next up previous
Next: Imperfect Communication Up: Communication Defined as Complementary Processes Previous: Basis for a Hierarchy

The Performance of Processes

All receiving informative processes within a communication system may be viewed as filtering processes, selectively providing information to a process immediately succeeding (above) it in the right half of the hierarchy. Some processes pass no information to succeeding neighboring processes, some processes selectively pass information, while others pass all input information to their neighbor.

A neighbor process whose output is input to a second process is referred to as a context for the second process. When all input to a particular process comes from the output of another process, the latter process is the full or complete context for the former process. The context may be viewed as providing a query or statement of information need for a filtering process.

A filtering process accepts information from the process below and, based upon the context and the interests and needs it represents, will transmit or not transmit information to the level above it, depending on the incoming information and the context provided from its predecessor and other processes. Information that is passed to a higher level process may modify the context itself and may thus affect future filtering actions. The filtering process may thus be recursive.

Consider a message with one binary characteristic which would occur in a noiseless informative process transmitting a single bit. The characteristic occurs with probability (and average frequency) pin messages to be passed (referred to as relevant messages) and probability (and average frequency) t in all messages [Los98]. If a filter attempts to pass only those messages at or above the average position of relevant messages in a list of messages ordered by decreasing relevance, the percent of all messages that will be passed is 1-(1-p+t)/2.Smaller values represent better filtering performance. Clearly, other filtering performance measures will be needed with more complex term relationships and probabilistic distributions, but this indicates the kind of result that may be obtained, increasing our understanding of communication systems.


next up previous
Next: Imperfect Communication Up: Communication Defined as Complementary Processes Previous: Basis for a Hierarchy
Bob Losee
1999-09-28