search

D-PLACE / dplace-legacy

D-PLACE: Database of Places, Language, Culture and Environment (OLD)
https://d-place.org/
MIT License
21 stars 4 forks source link

Language/Dialect visualisation problem in Glottolog Trees #324

Closed Bibiko closed 5 years ago

Bibiko commented 8 years ago

There's a tiny problem with a few circumstances where a society within a Glottolog family is assigned to a Glottolog language A and another society (same family) is assigned to a dialect of the language A. Example: societies Dan (Mande family, language Gweetaawu [gwee1242]) and Ngere (Mande family, language Dan [dann1241]). Here, Gweetaawu is a dialect of the language Dan. Since the Glottolog trees don't show the node names, only the society Dan (language Gweetaawu) is shown in the tree view (Ngere not), and in the Global Tree we see the society Dan AND a not-connected dot Ngere. I believe that the strategies to display these dendrograms differ. Here the societies I found so far: Ngere [dann1241] Iglulik Inuit [west2618] Tazarawa [haus1257] Cariña [gali1262] Bambote [taab1238] Igbo [nucl1417] Noatak Inuit [nord2945] What should we do?

SimonGreenhill commented 8 years ago

@kirbykat can we assign the 'parent' societies to another at the dialect level?

kirbykat commented 8 years ago

@Bibiko @SimonGreenhill Sorry for slow reply - catching up after weekend conference. The scenario that @Bibiko describes for the family-level tree is what @stefelisabeth and I decided was the best strategy for these cases where there are multiple societies for a single language AND some are resolved to dialect but others only to language (i.e., we decided to show the society(ies) linked to dialect(s), rather than the society(ies) for which a dialect could not be confirmed).

All of these dialect-linked societies COULD easily be linked to their parent language, but then we would randomly select one society to display on the tree. Often there are 4 societies linked to one language, 3 of which are linked to a dialect. So, if we prioritize display of dialect-linked languages, we can display 3/4 on the family or global tree (as separate tips, branching out from the language), instead of the 1/4 societies on the single (extended) language tip.

So, if I understand how display on glottolog trees is currently set up (feel free to correct me) I would suggest prioritizing the dialect-linked societies for display on the global tree, as we do with individual family trees. @stefelisabeth Does this make sense to you? Do you remember setting this up for the family-level glottolog trees?

stefelisabeth commented 8 years ago

Do you mean we only display the tree tips (so if there's a family and a dialect, only the dialect will be displayed)? Initially, it would have displayed both Dan and Ngere, but one of those would have been an inner node rather than a tree tip (I think?). I can change it back to display both if you'd like.

kirbykat commented 8 years ago

I think we decided to only display the dialect (on the tip), because it was confusing having some societies displayed at nodes and others at tips. It sounds like this hasn't been implemented for the global tree.

stefelisabeth commented 8 years ago

@Bibiko any ideas?

Bibiko commented 8 years ago

hmm, actually not really. Maybe one could link a society which is actually linked to a dialect D1 to its parent language L1 but by changing its label to L1 (D1) whereby L1 and D1 are hyperlinks to Glottolog.

kirbykat commented 8 years ago

@Bibiko, do you have an up-to-date mapping file for glottolog ids -> parent ids -> iso codes? I have one you gave me last May, but I imagine there are updates. The file you sent me last had a column indicating if a glottolog id was a dialect or language. @stefelisabeth, you might already have this file, but if not it could serve as the key to automating D1->L1 links? (or maybe @Bibiko you are suggesting hyperlinks so that the links are always up-to-date?)

stefelisabeth commented 8 years ago

Labels for tree tips are society names, not languages, though. So by L1 (D1), do you mean parent society (dialect society)? For the aforementioned example Dan (Mande family, language Gweetaawu [gwee1242]) and Ngere (Mande family, language Dan [dann1241]). Here, Gweetaawu is a dialect of the language Dan, this would be Ngere (Gweetaawu). Or does the L1 (D1) go in the tooltip that pops up when you hover on a marker?

Bibiko commented 8 years ago

oops, sorry, I'm used to think in languages ;) - hmm, the problem is that the trees can only display leaves which are at the same level, i.e. either all societies are linked to dialects or to languages. My suggestion is for displaying of the trees only to link all societies to a language not to a dialect; example: Dan (Mande family, language Gweetaawu [gwee1242]) should be Dan (Mande family, language Dan [dann1241]). The tooltips etc. should display the most fine-grained mapping - in that case Gweetaawu [gwee1242].

stefelisabeth commented 8 years ago

Sorry to bring this up yet again (going through the Github issues to see what still needs resolving...) - I think markers for languages are shown (as well as dialects) presently, but the markers for legends are just not at the tips. I can leave it like this, unless we want to go ahead with @Bibiko 's suggestion. I think the only possible problem with that suggestion is that when more than one society shares the same language but have different dialects, we'll have to choose one to display on the tree at the language level.