Products

The Inferencing Agent(tm)

Pontimax’s Inferencing Agent offers a novel approach to Complex Event Processing (CEP) through its use of both deductive and inductive inferencing capabilities.  Complex pattern recognition can be accomplished through direct specification of the fact patterns of interest via the use of predicate expressions.  These predicate expressions allow the user to powerfully yet concisely communicate sensor data pattern descriptions as well as “spatio-temporal” fact patterns to the Inferencing Agent to directly drive its inferencing operations.  The sensor data patterns describe the sensor occurrences in terms of the number, duration and location of the occurrence pattern of interest as well as the timeframe of interest. 

Spatio-Temporal fact patterns provide a way to pose assertions such as, “Was the object of interest continuously at a particular location for a specified time period in the past?  Have they been at their current location for some threshold duration without moving?  A spatio-temporal fact pattern provides a general means for assertions involving presence at location and time at locations to be posed to the Inferencing Agent for spatio-temporal predicated inferences.   Multiple dependent inferences can be specified, called predicate chains, to increase the assurance that the event or activity of interest has in fact occurred.  Hierarchical sets of inferencing entities can be defined such that very complex pattern recognition can be accomplished.   Hierarchical plans with goal tracking can also be specified.  The “inferencing entities” can be any logical concept representing what the user is interested in detecting.  For example, the occurrence (or non-occurrence) of one or more fact patterns can be defined as an “event”.  One or more events can be defined as inferring the occurrence of an “activity”.

Consequent actions can be configured to be performed upon the occurrence or non-occurrence of the fact pattern or inferencing entity.  Currently, there are four types of consequent actions that can be specified:   an e-mail message with a specified subject and body, a cellular SMS text message with a specified subject and body and a voice message.  The latter is comprised of text for which a text-to-speech generation is done.  The fourth type of consequent action is a database record update or insertion, based on an SQL meta specification.

The Inferencing Agent is currently in initial operation supporting an Aware Remote Caregiver System For Oversight of Developmentally Disabled Individuals Living Independently.  Development of this system was partially funded by a grant from the National Institute For Disability & Rehabilitation Research (NIDRR) in partnership with the Eugene Research Institute and Assistech Systems.