Setting Up Recommendation Indexes

Implementing a solution that uses the Recommendation Engine involves both developing the application itself and instituting a process for creating and updating the user and document profiles.

Both user and document profiles evolve over time, but it may be important to seed them initially with sufficient information to make them immediately useful. As noted earlier, document profiles are seeded with automatically extracted feature information that is generated when the documents are first indexed into a collection. User profiles do not automatically contain any initial information, but the Recommendation Engine provides an XML import interface you can use to seed that initial information. The information you import can come from many sources, including the following:

Customer data from a Customer Relationship Management (CRM) system


Employee data from human resource or LDAP databases


Job descriptions from organization charts


Document relationships from clustering algorithms


For more information on the user-visible features a recommendation application can implement, see Implementing Recommendations. For more information or recommendation indexes and entity profiles, see the Verity K2 Recommendation Engine Guide.