6.6. The Working Solution: Codexa in Action

Figure 6.5 is a visual representation of the flow of a data item through the Codexa system. The Harvester monitors Web and traditional data sources. When it finds an item of interest, the item is sent to an Extractor for conversion into a valid XML document. When the Extractor is finished, it stores the document in PCA and sends a JMS message to a Classifier, which decides which of the Wilshire 5000 companies are featured in the item. The Classifier stores its information, then passes the item on via JMS to appropriate Evaluators and Knowledge Filters, which operate in parallel to figure out what kind of information is in the item (for example, mergers and acquisitions, price manipulation, and so forth) and the relevance of the information. When the filtering is done, users are notified of items of interest via the KnowledgeMQ. Users can invoke reporting services to create real-time reports from items in the PCA or historical reports (trends and aggregates) of older items in the database management system.

From the standpoint of the layered architecture, all user interactions with the system are much the same as shown in Figure 6.6, which shows a user request for a report from the Codexa Service. The request is handled by a servlet, which calls an appropriate EJB, which in turn goes to the JNDI to invoke a service in a CORBA process. The CORBA process returns the requested information to the user in XML form via the EJB and servlet.