Database Reference
In-Depth Information
The lexical proplet representing read has the category (n-s3' a' v) . It matches
the RegEx pattern (n-s3' .* v) , which represents the non-third-person singular
present tense (i.e., unmarked) form of the English verb. The Kleene star * gen-
eralizes over varying oblique valency positions. 12 Because the surface of this
form may be used systematically 13 as the infinitive, it would violate the prin-
ciple of Surface Compositionality (SCG'84, FoCL'99, NLC'06) 14 to postulate
separate lexical analyses for (i) the unmarked finite verb form of English and
(ii) the English infinitive. The same holds for German, though it uses another
finite verb form, namely the first and third person plural present tense form,
e.g., lesen , for the infinitive. The infinitive in Latin, in contrast, has a separate
surface, e.g., amare , which does not equal any finite form.
Consider the content resulting from the hear mode derivation 8.4.1:
8.4.2 C ONTENT REPRESENTATION OF AN INFINITIVE CONSTRUCTION
noun: John
cat: nm
sem: sg
fnc: try
prn: 32
verb: try
cat: decl
arg: John read
mdr:
prn: 32
verb: read
cat: inf
fnc: try
arg: John book
prn: 32
noun: book
cat: snp
sem: indef sg
fnc: read
prn: 32
The proplet John has try as its fnc value, while try has John as its first arg
value; using schema (1) in 7.3.3, the bidirectional relation between John and
try is characterized as subject
verb. The proplet read has book as its second
arg value, while book has read as its fnc value; using schema (2) in 7.3.3, the
bidirectional relation between read and book is characterized as object
/
verb.
The bidirectional relation between try and read remains. One direction is
coded by the value read in the second position of the arg attribute of try .This
is opaque insofar as read is verbal rather than nominal because it also serves
as the core value of a verb proplet.
The other direction is coded by the fnc attribute of read with the value try .
This is opaque insofar as a verb proplet like read normally (i.e., in a transparent
\
12 The equivalent DBS notation is (n-s3' X v) . While early NEWCAT'86 implementations of LA-
grammar used RegEx for matching ordered category values, the pattern matching for proplets requires
a more differentiated method based on restricted variables.
13 The single exception is the auxiliary be with its non-third-person singular present tense forms am
and are . This exception is handled by introducing the variable INF (for infinitive) with the following
restriction: INF {(n-s3' .* v), (be-inf)} . Thereby be-inf is the cat value of be and (n-s3' .* v) covers
all other infinitives of English.
More specifically, the RegEx pattern (n-s3' .* v) matches the relevant forms of one-, two-, and
three-place main verbs as well as those of have and do , i.e., (n-s3' v), (n-s3' a' v), (n-s3' d' a' v),
(n-s3' hv' v), and (n-s3' do' v ), respectively. In contrast, the forms am of category (ns1' be' v) and
are of category (n-s13' be' v) as defined in NLC'06, 13.1.8, are not matched. Instead, the cat value
of be is defined as be-inf and added to the restriction of the variable INF.
14 Huddleston and Pullum (2002) postulate two lexical entries for the form in question, called “primary
plain” and “secondary plain,” and use the latter for the imperative, the subjunctive, and the infinitive.
Search WWH ::




Custom Search