is riding
piggyback on the think mode (3.3.3), the proplets in a Word Bank are suitable
(iii) for language production from stored content as well.
Another operation enabled by proplets in a Word Bank is (iv) retrieving an-
swers to questions. This operation is based on moving a query pattern along
a token line until matching between the pattern and a proplet is successful. A
query pattern is defined as a proplet with at least one variable as a value.
Consider an agent thinking about girls. This means activating the corre-
sponding token line, as in the following example:
2 The token line for any core value is found by using a trie structure (Fredkin 1960). The search for
a proplet within a token line may use the prn value of the address in relation to the strictly linear
increasing prn values. Technically, this may be based on binary search, in time O(log(n)) (Cormen
et al. 2009), or interpolation, in time O(log(log(n))) (Weiss 2005), where n is the length of the token
line.
Search WWH ::




Custom Search