20 CLAIMS

• 1. A probabilistic learning element, that sequentially receives objects and outputs sequences of recognized states and includes context driven searching, said learning element comprising:
• means for sequentially receiving objects;
• long term memory means for storing in sequential context,
• previously learned states,
• objects contained in said previously learned states, and
• predetermined types of knowledge relating to said stored previously learned states and said objects contained in said previously learned states, whereby from any stored information in
• said long term memory means the stored information which occurs next in context is directly accessible;
• short term memory means for storing in sequential context said received objects;
• means for correlating said received objects stored in said short term memory means with information stored in said long term memory means, said correlation being facilitated by using the context of said objects stored in said short term memory means as a pointer to the context of said information stored in said long term memory means, said correlating means assigning probabilities to possible next states in a sequence of recognized states;
• means, responsive to said probabilities of possible next states, for determining a most likely next state;
• means, responsive to said objects stored in said short term memory means and said information stored in said long term memory means, for providing a signal corresponding to a probability that a state has ended; and
• means, responsive to said end of state signal, for outputting said most likely next state as a recognized next state in a recognized state sequence.
• 2. A probabilistic learning element as described in claim 1, additionally comprising means for providing a rating of confidence in said recognized next state.
• 3. A probabilistic learning element as described in claim 2, additionally comprising:
• means for accumulating the ratings of confidence of recognized states of said recognized states sequence; and
• means, responsive to said accumulated ratings of confidence, for causing said long term memory means to store said recognized states, objects contained in said recognized states and the predetermined types of knowledge relating to said recognized states and said objects contained in said recognized states as previously learned states and information relating to previously learned states when said accumulated ratings of confidence exceed a predetermined threshold level.
• 4. A probabilistic learning element as described in claim 1, additionally comprising learning supervision means, responsive to external reinforcement signals, for causing said long term memory means to store said recognized states, objects contained in said recognized states and the predetermined types of knowledge relating to said recognized states and said objects contained in said recognized states as previously learned states and information relating to previously learned states.
• 5. A probabilistic learning element as described in claim 1, additionally comprising:
• means for providing a rating of confidence in said recognized next state;
• learning supervision means, adapted to receive said rating of confidence and an external reinforcement signal, said means being responsive to accumulated ratings of confidence of the recognized states of said recognized state sequence and said external reinforcement signal for providing an output signal when either said accumulated ratings of confidence exceed a predetermined threshold level or an external reinforcement signal is received; and
• means responsive to said output signal from said learning supervision means for causing said long term memory means to store said recognized states, objects contained in said recognized states and the predetermined types of knowledge relating to said recognized states and said objects contained in said recognized states as previously learned states and information relating to previously learned states.
• 6. A probabilistic learning element as described in claim 1, wherein the predetermined types of knowledge that are stored include the number of occurrences of: each object, sequences of objects, states and sequences of states.
• 7. A probabilistic learning element as described in claim 6, wherein the predetermined types of knowledge that are stored additionally include state lengths and the number of their occurrences and sequences of state lengths and the number of their occurrences.
• 8. A probabilistic learning element as described in claim 7, wherein the predetermined types of knowledge that are stored additionally include state-length pairs and the number of their occurrences and sequences of state-length pairs and the number of their occurrences.
• 9. A probabilistic learning element as described in claim 8, wherein the means for correlating includes a first means for determining the probabilities that possible states will span an object sequence having a particular begin time and end time and second means for determining the probabilities of state-length pairs given the previous state-length pair context.
• 10. A probabilistic learning element as described in claim 9, additionally comprising means responsive to the previously mentioned probabilities to implement an algorithm to provide probabilities of possible states, with a particular length that span an object sequence given the previous state-length pair context.
• 11. A probabilistic learning element as described in claim 1, wherein the long term memory means comprises a context organized memory including a plurality of tree structures for storing previously learned states and the objects contained therein and predetermined types of knowledge relating to the stored objects and states.
• 12. A probabilistic learning element as described in claim 11, wherein a tree structure used to store objects of learned states includes at each node thereof an object that is a part of a state whereby the tree includes sequences of objects shown in context.
• 13. A probabilistic learning element as described in claim 12, wherein each node of the tree has stored therein the number of occurrences of the object in the context represented by the node.
• 14. A probabilistic learning element as described in claim 11, wherein a separate tree structure is provided for storing in context learned states, objects contained in learned states, state lengths and state-length pairs.
• 15. A probabilistic learning element as described in claim 1, wherein said long term memory means comprises a context organized memory including a plurality of tree structures for storing the various types of stored information in sequential context, each tree includes a plurality of sequentially connected nodes with each node storing an item of information and having a record of the stored item, said record including:
• the number of occurences of the stored item in the sequential context of stored items represented by the node,
• a pointer to a node which represents the same sequential context stored items without the last item of the sequence,
• a pointer to a node which represents the same sequential context of stored items with one additional stored item at each end of the sequence, and
• a pointer to a node which represents the same sequentialy context of stored items without the first stored item of the sequence.
