Information Technology Reference
In-Depth Information
are retrieved, more than 70% of them are relevant for the user. The filter that
considers both type and capability is quite effective when few items are retrieved,
as confirmed by the highest value of
MAP
achieved for lower levels of
k
.
Tabl e 3.
Mean Average Precision
Filter
MAP@1 MAP@2 MAP@5
No filter
0.139
0.126
0.117
Type
0.149
0.142
0.135
Capability
0.305
0.284
0.273
Type+Capability
0.368
0.354
0.347
Finally, a qualitative analysis of the retrieval accuracy can be carried out by
studying some examples of returned functionalities. We queried the IR service
by a system requirement
r
such as “
The CS shall provide facilities for Human
Computer Interface presentation
”. Table 4 shows the retrieved functionalities
obtained by applying the combined filter (
type
and
capability
) to the input
r
.
It is clear that the returned functionalities have a quite good relevance for the
queries requisite, as the first hits in the Table show. Moreover, the quality of the
relevance decrease along with the ranking: the third returned hit is much less
relevant than the first two, despite properly respond to the requirement.
Tabl e 4.
Example of retrieved functionalities
r
The CS shall provide facilities for Human Computer Interface presentation
f
1
The CMS shall provide similar controls and means of interaction with all
displays, i.e. they should be the same where possible and consistent otherwise.
f
2
The CMS shall provide the following facilities at CMS consoles : screens,
pointing device, keyboards, MFKA, service settings.
f
3
The CMS shall display alerts on primary view area
5 Conclusions
While Semantic Technologies show a large set of promises in the Defense Sys-
tem Engineering domain, they are usually very demanding from the point of
view of complexity in design, optimization and maintenance. Traditional (i.e.
Knowledge-based) HLTs approaches are in this class of technologies. The results
achieved in Statistical Natural Language Processing by the adoption of robust
and accurate Machine Learning algorithms allowed to increase the applicability
of these methods in several domain, from Business Analysis, Web Communica-
tion as well Security. In this paper, a general architecture for large scale and
adaptive Requirement Analysis has been presented. Its application to Require-
ment Analysis in the specific Defense System Engineering domain is evaluated
and discussed. The main idea is to combine requirement classification and IE
for automation of most of the validation stages related to system behaviors. The
system is currently experimented in a specific scenario of Combat System Equip-
ment, applied to the management of the design and description of the Computer
Software Configuration Items interactions. The application of semantic technolo-
gies in the Defense System Engineering domain has shown its potentials in the
Search WWH ::
Custom Search