Linguistics

Interesting paper, about the alleged ability of LLMs* to judge the grammaticality of sentences - something that humans are rather good at. Eight phenomena were tested, and LLMs performed extremely poorly.

*LLM = large language model. Stuff like Bard, ChatGPT, LLaMa etc. I'd argue that they aren't actual language models due to the absence of a semantic component, as shown by the article.

I'm creating this thread to hopefully promote a bit more activity in the community.

If you want to talk about something Linguistics-related, but for some reason you don't want to create a new post just for that, feel free to post it here instead.

cross-posted from: https://slrpnk.net/post/4507295

Institution: MIT Lecturer: Prof. Norvin A. Richards University Course Code: MIT 24.900 Subject: #linguistics Description: This class provides some answers to basic questions about the nature of human language. Throughout the course, we examine a number of ways in which human language is a complex but law-governed mental system. Much of the class is devoted to studying some core aspects of this system in detail; we also spend individual classes discussing a number of other issues, including how language is acquired, how languages change over time, language endangerment, and others.

Even if not solid research, I think that this article is worth sharing as food for though.

The author mentions Duncan's five faces of poverty (material, social, spiritual, aspirational and identity), then focuses on the later two, and proposes that language also plays a role in social poverty.

Superficially it might seen that the author proposes "replacive bilingualism" (i.e. linguicide) as a solution for this problem; he doesn't, he is mentioning it to highlight how individuals seek to address this linguistic poverty.

Make sure to give a check to the references cited - there's a lot of good stuff there.

The article shows a case of language contact (Tsimané vs. Spanish) triggering the conceptual split of a colour into two.

The word éxito in Spanish (and cognates in other iberian romance languages) has the meaning of success, but it is a cognate of English "exit".

According to Wiktionary, they all come from Latin "exitus", which is a participle of "exire", which literally means "to go out/outside, to exit, to leave".

Also on the Wiktionary page for this word is someone asking about this apparent semantic shift in Spanish, which got me wondering as well. Further googling only told me that it's not just Spanish but also Galician and Portuguese, possibly more.

Does anyone have information on how this shift developed? Or is the written evidence we have so poor that it might just as well have suddenly acquired the current meaning overnight as gradually over several generations and we wouldn't be able to tell?

The languages in the Indo-European family are spoken by almost half of the world’s population. This group includes a huge number of languages, ranging from English and Spanish to Russian, Kurdish and Persian.

Ever since the discovery, over two centuries ago, that these languages belong to the same family, philologists have worked to reconstruct the first Indo-European language (known as Proto-Indo-European) and establish a “language family tree”, where branches represent the evolution and separation of languages over time. This approach draws on phylogenetics – the study of how biological species evolve – which also provides the most appropriate model for describing and quantifying the historical relationships between languages.

Despite numerous studies, many questions still remain as to the origin of Indo-European: where was the original Indo-European language spoken in prehistoric times? How long ago did this language group emerge? How did it spread across Eurasia?

A great video explaining why the standard English vowel transcription system is outdated.

Invidious link

Fluency in ‘inglese farlocco’ has become necessary in Italy as hybrid words and off-kilter meanings proliferate.

As Evile and Pesetsky show in a newly published paper, "whom of which" obeys very specific rules, whose nature contributes to a larger discussion about sentence construction. The paper, "Wh-which relatives and the existence of pied piping," appears this month in the journal Glossa.

"It seems to be brand new, and it's very colloquial, but it's extremely law-governed," says Pesetsky, the Ferrari P. Ward Professor of Modern Languages and Linguistics at MIT.

I was discussing this with my fiance, and she agreed with me in that she also speaks English in this manner.

I have found that, at least personally, I tend to speak several common homonyms in English in distinct ways: bear/bare, they're/there, where/ware. It's difficult to describe the differences in a concise way, but I'll do my best, and maybe use IPA where applicable, assuming I'm not using them incorrectly?

The traditional pronunciation of bare is [ˈbɛr]. I would completely agree with this, and while the dictionary might also say bear is pronounced this way, I would argue that I often hear it more as [ˈber] — a more closed sound with the lips pulled back in a smile. Sure, sometimes people will lazily say both in the same manner, but if I say [ˈber], the listener is going to recognize in a vacuum that I am speaking of the furry mammal, not the term to describe a naked person.

Similarly, there is rendered as [ðɚ]. There is a perfect rhyme with bare. I agree with this. However, they're is given the treatment of being a contraction of "they are", and it similarly has that closed sounded [e] instead of [ə].

Am I crazy, or does anyone else out there experience English this way?

without the filler:

Excavations have been taking place at Boğazköy-Hattusha for more than century under the direction of the German Archaeological Institute (DAI).

