Database Reference
In-Depth Information
The analysis is surface compositional in that each surface is analyzed as a
lexical proplet (lexical lookup, cf. Handl et al. 2009). The derivation is time-
linear, as shown by the stair-like addition of one lexical proplet in each new
line. Each line represents a derivation step, based on the application of the
specified LA-hear grammar rule, e.g.,
1N+FV
(defined in 3.4.1). The rules
establish grammatical relations by copying values, as indicated by the diagonal
arrows. The result of the derivation is the order-free set of proplets 3.2.1, ready
to be stored in the agent's content-addressable memory (4.1.1).
Based on the grammatical relations between the proplets stored in the agent's
memory, the second step in the cycle of natural language communication is a
selective activation
of content by navigating from one proplet to the next. The
following example is based on the content 3.2.1, derived in 3.3.1:
3.3.2 DBS
THINK MODE NAVIGATION
3
1
verb: know
cat: decl
noun: Julia
cat: nm
fnc: know
noun: John
cat: nm
arg: Julia John
fnc: know
2
prn: 625
prn: 625
prn: 625
4
The navigation is driven by an LA-think grammar which uses the grammati-
cal relations between proplets like a railroad system. By constructing proplet
addresses from the
arg, fnc
,and
prn
values, the navigation proceeds from the
verb to the subject noun (arrow 1), back to the verb (arrow 2), to the object
noun (arrow 3), and back to the verb (arrow 4).
Such a think mode navigation provides the
what to say
for language produc-
tion from stored content, while the third step in the cycle of communication,
i.e., the speak mode, provides the
how to say it
(McKeown 1985) in the natural
language of choice. Consider the following example of a speak mode deriva-
tion, resulting in a surface realization:
3.3.3 DBS
SPEAK MODE REALIZATION
.
Julia
knows
John
3
1
2
3
1
verb: know
cat: decl
noun: Julia
cat: nm
fnc: know
noun: John
cat: nm
arg: Julia John
fnc: know
2
prn: 625
prn: 625
prn: 625
Search WWH ::
Custom Search