Tracking initiative in collaborative dialogue interactions

Claim 3

The method of claim 1 further including the step of computing in conjunction with the current turn of the dialogue at least one of a task initiative index and a dialogue initiative index for the next turn of the dialogue, using (i) the at least one cue observed during the current turn, (ii) the probability assignment associated with the observed cue, and (iii) a current value of at least one of the task initiative index and the dialogue initiative index, and wherein the predicting step includes predicting at least one of the task initiative holder and the dialogue initiative holder for the next turn of the dialogue, using at least one of the task initiative index and dialogue initiative index computed for the next turn.

Claim 4

The method of claim 2 wherein the current value of at least one of the task initiative index and the dialogue initiative index includes initial values of the task initiative index and the dialogue initiative index assigned prior to performing the computing, predicting and adjusting steps for a first turn of the dialogue.

Claim 5

The method of claim 1 wherein each of the probability assignments assigns a value in the range (0,1) to each of a plurality of possible initiative holders, such that the values for all the possible initiative holders sum to 1, and the value assigned to a given initiative holder indicates the amount of support a corresponding cue provides for the given initiative holder actually holding the initiative in the dialogue.

Claim 6

The method of claim 1 further including the step of repeating the predicting and adjusting steps for each of a plurality of turns of the dialogue, while swapping roles of speaker and hearer for each turn.

Claim 7

The method of claim 1 further including performing the predicting and adjusting steps for each of a plurality of turns of a first dialogue in order to determine appropriate probability assignments for a given set of cues, and using the resulting probability assignments to predict initiative shifts in a second dialogue.

Claim 8

The method of claim 1 wherein the adjusting step adjusts the probability assignment associated with a given cue such that a value of the probability assignment for an actual initiative holder is incremented by a constant amount for each incorrect prediction of that initiative holder, while another value of the probability assignment is decremented by the constant amount for another possible initiative holder.

Claim 9

The method of claim 1 wherein the adjusting step associates with each probability assignment for each cue a counter which is incremented when a correct prediction is made, and dccremented when an incorrect prediction is made.

Claim 10

The method of claim 8 wherein if the counter for a given cue is negative, the counter is reset to zero, and the probability assignment associated with a given cue is adjusted such that a value of the probability assignment for an actual initiative holder is incremented by a constant amount for each incorrect prediction of that initiative holder, while another value of the probability assignment is decremented by the constant amount for another possible initiative holder.

Claim 11

The method of claim 8 wherein the value of the counter is used to determine an amount of adjustment for one or more values of a given probability assignment.

Claim 12

The method of claim 10 wherein the probability assignment associated with a given cue is adjusted such that a value of the probability assignment for an actual initiative holder is incremented by Δ/2 count+1 for each incorrect prediction of the actual initiative holder, while another value of the probability assignment is decremented by Δ/2 count+1 for another possible initiative holder, where Δ is a constant and count is the value of the counter.

Claim 15

The apparatus of claim 12 wherein the processor is further operative to compute at least one of a task initiative index and a dialogue initiative index for the current turn of the dialogue, using (i) the at least one cue observed during the current turn, (ii) the probability assignment associated with the observed cue, and (iii) a current value of at least one of the task initiative index and the dialogue initiative index, and further wherein the processor predicts at least one of the task initiative holder and the dialogue initiative holder for the next turn of the dialogue, using at least one of the task initiative index and the dialogue initiative index computed for the current turn.

Claim 16

The apparatus of claim 13 wherein the current value of at least one of the task initiative index and the dialogue initiative index includes initial values of the task initiative index and the dialogue initiative index assigned prior to performing the compute, predict and adjust operations for a first turn of the dialogue.

Claim 17

The apparatus of claim 12 wherein each of the probability assignments assigns a value in the range (0,1) to each of a plurality of possible initiative holders, such that the values for all the possible initiative holders sum to 1, and the value assigned to a given initiative holder indicates the amount of support a corresponding cue provides for the given initiative holder actually holding the initiative in the dialogue.

Claim 18

The apparatus of claim 12 wherein the processor is further operative to repeat the predict and adjust operations for each of a plurality of turns of the dialogue, while swapping roles of speaker and hearer for each turn.

Claim 19

The apparatus of claim 12 wherein the processor performs the predict and adjust operations for each of a plurality of turns of a first dialogue in order to determine appropriate probability assignments for a given set of cues, and then uses the resulting probability assignments to predict initiative shifts in a second dialogue.

Claim 20

The apparatus of claim 12 wherein the processor is operative to adjust the probability assignment associated with a given cue such that a value of the probability assignment for an actual initiative holder is incremented by a constant amount for each incorrect prediction of that initiative holder, while another value of the probability assignment is decremented by the constant amount for another possible initiative holder.

Claim 21

The apparatus of claim 12 wherein the processor is operative to associate with each probability assignment for each cue a counter which is incremented when a correct prediction is made, and decremented when an incorrect prediction is made.

