What is it?
In many intelligence applications (in security, business, finance, etc.), we want to monitor a stream of
data about some real phenomenon space of interest in order to detect occurrences or developments of
particular objects, situation or patterns of interest. LT’s Recognition Technology provides a solution
for decision makers to monitor big data streams to be alerted about occurrences of situations, etc., of
interest and to be warned early about the likely future emerge of such situations.
How does it work?
The data streams may be from multiple sources that may be of different kinds, depending on the
application domain. The information is extracted, harmonized, interpreted, etc., by the iExtractor
Examples of kind of sources are data sensors, video sensors, transaction (business, financial) data,
news feeds, security information and intelligence (multi INF/INT) and investigative findings.
Utilizing such data for recognition of situations, etc., of interest have several challenges that are
handled by LT’s Recognition Technology, namely:
The expected reliability of the source
All other thing equal, the more reliable the source, the more
trustworthy is the information it provides. On the other hand, a less reliable source may provide
information that is very interesting, if it is truthful.
The information provided by the data in a stream may be more or less compatible
with other existing information or knowledge.
Imperfections of the data
The data may be incorrect or erroneous, due to various causes, or only
provide imprecise information.
For effective recognition, the importance of attributes (aspects, variables)
extracted from the data stream must be considered. The recognizer will consider all relevant
aspects we can derive from the data streams and their influence in the recognition. The latter
implies considering criteria (or subset of criteria) that are sufficient, necessary or just contributive
to the recognition.
Integrability and aggregation
This relates primarily to the independence of the criteria. Thus, a
high satisfaction of two criteria will, all other things equal, contribute less to the recognition if the
criteria are interdependent than if they are independent.