This invention describes a system that automatically improves a structured knowledge base, called an ontology, for use in understanding human language. It works by creating a model of the ontology, then using a technique called 'spreading activation' to connect parts of a text to concepts in the ontology and identify related ideas. The system then checks how well this connection process performs for a specific language task, like figuring out word meanings, and adjusts either the ontology itself or how it's used to improve future results.
Why it matters: Filed before the widespread adoption of large language models. These models can now provide rich data for evaluating ontology performance and could be used to automate the modification process, making the system far more effective.
AI gives you a few directions you could take this. Pick one, and we check whether your version is different enough to patent, then write the filing.
Reinvent this with AI