Model-guided search, discovery and learning are critical technologies that enable users to organize and make sense of the inundation of data that an Internet or database query has returned. To effectively sort through petabytes of information, Sarnoff technology performs three functions: modeling (creating representations of user behavior to predict future actions), learning (developing methods for software to learn by example), and mining (analyzing data to identify patterns or relationships).

Sarnoff’s technologies use Concept Maps to enable personalization and create a model of the user’s interests. We offer capabilities that include the ability to quickly find results of particular interest (data mining) [1] within a deluge of returned data, as well as the model’s ability to stay current as the user’s interests change. Query contextualization capabilities use ontologies to automatically build context around a user’s search. Additionally, automatic hypothesis generation uses evolutionary computation to learn proven knowledge about a given topic.

Model-guided search technology has a number of real-world applications. Web searches depend on model-guided search and discovery capabilities to sort relevant information from vast amounts of irrelevant data. Personalization for text and image searches, from both the Internet and databases, also depend on Sarnoff’s technology to learn and create dynamic and adaptive behavior models. Identifying search patterns through data mining also utilizes Sarnoff’s user modeling technology.