Paper with @myp.bsky.social on agreement variation in Portuguese is out! What if ϕ-features don't come in a single bundle, and "uninterpretable" features are actually interpretable along a socio-expressive dimension of meaning? We also propose a register calculus! revistes.uab.cat/isogloss/art...
Depends what one means by “interface”. If it’s a loose cover term for lawful interactions among domains, then trivially yes. The harder question imo is whether there are pragmatic inferences that are purely form-driven, not due to subtle differences in semantic strength — legit manner implicatures.
Also @elinm.bsky.social on “man”. I think they’re a bit different (“man” functions more easily as an expressive modifier without implying reference to the addressee), but there is definitely some overlap: link.springer.com/article/10.1...
I cannot recall if the subsequent subjective response to my prior belief resembled a presupposition failure or a mere revision of a false belief. Nor can I produce any such judgment now.
Something similar (but less dramatic) actually did happen to me once. Sitting on the subway, I looked at what was seemingly my reflection on the window opposite to me and though “Is my hair really this long???”. Then I realized I was looking at someone who was sitting right next to me.
Vem! Tu certamente tornarás este espaço mais divertido!
Isso não significa que a tua conclusão esteja errada. Tem boas razões para acreditar que LLMs, por si sós, sejam irremediavelmente alucinatórios e incapazes de apreender generalizações abstratas (i.e. conhecimentos) como nós humanos.
Se formos olhar para um corpus produzido inteiramente por LLMs, é provavelmente veremos uma relação inversamente proporcional entre frequência de lemmas e ranqueamento na tabela de frequências, assim como em corpora produzidos por humanos.
Não sei se entendi teu ponto, mas a lei de Zipf se aplica à distribuição de palavras em um corpus, mas não diz nada a respeito da probabilidade de uma palavra X ocorrer após uma palavra Y em um dado contexto. É essencialmente isso que LLMs fazem.
I think one could make the argument that the morphosyntax of BP is easier for English speakers than that of other Romance languages. Several quirks of Romance that English speakers struggle with (clitics, null subjects, rich inflection, inversions) are for the most part absent in colloquial BP.
asking for an image of a professor and asking to make it "more professor", a thread.
#PhD fellowships in the SFB Register in 2024:
Do you want to work for 1 year in a register related project in Berlin?
The fellowship can be used to carry out a self-contained research project on the topic of the SFB. More information: sfb1412.hu-berlin.de/news/new-cal...
Please share!
#linguistics
Great news! The #SFB1412 „Register: Language Users’ Knowledge of Situational-Functional Variation“ is getting #DFG funding for phase 2! 🎈🍸🎈 #linguistics
sfb1412.hu-berlin.de
My Fair Lady is basically about sociophonetics
I am a big fan of Peter Culicover’s “Natural Language Syntax”. It is not prototypical generative syntax, but he does cover the history and all of the major landmarks (passive, wh movement, etc). David Adger’s “Core Syntax” is also great and probably closer to what one sees in papers nowadays.
today in clitics
Academia isn’t about wealth, or fame, or approval. It’s about downloading as many pdfs as you possibly can until you retire without reading any of them
Hey @verbingnouns.bsky.social, could I be added to the feed?
Completely new to this. Happy to bring more #linguistics into this part of the universe (the existence of which I only discovered this week). Also interested in keeping up with what you all have been up to.