Information Technology Reference
In-Depth Information
The association rules obtained can be used to suggest other
relevant services frequently used by peers, when a scientist has
already put some services into his/her incomplete workflow. Due
to the lack of a large number of transactions (i.e., workflows), the
association rules we have obtained have low support value and may
not all make much sense. Despite this fact, the feedback from caBIG
users shows that these rules are quite informative to introduce relevant
services from a large set into their experiment. The reason is that
scientific workflows are explorative and less repetitive than their
business counterpart, and therefore even the association rules with
low support are noteworthy. Some users have told us that they are
enthusiastic of the services recommended to them,
in terms of
generating innovative data and unexpected results.
8.3.4 What is an Operation Chain Between
Services/Operations
Now we present the ServiceMap approach to address Q2:
Given two
or more services I want to use together, what is a path between
them?
Current scientific artifact repositories such as BioCatalogue [172]
and myExperiment typically adopt keyword-based search. The idea is
to index an entity as a vector of keywords and use the TF-IDF (term
frequency-inverse document frequency) algorithm [226] to measure the
weight of each keyword. Methods have recently been developed to
search substructures in a tree- or graph-like structure [227] or over
nested workflows [228]. These approaches suffer from two limitations.
First, each result comes from a single document. For example, two
workflows cannot be concatenated as a result, even by doing that you
can chain two services. Second, sequential relationships cannot be
established between keywords. For example, one can search for a
workflow containing both services
foo
and
bar
, but cannot define the
order of their appearance.
To answer Q2, we make use of the directed
operation network S
0
in
which nodes represent operations in services, and a directed edge
represents a data link between two operations in an abstract workflow.
S
0
can be seen as a directed map in which service operations are places,
and workflows are routes connecting them.
