Family Trees and evaluation

[PhyloStar]

I looked at the neighbornets for the different families. The reticulation scores are within the range of 0.29-0.35 which is an indicator of tree signal. Here are my observations:

• Dravidian: The network captures the major subgroups right except for mixes a Central Dravidian language with South Dravidian language. This is mainly due to 20 languages size dataset.
• Indo-Iranian: Network shows a chain like structure with three subgroups: Iranian, Indo-Aryan, and Dardic coming out separate.
• Sino-Tibetan: Network is more of star shaped with the major subgroups coming out.
• Austro-Asiatic: Looks similar to Dravidian because of the 26 languages in the dataset.

Next steps (may be) to infer trees,

• remove the ancient languages and run MrBayes (uniform tree prior without calibration points)
• Compare the majority consensus tree with the reference standard tree. Is there any way to obtain the family reference tree from Glottolog using PyGlottolog without manual editing of the newick file.

Finally, show the results and write up the paper. :)

[LinguList]

Yes, this is advancing well. I think, we may even do without the neighbor nets. If we have a nice CLLD app, this may even be more useful for now, we are currently discussing how this could be done with @xrotwang (one feature we discuss is how to represent sound inventories).

[PhyloStar]

Really cool! If both languages and phoneme inventories can be included in one map it would be really cool.

Do you want me to go ahead and set up MrBayes analysis? I think it is independent of the clld app, right? We will have the results in 2 days or so and then we can evaluate against Glottolog trees...

[LinguList]

Yes, why not? I'll try and check the code done so far during this week, so we have some review by two people here, which is probably useful.

[PhyloStar]

Done setting up the MrBayes analysis. We can expect results for larger families within a day. Generating the reference tree is remaining. Hope to get that done soonish and we can submit the paper by the end of May, may be.

[PhyloStar]

Yes, this is advancing well. I think, we may even do without the neighbor nets. If we have a nice CLLD app, this may even be more useful for now, we are currently discussing how this could be done with @xrotwang (one feature we discuss is how to represent sound inventories).

@lingulist Is this app web based or related to Python? Is there anything I can do? I will try to do my best.

[LinguList]

https://clld.org/2020/05/07/update.html

This is a recent post, showing, how you can use CLDF to make an app that you can browse. We still discuss the representation of phoneme inventories, but the other aspects of showing words in space, etc. are already there.

You could thus give it a try and see if it works.

[xrotwang]

@PhyloStar what do you mean by "languages and phoneme inventories in one map"? Something similar to PHOIBLE would be pretty straightforward, i.e. charts like here and absence/presence maps per phoneme à la https://phoible.org/parameters/1507B68E1E3108371C1F882C40902AA5#1/26/155

[PhyloStar]

@xrotwang I see. Thanks. I guess @lingulist had a different idea about showing phoneme inventories. I am somewhat confused here.

[PhyloStar]

@xrotwang I had a chat with @lingulist and a CLLD app like PHOIBLE would be really cool to have. How do we start off?

[PhyloStar]

I started to work from here: https://clld.org/2020/05/07/update.html

I am in the virtualenv created for cldfying lsi. I tried to do the 4th step and didn't succeed. clld initdb development.ini --cldf PATH/TO/cldf/petersonsouthasia/cldf/StructureDataset-metadata.json --glottolog PATH/TO/glottolog/glottolog

Obviously, I require the StructureDataset-metadata.json file within lsi/cldf/ path. How do I create this json file? Is it through cldfbench?

[LinguList]

cldf/cldf-metadata.json

[LinguList]

It's in our lsi directory.

[LinguList]

The metadata file contains all the information on the content in the cldf data. It is the core of cldf.

[xrotwang]

Ah, for lsi the metadata is in cldf/cldf-metadata.json. @PhyloStar I can setup an initial repos at clld/lsi later today, and add you as colaborator.

[PhyloStar]

After running setup.py in clld_lsi, I tried the cldf/cldf-metadata.json, I get the error: KeyError: 'CodeTable'

I will wait for the initial repos. There might be something that needs to be fixed.

[LinguList]

Did you try and debug, tracing the error?

[xrotwang]

@PhyloStar you may have created the project template for the wrong CLDF module, i.e. not for a Wordlist but for a StructureDataset. In that case, any steps down the road would fail. But yes, wait for the initial repos. This functionality is rather new and may still be buggy.

[PhyloStar]

@xotwang: I did create template for StructureDataset. Ahh, I see. I will wait.