New Schema for Cluster Algorithms

lingpy / lingpy

LingPy: Python library for quantitative tasks in historical linguistics

http://lingpy.org

GNU General Public License v3.0

124 stars 34 forks source link

New Schema for Cluster Algorithms #46

Closed LinguList closed 11 years ago

LinguList commented 11 years ago

Currently, we cluster with

flat linkage algorithms and
markov clustering (newly added today!)

In the future, we might also want to carry out link clustering based on the link community detection algorithm. This, however, requires a new representation schema for cognate clusters, since they can be fuzzy, and currently, fuzzy cognate sets are NOT supported.

LinguList commented 11 years ago

Link clustering is now implemented, and the fuzzy results can also be written to file and handled in etymological dictionaries. However, in order to provide full compatibility, we also need to test these methods. Unfortunately, currently I don't know of any way to carry out tests of fuzzy cluster results in any consistent way. So this needs some future thinking.

LinguList commented 11 years ago

Closing this issue for the moment: fuzzy clustering is implemented in the classes, it can also be used to export to etymological dictionaries. Only thing that does not work at the moment is to test the accurracy of fuzzy clustering, but this is another issue for the future.