Around 30,000 clay tablets have been found at the site to date, which have shed light on various aspects of life during the Hittite period, according to the Julius-Maximilians-Universität Würzburg. The tablets contain inscriptions in cuneiform—what is generally considered to be the oldest known writing system. Developed by the ancient Sumerians of Mesopotamia more than 5,000 years ago, cuneiform is a script that was used to write several languages of the ancient Near East.

Most of the inscriptions found at Boğazköy-Hattusha record the extinct Hittite language, which is the oldest attested member of the Indo-European family. Other languages, such as Luwian and Palaic, are also represented at the site.

However, excavations conducted this year, led by professor Dr. Andreas Schachner of the DAI's Istanbul Department, surprisingly uncovered a recitation of a previously unknown extinct language. The language was hidden on a cuneiform tablet containing a ritual text written in Hittite. The Hittite ritual text refers to the lost tongue as the language of the land of Kalašma, an area that likely corresponds to where the towns of Bolu or Gerede in northern Turkey are located today.

"The new language was written in cuneiform," Schachner told Newsweek. "It is the same writing system the Hittites used. The text is part of a longer text starting in Hittite. As it continues it says at one point: 'Continue in the language of the Land [of] Kalašma.'"

"The Hittites were uniquely interested in recording rituals in foreign languages," Daniel Schwemer, head of the Chair of Ancient Near Eastern Studies at Julius-Maximilians-Universität Würzburg, said in a press release.

The recently discovered language remains largely incomprehensible. However, Professor Elisabeth Rieken with the Philipps University of Marburg, Germany, a specialist in Anatolian languages, has confirmed that the Kalasmaic tongue belongs to the Indo-European family, according to Julius-Maximilians-Universität Würzburg.

EDIT: a more readable article with some other details here - https://www.uni-wuerzburg.de/en/news-and-events/news/detail/news/new-indo-european-language-discovered/

This video offers a nice introduction on the comparative method, used to reconstruct languages without direct attestation, and then talks a bit about the reconstruction of Proto-Indo-European. It's full of examples and rather accessible, even for people not well-versed in Historical Linguistics (or even Linguistics).

It's sometimes claimed that languages spoken by societies with large numbers of non-native speakers, and large heterogeneity of their native speakers, tend to simplify themselves over time. This study contradicts the claim, based on data for morphological complexity from 1k+ languages.

Phrases like know one's [general subject of interest] are very annoying to me because they seem rather self-centered. I am obviously fine with knows his way around or Know Your Customer because the use of possesive pronouns is appropriate.
On the other hand, now I know my ABCs is atrocious because the modern Latin alphabet obviously does not and never did belong to a single person, and has been used by billions of people in the last few centuries.

Do you know other English phrases with unnecessary posessive or personal pronouns? Do they exist in other languages? Is there a name for this linguistic phenomenon? Where do I complain? ~/s~

The study involved linguists and geneticists, and estimated the family to be around 8100 years old, with five main branches splitting off 7000 years ago or so. That fits neither the Kurgan/Steppe hypothesis nor the farming/Anatolian one. Instead the authors propose a hybrid hypothesis, with PIE spreading initially from the southern Caucasus; and then an IE branch going north, into the steppes, and spreading from there.

Personal note: that further hints that the similarities noticed between the NW Caucasian languages and the current PIE reconstructions aren't just a result of coincidence; they might be areal features. I wouldn't be surprised for example if what's currently reconstructed as *e *o was originally vertical, something like **ə **a (Ubykh style).

From what I’ve seen, writing was independently invented somewhere between 3 to 6 times. With so many languages and linguistic communities, why have so few independently invented writing?

I'm a mathsy scientist, not a linguist, so I'm coming at this from a different angle, but I find this blog by a linguist gives a great informal overview of applied category theory in linguistics.

Similar concepts from a mathematician's angle is here: https://www.math3ma.com/blog/language-statistics-category-theory-part-1 I really enjoy how complementary these perspectives are

The article provides a global analysis to model patterns of current and future language endangerment. In other words, it's trying to explain and predict where and how language loss happens, by measuring stuff.

Interesting excerpts of the article:

Our best-fit model explains 34% of the variation in language endangerment (comparable to similar analyses on species endangerment.

That's actually rather good, considering the global scale of analysis for something as messy as human beings, and how local political factors can revive or kill languages.

Five predictors of language endangerment are consistently identified at global and regional scales: L1 speakers, bordering language richness, road density, years of schooling and the number of endangered languages in the immediate neighbourhood.

I feel like linguists handling minority languages should already know thing by "gut feeling": small community, with lots of nearby languages, well-connected to other communities, being drilled by the government = threatened linguistic community. However, it's still great that the article is grounding that "gut feeling" into data.