Creating Profile Nets

With the K2 Profile Service, developers can build applications that classify documents for purposes other than taxonomy browsing.

The Profile Service is based on sophisticated technology that matches the terms and concepts in incoming documents to profile nets, which are sets of topics (stored queries) called interest profiles. The Profiler Service uses interest profiles to determine whether a given document is about a given subject. The application can then assign the document to the appropriate category or categories, which can in turn trigger messaging systems or cause the documents to be sent directly to interested users. The Profiler model is ideal for document classification, message routing, or push technologies.

(Note that, in contrast to the Verity engine, which indexes a number of documents and then evaluates individual queries against them, the Profile Service indexes a number of queries and then evaluates individual documents against them.)

For example, each email message arriving into a customer response system (Figure 3-7) might need to be routed to one of several service specialists. The Profiler application can match the content of each email to the stored interest profiles, and thereby know which specialist should receive the message.


Figure 3-7    Using the K2 Profiler for email routing



In this example, the content of email B most closely matches the profile associated with support specialist B, so the application sends the email to B.

The Verity administrator creates the interest profiles and compiles them into a profile net by using the mkprf command-line tool. Each profile in a profile net is a topic. In fact, a topic set can be converted directly into a profile net.



For more information on creating interest profiles and writing a K2 Profiler application, see the Verity K2 Profiler Programming Guide.