Scientific Discovery Within Data Streams - Ambient Intelligence for Scientific Discovery - page 69

Fig. 1. Data Stream Processing Engine Concept.
If, on the other hand, a classification mode of operation is required, a different
set of algorithms could be substituted to perform clustering of documents into
specific folders dependant on content. This 'plug and play' nature of our streams
processing concept can be seen in Fig. 2.
Viewing this process from the user's standpoint, prior to any stream process-
ing the user of the system creates a 'stream tap': a description of what particular
stream they are interested in, and what operations they wish to perform. This
tap also defines report settings such as how they wish to be notified (e.g. email,
SMS text messaging, etc). Figure 3 shows a schematic of the user interface el-
ements that produce the steam tap. First of all, the set of streams available to
the user are presented. This usually includes all the streams that have been pro-
cessed for inclusion in our prototype (i.e. had XML stream format descriptors
created) although security mechanisms could be incorporated so to present only
streams that a particular user is cleared to use. After selecting a stream, the
prototype verifies that the stream is available and running (i.e. currently being
received). It would be inecient to have our system receive streams that are
not currently being tapped, so a stream is only received when there is interest
in it. The next step is to allow the user to reduce the stream, if applicable. For
operations like change detection, users are often interested in specific parts of a
Fig. 2. Component Nature of the Data Stream Processing Engine.