Information Technology Reference
In-Depth Information
1.
John bought a
house
.The
windows
are wooden.
2.
John was bitten by a
snake
on the foot. The
poison
had gone up to the knee
by the time ambulance arrived.
In the example (1) above, the noun
windows
can only be interpreted fully in the context
of the noun
house
mentioned in the previous sentence. In this case, the anaphor-noun
windows
is related to the antecedent-noun in an indirect way, different from the one-
to-one co-reference type relation. The example in (2) shows another indirect relation
between
poison
and
snake
, however it is not the same as the one in (1). NP anaphora
resolution studies (e.g. [24,8,27]) treat these indirect relations as a single category and
refer to them as
associative
or
bridging
anaphora. In this study we propose a relational
framework that distinguishes between the different types of anaphoric relations that can
exist between two nouns, one of which represents the anaphor and the second one repre-
sents the antecedent. Hence in this framework, the task of anaphora resolution involves
identifying the antecedent as well as the type of relation to the antecedent. To distin-
guish from the previous works, we will use the term
anaphora interpretation
, instead
of
anaphora resolution
, where the latter involves only identification of the antecedent.
Since we are also identifying the type of relation, it is possible for an anaphor to have
multiple antecedents, related by the same or a different relation. This is a significant
departure from the conventional notion of anaphora resolution where an anaphor is
resolved to a single, previously mentioned entity. In the case in which the antecedent is
also an anaphor, it is assumed to be already resolved, forming a sequential chain. For
some NP anaphora this is inadequate. As an illustration, consider the excerpt below:
The robber jumped out of the
window
1
.
The
house
2
belonged to Mr Smith.
The
window
3
is thought to have been unlocked.
If we allow a single resolution relation for an anaphoric NP, then
window
3
would have
to be resolved to either
house
2
or
window
1
. In either case, a part of the information
would not be captured. A common strategy in most studies (eg. [24,8]) is to resolve to
the most recent antecedent. In the case of the above excerpt, this would mean that we
resolve
window
3
to
house
2
which can be assumed to be already resolved to
window
1
.
There are two inadequacies in this strategy; firstly the semantic difference between the
relation of
window
3
to window
1
and
window
3
to
house
2
is approximated by a single
co-reference relation, and secondly as a consequence, the direct relation between
win-
dow
3
to
window
1
is not captured. In the proposed framework, we will identify both
house
2
and
window
1
as antecedents and interpret each of them with a different relation.
It can be argued that this can be overwhelming since we can form a relation even be-
tween a pair of very remote entities. However the constraint in our case is that we are
only interested in
relations that give rise to anaphoric use of NPs
. The interpretation
framework involves
specifying a relation
between an anaphor and the antecedent hence
a consequence of this is that an NP can form relations with
more than one antecedent
.
This allows us to represent and interpret anaphoric uses of a noun such as
window
to an
occurrence of
house
and
another occurrence of
window
.
