tag:blogger.com,1999:blog-1300680252997007251.post6511970427265981882..comments2024-03-01T18:56:18.314+11:00Comments on Humans Who Read Grammars: A Global Tree of LanguagesHedvig Skirgårdhttp://www.blogger.com/profile/03689179680848604827noreply@blogger.comBlogger3125tag:blogger.com,1999:blog-1300680252997007251.post-66151803966174571792016-07-12T03:21:09.354+10:002016-07-12T03:21:09.354+10:00Hi Erich,
Thanks for the reminder. Yes, you are r...Hi Erich,<br /><br />Thanks for the reminder. Yes, you are right, there is some previous work on phonological inventories and phylogenetics, including your paper (Rama and Borin 2015 also look at bigram inventory distances: Rama, Taraka, and Lars Borin. "Comparative evaluation of string similarity measures for automatic language classification." Sequences in Language and Text 69 (2015): 171). This topic definitely merits some more thorough discussion in a longer paper. <br /><br />With "methodological innovation" I was referring to the idea of combining lexical and phonetic information during phylogenetic inference.<br /><br />Cheers, GerhardAnonymoushttps://www.blogger.com/profile/02843330944287807134noreply@blogger.comtag:blogger.com,1999:blog-1300680252997007251.post-38408969626712268692016-07-11T21:45:48.501+10:002016-07-11T21:45:48.501+10:00Hi Erich - Thanks very much for the interesting pa...Hi Erich - Thanks very much for the interesting paper, I will have a look at it and also describe it in a follow-up post (with your permission). I think that is the main thing to continue to explore in the ASJP - if as Gerhard says the signal for macro-families is mainly from phonological similarity, then we need to work out what is going on there (how much is contact for example).Jeremy Collinshttps://www.blogger.com/profile/02949376439100679223noreply@blogger.comtag:blogger.com,1999:blog-1300680252997007251.post-45849253754908474162016-07-11T21:01:30.862+10:002016-07-11T21:01:30.862+10:00Hi Jeremy (and Gerhard)
This is a great discussio...Hi Jeremy (and Gerhard)<br /><br />This is a great discussion of very interesting research, and I'm happy to see some of my own recent work being picked up on, even if not explicitly.<br /><br />Jeremy, you write about the use of bigrams, "This method was not used in Jäger's PNAS paper, and the authors give no justification for using it here". Gerhard refers to it as “the most important methodological innovation of the paper”.<br /><br />The justification and the antecedent for this approach, I would guess, is a paper by Jayden Mackiln-Cordes and me, published after Gerhard’s PNAS paper and before the EvoLang paper, and presented at a conference hosted by Gerhard's research group in November 2015. It's available here:<br /><br />https://www.academia.edu/20320642/High-Definition_Phonotactics_Reflect_Linguistic_Pasts<br />http://dx.doi.org/10.15496/publikation-8609<br /><br />In that paper we show that simple bigram phonotactics contains phylogenetic signal. Interestingly however, we demonstrated that using binary bigram presence/absence scores recovers poor phylogenetic signal at best, whereas richer, continuous probabilistic data does recover good signal, at least in shallow genealogical groups.<br /><br />I'll be interested to read discussion of the EvoLang results in relation to our own findings in the follow-up paper Gerhard has foreshadowed here.<br /><br />Very best, Erich.Erich Roundhttp://languages-cultures.uq.edu.au/ancient-language-labnoreply@blogger.com