Compute a Core Set of Meanings with Maximal Coverage for Languages and Concepts

[LinguList]

We need a function that searches for a maximal mutual coverage subset of many languages and many concepts. This is not trivial, but an approximate solution may suffice, e.g., a solution which sets a threshold, like 200 concepts -> give me the number of languages and which concepts. Or 1000 languages, give me the concepts for which I'll reach a mutual coverage of at least 0.8 per language pair.

[LinguList]

I am still hesitating whether this function would be needed for the lexibank study or not. Maybe we can ignore it, but it is a definite asset for later studies.

[xrotwang]

I think this is both, non-trivial and research-question related. Thus, coming up with an algorithm would probably require a concrete study which also shows the effects. For lexibank, I'd say chosing thresholds heuristically/ad-hoc is ok, in particular if we point out that the degree of comparability in lexibank already means progress in providing the basis for computing mutual coverage.

[LinguList]

I have thought about it again, and it turns out there is one concrete straightforward solution that can be implemented. The goal is to set the mutual coverage (which we introduced in already in lingpy) to some value (e.g., 0.8), and to then select the subset of languages which conform to this value. Definitely: an algorithm that proposes this would best be presented with a concrete use-case in mind (e.g., make an alternative to ASJP with more concepts or similar).

[LinguList]

So I think if I manage to prove my solution works, I'd at some point add it to cltoolkit, but you are right that a concrete study is useful, and this is definitely (as indicated by the milestone) post-lexibank.