Relevance Engine: The relevance engine performs real-time matching of content with consumer profiles. It produces a list of content ordered by consumer propensity for consumption of that information.
Inference Engine: The inference engine learns the preferences of both the content editor and content consumers, and updates dynamically the appropriate consumer propensities. The inference engine also creates additional relationship links when necessary. The behavior of the content editor and/or content consumer in assigning personalization values to content is learned, using Bayes Nets and integrated with explicit profile information to update the consumer tags in an event driven manner. Additional engines for new algorithms can be created very easily and added to the server network using our engine factory interfaces. Our server presents a template that could be used to create weights and values for pre-population.
Concept Extraction Engine: The concept extraction engine looks into a document, a web page or a file of a given type and extracts key words, phrases and information using a Naïve Bayes’ engine in our engine factory server. This helps classify the file into the ontology based on the characteristics of the nodes including explicit and implicit meta information. These are used to drive certain events based on consumer tags.
Architecture: The software platform is built as a package consisting of several Java based servers that implement the complex Artificial Intelligence algorithms. These servers work based on templates where they can be adjusted for the business conditions. The servers are built as a robust, fault-tolerant and load-balanced servers that can use a communication interface of choice. It comes with native TCP/IP smart sockets and it can be integrated with messaging systems such as MQ, TIBCO or others. It is built to work on all relational database and file based systems. The package consists of an extensive API for integration. It is built for very high scalability.
