Modeling Coevolution between Language and Memory Capacity during Language Origin

Accelerating language emergence by functional pressures

In language emergence, neural agents acquire communication skills by interacting with one another and the environment. Through these interactions, agents learn to connect or ground their observations to the messages they utter, forming a shared consensus about the meaning of the messages. Such connections form what we refer to as a grounding map. However, these maps can often be complicated, unstructured, and contain redundant connections. In this paper, we introduce two novel functional pressures, modeled as differentiable auxiliary losses, to simplify and structure the grounding maps. The first pressure enforces compositionality via topological similarity, which has been previously discussed but has not been modeled or utilized as a differentiable auxiliary loss. The second functional pressure, which is conceptually novel, imposes sparsity in the grounding map by pruning weaker connections while strengthening the stronger ones. We conduct experiments in multiple value-attribute environments with varying communication channels. Our methods achieve improved out-of-domain regularization and rapid convergence over baseline approaches. Furthermore, introduced functional pressures are robust to the changes in experimental conditions and able to operate with minimum training data. We note that functional pressures cause simpler and more structured emergent languages showing distinct characteristics depending on the functional pressure employed. Enhancing grounding map sparsity yields the best performance and the languages with the most compressible grammar. In summary, our novel functional pressures, focusing on compositionality and sparse groundings, expedite the development of simpler, more structured languages while enhancing their generalization capabilities. Exploring alternative types of functional pressures and combining them in agent training may be beneficial in the ongoing quest for improved emergent languages.

Informativeness across Interpreting Types: Implications for Language Shifts under Cognitive Load

Previous quantitative studies discussing interpreting types have focused on various features of linguistic forms in outputs. However, none of them has examined their informativeness. Entropy, as a measure of the average information content and the uniformity of the probability distribution of language units, has been applied to quantitative linguistic research on different types of language texts. In the present study, entropy and repeat rate were used to investigate the difference of overall informativeness and concentration of output texts between simultaneous interpreting and consecutive interpreting. We intend to figure out the frequency distribution patterns of word and word category in two types of interpreting texts. Analyses of linear mixed-effects models showed that entropy and repeat rate can distinguish the informativeness of consecutive and simultaneous interpreting outputs, and consecutive interpreting outputs entail a higher word entropy value and a lower word repeat rate than simultaneous interpreting outputs. We propose that consecutive interpreting is a cognitive process which reaches an equilibrium between production economy for interpreters and comprehension sufficiency for listeners, especially in the case where input speeches are more complex. Our findings also shed lights on the selection of interpreting types in application scenarios. The current research is the first of its kind in examining informativeness across interpreting types, demonstrating a dynamic adaptation of language users to extreme cognitive load.

Rethinking foundations of language from a multidisciplinary perspective

The issue of language foundations has been of great controversy ever since it was first raised in Lenneberg's (1967) monograph Biological Foundations of Language. Based on a survey of recent findings relevant to the study of language acquisition and evolution, we propose that: (i) the biological predispositions for language are largely domain-general, not necessarily language-specific or human-unique; (ii) the socio-cultural environment of language serves as another important foundation of language, which helps shape language components, induce and drive language shift; and (iii) language must have coevolved with the cognitive mechanisms associated with it through intertwined biological and cultural evolution. In addition to theoretical issues, this paper also evaluates the primary approaches recently joining the endeavor of studying language foundations and evolution, including human experiments and computer simulations. Most of the evidence surveyed in this paper comes from a variety of disciplines, and methodology therein complements each other to form a global picture of language foundations. These reflect the complexity of the issue of language foundations and the necessity of taking a multidisciplinary perspective to address it.