The REST API’s allow us to access the data of the
services with different query term-combinations. To
ensure a valued outcome, different algorithms and
term-combinations are implemented in the system. All
services are queried with the same term combinations,
beginning with the five most valuable words, followed
by the four most valuable words and so on. To avoid
duplicate query outcomes, a near-duplicate detection
based on the Shingling-Jaccard algorithm was
implemented [3]. In the next step, the results are
presented to the idea generator in a spiral shape
ordered according to their relevance (see Figure 2, S4
and Figure 3). For the sorting of the results we use the
vector space model, in which we first collect all of the
words of a result and then calculate the according
frequency vector. The same is done with the words of
the idea. We then calculate the distances between the
points (idea frequency vector to result frequency
vector) and sort this conclusion in ascending order, i.e.
shorter distances are better. This also ensures that the
results, which are more closely related are visually
closer to the idea. The following screenshot shows the
presentation view with Alan’s messages and the
functionality used to mark and attach them to the idea.