koenedaele
commented
7 years ago Hi Rémi,
glad that you find it interesting. I'll get back to you on the specific license, but it's certainly open source. So, no problem to use in research whatsoever.
The KVD relations are a few extra relations I thought were useful. I talk about them on page 32 of my thesis, but unfortunately that's all in Dutch. The kvd_before (or bef) is basically allen_before or allen_meets. The idea came from the observation that under the allen_relations '1900-1914' before '1914-1918' would be false. Which seems rather strange from an archaelogist or historian's point of view. The kvd_during (or dur) actually appears in the Allen paper about relations. The kvd_intersect is just a simple do these two intervals have a point in common, similar to the postgis functions. From my end users point of view this is more often than not the only relation they care about, which was noted by Nagypal and Motik (G. Nagypal & B. Motik 2003. A fuzzy model for representing uncertain, subjective and vague temporal knowledge in ontologies).
I'm not sure what kind of docs you've already found. Possible docs:
- https://www.researchgate.net/publication/259717493_Imperfecte_tijdsmodellering_in_historische_databanken (The Thesis)
- https://www.researchgate.net/publication/266750215_Modelling_Imperfect_Time_in_Datasets (A paper written for CAA 2012)
- https://www.researchgate.net/publication/265844759_HANDLING_IMPERFECT_TEMPORAL_INFORMATION_IN_A_DATABASE (A presentation I gave in Paris in 2012)
- https://www.researchgate.net/publication/265844950_STORING_IMPERFECT_TEMPORAL_INFORMATION_ABOUT_CULTURAL_HERITAGE_IN_A_RELATIONAL_DATABASE (A second presentation I gave in Paris, this time in 2014)
There are two repositories with code. This one and https://github.com/koenedaele/pgFTI. They use the same underlying algorithms, but this one only implements the efficient version by Schockaert et al in a C module. The other one implements different algorithms, but all of it in PL/PGSQL and by using Postgis (which did have the benefit of being able to load a list of periods in QGIS). If you're concerned about performance and are ok with using simple fuzzy intervals, this one is best (and what we will be using). If you want to be able to use more complicated Fuzzy Time Intervals and don't need as much performance, the other repository can help you.
PR's are certainly welcome.
Hey @koenedaele, I'm very glad to have found this work, it suits perfectly what I'm trying to do. I have a few questions/ remarks:
what paper are the "Kvd relations" referring to ? I found your master thesis, it seems to be
[6] Steven Schockaert, Martine De Cock, and Etienne E. Kerre, “An efficient characterization of fuzzy temporal interval relations,” in Proceedings of the 2006 IEEE World Congress on Computational Intelligence, 2006, pp. 9026–9033.
There are plenty of option for string/date/int -> sfti, but I can't find any option for sfti->array/record. Sure I can do this manually , but it is ugly and ineffective :
SELECT sfti_ar FROM sfti_makesfti(1783, 1785, 1791, 1799) as f , trim(both '()' from f::text) as ar , regexp_split_to_array(ar, ',') as sfti_ar --{1783.000000,1785.000000,1791.000000,1799.000000,1.000000}
If you are OK I can extend the doc in the readme and make a pull request. I'll need more functions anyway, such as cast to geometry for easy visualisation, interesection function and so
Cheers, Rémi-C