Overview Index

Event correlation analyzer

Event correlation analyzer is a hardcoded routine which searches for correlations between actions and events.

Event correlation analyzer is the key part of the experiment (and therefore is the key part of the learning process).

Event correlation analyzer finds out cause-effect relations.

It helps to find answers to the questions:

1) Which events are effects of the specified action?

2) Which actions are causes for the specified event?

Event correlation analyzer saves cause-effect relations into cause-effect relation table.

 

Event correlation analyzer analyzes the results of “Active experiment” and “Passive experiment”.

“Active experiment” scenario

PCnous does some actions. These actions affect the world. As a result some events happen in the world. Information about events is received by hardcoded units (for instance, by chat reader). Both actions and events are logged.

Event correlation analyzer tries to find cause-effect relations between the actions and the events.

Example 1:

PCnous accomplished “throw a stone” action. Event “window is broken” is received.

Event correlation analyzer saves information about relations between action “throw a stone” and event “the window is broken” into cause-effect relation “’throw a stone’ ->’ window is broken’”.

 

Example 2:

PCnous says “Hi” in chat client. Then PCnous can get response “Hello”.
Event correlation analyzer saves information about cause-effect relation between word “Hi” and word “Hello”.

 

“Passive experiment” scenario

PCnous perceives several events from the outer world. Event correlation analyzer tries to find relations between these events.

Example:

PCnous received 3 events:

1)      “John said ‘I love you’ to Mary”. (#1)

2)      “Mary said ‘I love you’ to John”. (#2)

3)      “John kissed Mary”. (#3)

Event correlation analyzer saves information about correlation between event #1 and event #2 into cause-effect relation “#1->#2”

Event correlation analyzer saves information about correlation between event #2 and event #3 into cause-effect relation “#2->#3”

Event correlation analyzer saves information about correlation between event #1 and event #3 into cause-effect relation “#1->#3”

 

Know-how

Event correlation analyzer assumes that the less time period between an action and an event the higher is probability that the action is a cause of the event. The less the time interval between the action and the event is, the more coherence attribute is increased by the event correlation analyzer.

 

Event Correlation Analyzer implementation

Event correlation analyzer analyses event log.

For each event in event log event correlation analyzer:

1) Finds previous events

2) Creates cause-effect relation between current event and the previous events. Coherence value is set in inverse proportion to the time interval between the current event and the previous event.

If cause-effect relation already exists then the coherence value of the relation is incremented.

 

See also:

Discover concept correlations routine

Hardcoded routines

Hardcoded routine list

Know-how