A memory-based theory of verbal cognition.

Area of Science:

• Cognitive Science
• Computational Linguistics
• Natural Language Processing

Background:

• Existing models of verbal processing often rely on explicit grammars or predefined semantic roles.
• A need exists for memory-based computational models that can process natural language without these constraints.

Purpose of the Study:

• To introduce and describe the syntagmatic paradigmatic model for verbal processing.
• To demonstrate the model's capability in handling various linguistic and cognitive phenomena.
• To validate the model's effectiveness in extracting propositional information from unstructured text.

Main Methods:

• The syntagmatic paradigmatic model, a distributed, memory-based approach.
• Bayesian interpretation of string edit theory for cognitive control.
• Expectation maximization algorithm for lexical information extraction.
• Application to sentence processing, semantic categorization, recall, and inference tasks.

Main Results:

• The model effectively accounts for phenomena like sentence processing and semantic categorization.
• It demonstrates successful short-term serial recall and analogical/logical inference.
• The model extracted propositional information from a corpus of tennis news articles.

Conclusions:

• The syntagmatic paradigmatic model offers a viable alternative to grammar-based NLP.
• It shows that propositional information can be extracted from natural text without predefined structures.
• This approach advances memory-based computational models of verbal cognition.