Database Reference
In-Depth Information
Table 6. TimeML expression for Metadata 1 and Metadata 2
Lys Glacier
<EVENT eid=”e10” class=”OCCURRENCE”>
is melting
</EVENT>
<MAKEINSTANCE eiid=”ei1” eventID=”e10” pos=”VERB” tense=”PRESENT” aspect=”PROGRESSIVE” />
<TIMEX3 tid=”t1” type=”DATE” value=”2007-07-01”>
On July 1
st
, 2007
</TIMEX3>
<TLINK eventInstanceID=”ei1” relatedToTime=”t1” relType=”DURING”/>
<MAKEINSTANCE eiid=”ei2” eventID=”e10” pos=”VERB” tense=”PRESENT” aspect=”PROGRESSIVE” />
<TIMEX3 tid=”t2” type=”DATE” value=”2007-09-03”>
On September 3
rd
, 2007
</TIMEX3>
<TLINK eventInstanceID=”ei2” relatedToTime=”t2” relType=”DURING”/>
<CONFIDENCE tagType=”TIMEX3” tagID=”t2” confidenceValue=”0.7”/>
<TIMEX3 tid=”t3” type=”DATE” value=”2007-09-04”>
On September 4
th
, 2007
</TIMEX3>
<TLINK eventInstanceID=”ei2” relatedToTime=”t3” relType=”DURING”/>
<CONFIDENCE tagType=”TIMEX3” tagID=”t3” confidenceValue=”0.3”/>
a graphic user interface that allows depicting the semantics of the (possibly) soft temporal conditions
on the timeline such as in Figure 3.
The soft temporal conditions are defined as soft constraints, i.e., with a desired membership function
μ
Q
, defined on a time line with a given granularity, chosen among the available ones (Zadeh, 1978).
An example of user selection condition for the
Lys Glacier
melting example reported in table 6 could be:
Search for observations of glacier melting events, occurring close to late Summer 2007.
The soft constraint of this example can correspond to a fuzzy
time interval,
defined as follows: [t=1-8-
2007, Δd={ 0, 15, 43, 17},
day
], where Δd specifies the fuzzy time span in days from the date 1-8-2007.
This definition corresponds to a membership function with a trapezoidal shape such as the obe depicted
in Figure 4 [1-8-2007, 15-8-2007, 23-9-2007, 10-10-2007] (Bosc and Pivert, 2000).
The matching is performed by first filtering the observations that are related to the “Lys Glacier”.
This is done by matching the strings in the metadata “EVENT” and “MAKEINSTANCE” fields.
Then, if an instance is found, in order to obtain two homogeneous temporal representations, we have
to convert either the temporal metadata values in TIME3X or the soft query condition to which it is
matched into a temporal indication, defined on a common basic domain with the same granularity. This
can be done as explained in the following paragraph.
Conversion of a Time Indication
In order to convert a temporal indication defined on a domain into a semantically equivalent one, defined
on a domain with different granularity, we need a time hierarchy such as the one depicted in Figure 5.
This hierarchy defines the relationships between pairs of temporal granules with distinct granularity.
It is a graph in which each node is a time granule, and an edge connecting two nodes
i
and
j
expresses
the composition of the coarser time granule into the finer time granule: a mapping function F
i,j
allows
