Linkky: link key extraction based on relational concept analysis

Linkky is a proof-of-concept implementation of link key extraction based on formal and relational concept analysis [Atencia 2014d].

It takes into input two data sources and can extract the set of global link key candidates. It uses the Norris algorithm for performing FCA extended to deal with pairs of objects in the extent and quantified and qualified pairs of properties in the intent. The RCA processing has been implemented by iteratively applying the scaling operators to the formal contexts. RDF data can be loaded in the system through the RDF Library. The best global link key may be extracted through an extension of the unsupervised link key selection measures [Atencia 2014b] to global link key candidates.

It is currently possible to run:

$ python3 linkky.py -t ./txt_data/jd-rca2.txt -e ./export -f tikz -m
which will generate a full LaTeX report of the processing and/or the global link key lattices as this one:
The ideal way to use this tool would be to perform (not implemented yet):
$ python3 linkky.py -t0 datasource1.ttl -t1 datasource2.ttl -o alignment.rdf

Linkky is written in Python 3. It is not meant to be a robust and efficient implementation of the proposed ideas and it not planned to be maintained. It is only used for the purpose of performing research.

Linkky development/download site: https://gitlab.inria.fr/moex/linkky

Linkky was initially developed by Jérémy Vizzini for the purpose of his master thesis [Vizzini 2017a]. Linkky is made available under an MIT license.

References