kbrbe / work-set-clustering Goto Github PK

Based on descriptive keys internal data structures are created and elements with overlapping descriptive keys are grouped together. Every time the clustering is performed on certain input data, new cluster identifiers are generated because UUIDs are used.

Currently there is no way to cluster possibly new elements to already existing clusters, and thus reusing existing cluster identifiers. The script should be equipped with new command line options to read existing clusters and their descriptive keys.

Technically we only have to prefill the internal data structures related to clusters and the rest of the clustering can stay the same.

This feature is needed for the BELTRANS project. More concretely for kbrbe/beltrans-data-integration#204 and possibly for kbrbe/beltrans-data-integration#234

Currently the description of the needed input is quite vague, we could add an example CSV with a few lines to the README.

Currently one list of descriptive keys can be given via the argument -i / --input.
For data integration of the MetaBelgica project we want to cluster with descriptive key files from multiple data sources.

We could either allow multiple values for the --input parameter or we could provide the list of input files via positional arguments.

Besides a unit test, there are no overall integration tests.