We Hebben Een Serieus Translatie: Modeling Intercomprehension as Probabilistic Inference

2026-07-13

AI Summary

This study investigates the cognitive mechanisms enabling native speakers to achieve cross-linguistic mutual intelligibility without prior exposure to the target language, under conditions of linguistic relatedness. It introduces the first zero-shot computational model that formalizes mutual intelligibility as a Bayesian inference task within a noisy channel framework: the model scores potential translations of L2 utterances using only an L1 language model and infers L1–L2 lexical mappings through a general-purpose noise model grounded in orthographic similarity or symbolic regularities. Human behavioral experiments demonstrate that the model’s predictions significantly outperform both ablated variants and zero-shot prompting of large language models, closely aligning with the distribution of human mutual intelligibility judgments. These results substantiate the model’s cognitive plausibility and practical utility in modeling spontaneous cross-linguistic comprehension.

Abstract

Intercomprehension refers to partial intelligibility of an unfamiliar language (L2) by a speaker of a related language (L1). How is this zero-shot cross-language comprehension possible? In this work, we extend past work on algorithmic models of noisy-channel inference to model intercomprehension in a Bayesian framework. The model uses an LM in L1 only for scoring latent hypotheses about the translations of observed L2 utterances, and a general-purpose noise model to infer a mapping between L2 and L1 words based on either form-based similarity or symbolic rules. We then conduct a human behavioral experiment, eliciting inferences for utterances in Dutch, Italian, and Ukrainian from speakers of English, Spanish, and Russian, respectively. Our full model shows a closer alignment to the distribution of human intercomprehension performance than ablations, and also compares favorably to zero-shot prompting of much larger models. These results provide a cognitively plausible computational model of intercomprehension, and highlight the flexible inferences made by comprehenders under wide uncertainty in real-world cross-language scenarios. We share our code publicly.

Problem

• intercomprehension
• cross-language comprehension
• zero-shot inference
• language intelligibility
• Bayesian modeling

Innovation

• intercomprehension
• Bayesian inference
• noisy-channel model
• zero-shot cross-lingual comprehension
• cognitively plausible modeling