Insights from a Random Language Model
Dr. Fatemeh Lalegani Dezaki
Toronto Metropolitan University
Friday, August 1, 2025
1:30 p.m.
In-person: QNC 1201
Abstract: Understanding how a child acquires syntax—the structural rules governing sentence formation—is fundamental to unraveling the mysteries of language acquisition. Various theories have been proposed to explain this process. One recent approach is the Random Language Model (RLM), introduced by Eric De Giuli, which applies concepts from statistical physics to the problem [1]. The model suggests that the acquisition of syntax in human language may involve a continuous transition from non-grammatical to grammatical structure.
In this talk, I will begin with a review of the Random Language Model. I will discuss how the model can be extended to more realistic scenarios—for example, by introducing an external field—and show that the qualitative behavior of the model remains unchanged. We also test the model using real human language data, demonstrating its robustness [2].
Next, I will describe an attempt to solve the model analytically in the high-temperature regime [3, 4]. Finally, I will present our theoretical approach for analyzing the model in the low-temperature regime, where we observe signs of a more stable phase compared to the unstructured, uniform- rule phase. This analysis employs the replica method, a powerful tool in the study of disordered systems, along with Feynman diagram techniques.
1. Eric De Giuli. Random language model. Phys. Rev. Lett, 122:128301, 2019.
2. Fatemeh Lalegani and Eric De Giuli. PRE, 109: 054313, 2024
3. Eric De Giuli. Emergence of order in random languages. J. Phys. A: Math. Theor., 52:504001, 2019.
4. Eric De Giuli. Corrigendum: Emergence of order in random languages. J. Phys. A: Math. Theor., 52:509501, 2019.