When can we expect the code to be shared?

[LinguList]

I read the paper with great interest and was also happy to see that the code is shared (which many people still don't do), but now I ask myself, given that the paper has been published, when we can expect the code to appear? Can you give a time line, or remind me by solving the issue, once this is done?

[j-luo93]

Hi Prof. List, huge fan of your work here! And thanks for your interest in our work.

Regarding the codebase, sorry for being late on this. I've been quite occupied with an ongoing project on sound laws recently, but I plan to finish it at the end of this month. So I think this codebase should be ready before mid-April.

[LinguList]

Nice, thanks for getting back to me and keeping me posted. BTW: if you are interested in testing the approach on fresh data from Sino-Tibetan languages, where one could run a corpus of Pyu languages against wordlists in Sino-Tibetan, let me know, and I can get you in contact with colleagues who have the data.

[justeuer]

Hi,
I would like to do some tests with the phoneme embeddings created by your model. Do you already know when the code & data will be available?

[j-luo93]

Hi,
I would like to do some tests with the phoneme embeddings created by your model. Do you already know when the code & data will be available?

Do you need the pretrained embeddings (optimized for decipherment) or just the code that composes them and you want to train them later for your task?

[justeuer]

The paper is not explicit about the embedding model used, but from the linked repository I infer that it is something CBOW-like. So both would be good, my idea is to first train embeddings on your data, compare them to your embeddings obtained from the decipherment task as a sanity check, and then train on my own data (which is provided by Mattis).

[j-luo93]

The paper is not explicit about the embedding model used, but from the linked repository I infer that it is something CBOW-like. So both would be good, my idea is to first train embeddings on your data, compare them to your embeddings obtained from the decipherment task as a sanity check, and then train on my own data (which is provided by Mattis).

Got it. I will then release the pretrained embeddings (I'll provide a link to download them), and prioritize the part of codebase that composes phonological embeddings from features. Both should be ready by the end of next week. Sorry for taking this long -- I've got things piling up towards my graduation

[txertatua]

Nice work! I'm interested on the iberian and basque corpus you used as system input, I was not able to find it within the shared code; could you commit it? Thank you very much.