Fisica statistica

What is the landscape of natural language? Insights from a random language model

by Eric De Giuli (ENS)

Europe/Rome
Conversi (Dipartimento di Fisica - Ed. G.Marconi)

Conversi

Dipartimento di Fisica - Ed. G.Marconi

Description

Many complex systems have a generative, or linguistic, aspect: life is
written in the language of DNA; protein structure is written in a language
of amino acids, and human endeavour is often written in text. Are there
universal aspects of the relationship between sequence and structure? I am
trying to answer this question using models of random languages. Recently I
proposed a model of random context-free languages [1] and showed using
simulations that the model has a transition from an unintelligent phase to
an ordered phase. In the former, produced sequences look like noise, while
in the latter they have a nontrivial Shannon entropy; thus the transition
corresponds to the emergence of information-processing in the language.
 
In this talk I will explain the basics of natural language syntax, without
assuming any prior knowledge of linguistics. I will present the results
from the model above, and explain how the model is related to complex
matrix models with disorder [2].
 
[1] https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.122.128301
[2] https://arxiv.org/abs/1902.07516

Organised by

Gabriele Sicuro