CAPÍTULO 2: MARCO DE REFERENCIA
2.1 Categoría 1: Paradigma Pedagógico Ignaciano
2.1.2 La educación personalizada como principio de formación integral
I will develope, in a very informal fashion, a possible way to translate a given sentence in natural language to a semantic representation in EC, taking into account ourCGapproach.
Consider example (7.3), here reproduced as:
(7.5) JohnJohn[x1] threw Mary the ball
f throw[x4]e M ary[x2]f the ball[x3]f
We have supposed here that we already identified the semantic representation of the lexicon. Now, we face the following problem:
1. How to match the free variables of the objects with the appropriate vari- ables of the construction? (And how to choose the correct construction in the first place?)
Since we are following Goldberg’s proposal, the ditransitive construction is determined by the sentence type of the formN P V N P N P. Therefore,
the free variables of the objects and the variables of the construction are linked in a straightforward way.1
2. How to link the verb fluent with the scenario of the construction? We will focus on one aktionsart, namely, accomplishments. The scenario for this aktionsart provides an activity. It is this activity that we will unify with thetrans[x]act activity of the scenario of the construction.
The outcome of the process would look like a set of unification constraints. So, for our example (7.3), the set of unifications would be this:
{x=x1;y=x2;z=x3;
trans[x]act=T hrow[x4]act;x=x4}
This set of constraints can be added to the scenario determined by the construc- tion and the lexical elements. If we could work out an integrity constraint from the tense of the verb, we would have an interpretation of the sentence inEC.
Let us consider an example where the construction supplies argument roles that do not have a corresponding participant role, and let us try to work it in more detail. Consider the example:
Daniel kicked Tom the ball
Daniel[x1]f kick(x4, t) T om[x2]f the ball[x3]f
x1=Daniel Tense=past x2=Tom x3=the ball
We get a set of unification constraints between the objects and the participant roles straightforwardly from the order of the objects:
{x=x1;y=x2;z=x3}
We need now to identify the aktionsart of the sentence. Since the sentence is
N P V N P N P, the relevant VP is given by the verb and the second NP in the sentence. We can then look up in a database what its aktionsart is. I will assume we identified in our example the aktionsart, which is an accomplishment. We continue the representation by introducingECconstants for the accomplishment and its respective scenario. The constants consist of the predicatekick(x4, t) and the activity constructed from it: kick[x4]act. The unification of this activity
and the activity of the construction can be introduced just as another constraint:
trans[x]act=kick[x4]act
1This question takes on a more complicated form if we were following Michaelis. For the
linking between the objects of the sentence —that is, ‘John’, ‘Mary’, and ‘the ball’— and the argument roles is not straightforwardly given by their order in the sentence. For one thing, in the sentence:
(7.6) Mary was thrown the ball by John,
the link between the objects and the argument roles is not given by their order in the sentence. Furthermore, there is no principle way in which we can distinguish ‘Mary’ from ‘John’. If our criteria is purely semantic, we can make any of them the agent argument role. We will need aid from the syntax of the sentence to link objects and argument roles.
We can now build up a scenario out of the constraints on the constants for the objects, the aktionsart of the VP, the scenario for the construction, and the linking constraints. Adding the appropriate integrity constraint for the past tense of an achievement we get the full semantic representation inEC.
Finally, consider a sentence like:
Patrick sleeps George the piano
P atrick[x1]f sleep[x4]f George[x2]f the piano[x3]f
This sentence cannot be represented. For one thing, the verb is a state, and thus is represented by means of the fluentsleep[x4]f. This cannot be unified
with the activity trans[x]act. So the whole unification fails. This failure on
unification should be enough to discard a representation of this sentence. By the same reason, no state can be represented this way. The proposal is to use this mechanism of unification to handle with any other constraints we require either on the argument roles of the construction or on the meaning of the verb.
Further constraints are needed, indeed. Consider the following example: (7.7) *A gift was given Pat by Jennifer.
This sentence can be unified in the set up I have laid out so far. Since this sentence is clearly wrong, we need to specify constraints that rule it out. A constraint requiring the ‘recipient’ to be animate would certainly help us to rule (7.7) out. However, this would be too strong, since we would rule out also sentences which are grammatical, like:
(7.8) The music lent the party a festive air.
The way out taken by Goldberg is to keep the constraint on the ‘recipient’, and treat (7.8) as a metaphor. The adequacy of this solution needs to be assessed with care, since it is not clear how to treat metaphors within a formal framework.