• 16. A probabilistic learning element, that sequentially receives objects and outputs sequences of recognized states and includes context driven searching, said learning element comprising:
• means for sequentially receiving objects;
• long term memory means for storing in sequential context,
• a plurality of previously learned states,
• objects contained in said previously learned states, and
• predetermined types of knowledge relating to the stored previously learned states and said objects contained in said previously learned states, whereby from any stored information in said long term memory means the stored inforamtion which occurs next in context is directly accessible;
• short term memory means for storing in sequential context said received objects;
• means for correlating said received objects stored in said short term memory means with information stored in said long term memory means by implementing an nth order Markov process correlating several levels of stored context, said correlation being facilitated by using the context of said objects stored in said short term memory means as a pointer to the context of said information stored in said long term memory means, said correlating means assigning probabilities to possible next states in a sequence of recognized states;
• means, responsive to said probabilities of possible next states, for determining a most likely next state;
• means, responsive to said objects stored in said short term memory means and said information stored in said long term memory means, for providing a signal corresponding to a probability that a state has ended; and
• means, responsive to said end of state signal, for outputting said most likely next state as a recognized next state in a recognized state sequence.
• 17. A probabilistic learning element as described in claim 16, wherein said long term memory means comprises a context organized memory including a plurality of tree structures for storing the various types of stored information in sequential context, each tree includes a plurality of sequentially connected nodes with each node storing an item of information and having a record of the stored item, said record including:
• the number of occurrences of the stored item in the sequential context of stored items represented by the node,
• a pointer to a node which represents the same sequential context of stored items without the last item of the sequence,
• a pointer to a node which represents the same sequential context of stored items with one additional stored item at each end of the sequence, and
• a pointer to a node which represents the same sequential context of stored items without the first stored item of the sequence, whereby a variable order Markov process is efficiently implemented.
• 18. A method of recognizing a sequence of states from sequentially inputted objects utilizing a probabilistic learning element, comprising the steps of:
• sequentially receiving said objects;
• storing in a short term memory means said received objects in sequential context;
• storing in a long term memory means in sequential context previously learned states, objects contained in said previously learned states, and predetermined types of knowledge relating to said stored previously learned states and said objects contained in said previously learned states, whereby from any stored information in said long term memory means the stored information which occurs next in context is directly accessible;
• correlating said received objects stored in said short term memory means with information stored in said long term memory means, said correlation being facilitated by using the context of said objects stored in said short term memory means as a pointer to the context of said information stored in said long term memory means;
• assigning probabilities to possible next states in a sequence of recognized states;
• determining a most likely next state each time a new object is received;
• determining when a state has ended; and
• outputting said most likely next state as a recognized next state in a recognized state sequence when a state has ended.
• 19. A probabilistic learning element, that sequentially receives objects and outputs sequences of recognized states and includes context driven searching, said learning element comprising:
• means for sequentially receiving objects;
• long term memory means for storing in sequential context,
• previously learned states,
• objects contained in said previously learned states, and
• predetermined types of knowledge relating to said stored previously learned states and said objects contained in said previously learned states, whereby from any stored information in said long term memory means the stored information which occurs next in context is directly accessible;
• short term memory means for storing in sequential context said received objects;
• means for correlating said received objects stored in said short term memory means with information stored in said long term memory means, said correlation being facilitated by using the context of said objects stored in said short term memory means as a pointer to the context of said information stored in said long term memory means, said correlating means assigning probabilities to possible next states in a sequence of recognized states;
• means, responsive to said probabilities of possible next states, for determining a most likely next state;
• means, responsive to said objects stored in said short term memory means and said information stored in said long term memory means, for providing a signal corresponding to a probability that a state has ended;
• means, responsive to said end of state signal, for outputting said most likely next state as a recognized next state in a recognized state sequence;
• learning supervision means, responsive to external reinforcement signals, for causing said long term memory means to store said recognized states, objects contained in said recognized states, and the predetermined types of knowledge relating to said recognized states and said objects contained in said recognized states as previously learned states and information relating to previously learned states; and
• means for correcting a recognized state sequence prior to initiating an external reinforcement signal.
• 20. A probabilistic learning element, that sequentially receives objects and outputs sequences of recognized states and includes context driven searching, said learning element comprising:
• means for sequentially receiving objects;
• long term memory means for storing in sequential context,
• previously learned states,
• objects contained in said previously learned states, and
• predetermined types of knowledge relating to said stored previously learned states and said objects contained in said previously learned states, whereby from any stored information in said long term memory means the stored information which occurs next in context is directly accessible;
• short term memory means for storing in sequential context said received objects;
• means for correlating said received objects stored in said short term memory means with information stored in said long term memory means, said correlation being facilitated by using the context of said objects stored in said short term memory means as a pointer to the context of said information stored in said long term memory means, said correlating means assigning probabilities to possible next states in a sequence of recognized states;
• means, responsive to said probabilities of possible next states, for determining a most likely next state;
• means, responsive to said objects stored in said short term memory means and said information stored in said long term memory means, for providing a signal corresponding to a probability that a state has ended;
• means, responsive to said end of state signal for outputting said most likely next state as a recognized next state in the recognized state sequence;
• means for providing a rating of confidence in said recognized next state;
• learning supervision means, adapted to receive said rating of confidence and an external reinforcement signal, said means being responsive to accumulated ratings of confidence of the recognized states of said recognized state sequence and said external reinforcement signal for providing an output signal when either said acumulated ratings of confidence exceed a predetermined threshold level or an external reinforcement signal is received;
• means responsive to said output signal from said learning supervision means for causing said long term memory means to store said recognized states, objects contained in said recognized states and the predetermined types of knowledge relating to said recognized states and said objects contained in said recognized states as previously learned states and information relating to previously learned states; and
• means for correcting a recognized state sequence prior to initiating an external reinforcement signal.