Understanding the nature of human language and developing computational systems with human-level language abilities are important goals from both scientific and medical perspectives. The increasing role of information technology within health-care settings creates a need for computational systems to communicate naturally with humans. Furthermore, computational models with human-level language ability would be a key step toward understanding the unique character of human intelligence and, since language involves such diverse cognitive processes, it provides a strong test of a unified theory of cognition.
This project will achieve four major advances:
1. A computational theory of language use that, for the first time, unifies several aspects of linguistic theory and cognitive science by grounding them in a single cognitive architecture.
2. An architecture that is consistent with knowledge about human cognitive processes and that provides a unified account of scalable abductive inference, reasoning about people’s beliefs, hierarchical skill execution, and structural learning.
3. A large-scale knowledge base of linguistic and domain content that supports human-level language use.
4. An implemented dialog system that demonstrates a significant advance in the robustness and complexity of natural language processing.