This review identifies the impact of semantic underspecification on NLP and cognitive efficiency, suggesting integrated models may improve context utilization.
Key Points
Semantic underspecification can create challenges in understanding language, impacting both NLP and human cognition.
Five stages of development trace the evolution of models to address these complexities, from Frege to modern language models.
Pragmatic inference plays a crucial role in how humans efficiently navigate language, despite inherent ambiguities.
Integrated neuro-symbolic architectures with uncertainty modeling could advance AI's language comprehension capabilities.