Simultaneous Amalgamation - Computational Linguistics and Talking Robots - page 154

FoCL'99, Sect. 12.5, this is simply a bad approach; there is the empirically
adequate alternative of semantic doubling , which is of only linear complexity.
In short, LA-content may (i) contain recursive structures and (ii) have ambi-
guities, but still be a C1-LAG (if there are no recursive ambiguities). Whether
or not LA-content as defined in 8.2.1 contains recursive ambiguities may be
determined in two ways: (i) investigate the grammar by following the parallel
possible continuations through the rule packages 9 or (ii) investigate the content
phenomena which the grammar is designed to model.
If a recursive ambiguity is detected in the rule system, it may be empirically
legitimate or not. If not, the recursive ambiguity must be eliminated from the
grammar, for example, by modifying a rule package (debugging).
The first case, in contrast, would constitute an interesting linguistic discov-
ery. However, as long as we can't find a recursive ambiguity as a phenomenon
of natural language, and thus as a phenomenon of the corresponding content,
LA-content must not exceed the linear complexity of the C1-languages - in
concord with the CoNSyx hypothesis of FoCL'99, 12.5.7.
8.4 Infinitive Content Constructions
Let us turn now to a content construction which has not yet been treated in
DBS, namely the infinitive. Its analysis will provide us with an opportunity to
compare the assembling of content (i) by means of nonlanguage recognition
(LA-content) and (ii) by means of natural language interpretation (LA-hear).
Good entry points for analyzing infinitive content constructions are corre-
sponding surfaces of English. 10 At their center is a form of the verb, for exam-
ple, see . Following the grammars for classical Latin, traditional grammars of
English call it the infinitive, classify it as non-finite and unmarked for person,
number, and tense, 11 and use it as the verb's base form (“citation form”). It
should not be overlooked, however, that the surfaces of such infinitives as (to)
see, (to) eat, and (to) give are the same as the unmarked finite non-third-
person singular present tense forms in (I) see, (you) eat, and (they) give ,
respectively.
In many grammatical functions infinitives may be viewed as a stripped-down
version of corresponding subclauses, e.g., a subject sentence or an object sen-
9 Handl (2011) presents an algorithm which determines for any C-LAG whether it is recursively am-
biguous or not. The algorithm runs in r 2 ,where r is the number of rules in the grammar.
10 For an analysis of infinitive constructions in different languages, see Wurmbrandt (2001) and the
review by Reis and Sternefeld (2004).
11 We are referring here to an elementary verb form such as (to) see , leaving phrasal constructions such
as (to) have seen or (to) have been seen aside.