Scenarios for the Pass\'e Simple and Imparfait: An Event Calculus Approach to French Semantics
Fabrice Nauze

Abstract:
Scenarios for the Pass\'e Simple and Imparfait
An Event Calculus Approach to French Semantics
Fabrice Nauze

Abstract:
The goal of this thesis is to apply an AI technique for planning to natural
language semantics (pragmatics). It could seem strange at first sight to
relate both areas but let us formulate more precisely the topic and then try
to explain the relevance of this approach.
  The idea behind this thesis is to apply the AI formalism called Event
Calculus to the semantics of two past tenses in French, namely the Pass\'e
Simple and the Imparfait, to obtain a cognitively plausible explanation of
some phenomena related to tense and aspect.
  The event calculus is a formalism to model reasoning with time and
change and as such it is usually used to solve planning problems. But what
has planning to do with tenses?

The main hypothesis underlying this thesis is that our experience of time
is not primitive (i.e. time is not an input to our perceptions) but could
lie in the necessity to solve planning problems. This hypothesis granted, it
doesn't appear strange anymore to think that the comprehension of tenses
in natural language could rely somehow on our ability to plan.
  It seems then natural to first expose several methods for constructing
time from events and this will be our task in chapter 1. The influence of
this chapter on the rest of the thesis will be apparent to the reader when
we will comment some methods used to treat the semantics of the French
past tenses. In chapter 2 we sum up the needed data to explore the se-
mantics of the Pass\'e Simple and the Imparfait. We introduce the common
linguistics views on those tenses and try to insert aspect in this picture.
We will then turn to some theories used to account for temporal and as-
pectual problems in chapter 3. We will first discuss Kamp's framework,
i.e. Discourse Representation Theory (from now on DRT), which doesn't
account for aspectual problems. Then we will turn to de Swart's method
which incorporates in Kamp's framework a treatment of aspectual informa-
tion and extends DRT to Segmented DRT which "offers a theory of the
semantics/pragmatics interface".1 Finally we will expose Gosselin's frame-
work which is much more influenced by cognitive considerations. The last
two chapters will be devoted respectively to an introduction to the event
calculus and to a treatment within the event calculus framework of some of
our previous data. At this point we will try to make obvious why the use
of the event calculus is relevant in the study of the Pass\'e Simple and the
Imparfait semantics. We can already say that the way aspectual classes are
incorporated in the formalism and the notion of a scenario will be two of
the features that make the event calculus formalism particularly well-suited
for dealing with both aspectual and tense information in sentences with a
Pass\'e Simple and/or an Imparfait.