borrowings not indicated

[LinguList]

In the paper, they use the color-code to mark where words come from, and if a word comes from Turkic but is something else, this would be a borrowing. The question is, how we should render this in cldf. Colorcodes can be detected, butt so far, they are not assigned to any cognate set.

[xrotwang]

They do not link to source words, though, right? In that case, I'd just add a "loan" column to FormTable.

[LinguList]

that's what I meant, basically.

[SimonGreenhill]

@chrzyki -- can you have a look at this one sometime please?

[chrzyki]

Yepp!

[chrzyki]

@LinguList could you please elaborate on this part?

... and if a word comes from Turkic but is something else, this would be a borrowing.

I've had a look at the source material and understand how the color coding works. However, what is not clear to me is what coding for which source language should be considered a Loan=True or Loan=False respectively.

[LinguList]

Okay, we need to add another layer of "cognate identifier" here. They do not seem to indicate source of a borrowing, or do they?

What we need is:

1. cognate identifier, normal cognate set, inside language families alone (including borrowings)
2. extended cognate set, only applying to those words that related across families (so here, you take the letters ABCD and turn them to cognate sets directly, before, you restrict them to families)
3. turn color code into source information: we have four color codes
   Green: terms of Turkic origin
   Blue: terms of Iranian origin
   Yellow: terms of Arabian origin
   Red: terms of Russian origin
   So the logic here is if the origin (by color) is different from the language group, it is a borrowing, and we also have the source, it is not only LOAN, but it is LOAN, SOURCE=Family, so the lexeme information should be expanded by a Loan_Source column, translating color-code to the Language, like Russian, Iranian, etc. But it should NOT be applied when the word is not borrowed.

This latter information, the source, could be added to the Lexeme or the Cognateset, I don't have a strong opinion here.

[chrzyki]

Thanks for the explanation. I'll see how this can be implemented.