Claim 22

The apparatus of claim 19 wherein if the counter for a given cue is negative, the processor resets the counter to zero, and adjusts the probability assignment associated with a given cue such that a value of the probability assignment for an actual initiative holder is incremented by a constant amount for each incorrect prediction of that initiative holder, while another value of the probability assignment is decremented by the constant amount for another possible initiative holder.

Claim 23

The apparatus of claim 19 wherein the processor is operative to utilize the value of the counter to determine an amount of adjustment for one or more values of a given probability assignment.

Claim 24

The apparatus of claim 21 wherein the processor adjusts the probability assignment associated with a given cue such that a value of the probability assignment for an actual initiative holder is incremented by Δ/2 count+1 for each incorrect prediction of the actual initiative holder, while another value of the probability assignment is decremented by Δ/2 count+1 for another possible initiative holder, where Δ is a constant and count is the value of the counter.

Claim 25

A processing system for tracking shifts in initiative during a dialogue using a set of cues, the dialogue including a sequence of turns, the system comprising: a speech recognition unit for converting an input speech signal representing at least a portion of an actual dialogue to a digital format, wherein during a dialogue mode of operation the processing system acts as a participant in the actual dialogue; a processor having an input coupled to an output of the speech recognition unit for receiving the converted speech signal therefrom, wherein during a training mode of operation the processor is operative to predict in conjunction with a current turn of a training dialogue at least one of a task initiative holder and a dialogue initiative holder for a next turn of the training dialogue, using at least one cue observed during the current turn, and a probability assignment associated with the observed cue, and to adjust the probability assignment associated with the observed cue if at least one of the predicted task and dialogue initiative holders do not agree with corresponding respective actual task and dialogue initiative holders for the next turn of the training dialogue; a memory coupled to the processor for storing the probability assignment and the task and dialogue initiative indices; and a speech synthesizer having an input coupled to an output of the processor, the speech synthesizer receiving output signals from the processor corresponding to portions of the actual dialogue; wherein during the training mode of operation the processor performs the predict and adjust operations for each of a plurality of turns of the training dialogue in order to determine appropriate probability assignments for a given set of cues, and then uses the resulting probability assignments to predict initiative shifts when the system is acting as a participant in the actual dialogue during the dialogue mode of operation.

Claim 26

A processing system for tracking shifts in initiative during a dialogue using a set of cues, the dialogue including a sequence of turns, the system comprising: a speech recognition unit for converting an input speech signal representing at least a portion of an actual dialogue to a digital format, wherein during a dialogue mode of operation the processing system acts as a participant in the actual dialogue; a processor having an input coupled to an output of the speech recognition unit for receiving the converted speech signal therefrom, wherein during a training mode of operation the processor is operative to predict in conjunction with a current turn of a training dialogue at least one of a task initiative holder and a dialogue initiative holder for a next turn of the training dialogue, using at least one cue observed during the current turn, and a probability assignment associated with the observed cue, and to adjust the probability assignment associated with the observed cue if at least one of the predicted task and dialogue initiative holders do not agree with corresponding respective actual task and dialogue initiative holders for the next turn of the training dialogue; a memory coupled to the processor for storing the probability assignment and the task and dialogue initiative indices; and a speech synthesizer having an input coupled to an output of the processor, the speech synthesizer receiving output signals from the processor corresponding to portions of the actual dialogue; wherein during the training mode of operation the processor performs the predict and adjust operations for each of a plurality of turns of the training dialogue in order to determine appropriate probability assignments for a given set of cues, and then uses the resulting probability assignments to predict initiative shifts when the system is acting as a participant in the actual dialogue during the dialogue mode of operation.

Claim 29

The method of claim 24 further including the step of computing in conjunction with the current turn of the dialogue at least one of a task initiative index and a dialogue initiative index for the next turn of the dialogue, using (i) the at least one cue from the current turn, (ii) the probability assignment associated with the at least one cue, and (iii) a current value of at least one of the task initiative index and the dialogue initiative index, and wherein the predicting step includes predicting at least one of the task initiative holder and the dialogue initiative holder for the next turn of the dialogue, using at least one of the task initiative index and dialogue initiative index computed for the next turn.

Claim 30

The method of claim 25 wherein the current value of at least one of the task initiative index and the dialogue initiative index includes initial values of the task initiative index and the dialogue initiative index assigned prior to performing the computing and predicting steps for a first turn of the dialogue.

Claim 31

The method of claim 24 wherein the probability assignment assigns a value in the range (0,1) to each of a plurality of possible initiative holders, such that the values for all the possible initiative holders sum to 1, and the value assigned to a given initiative holder indicates the amount of support a corresponding cue provides for the given initiative holder actually holding the initiative in the dialogue.

Claim 34

The apparatus of claim 28 wherein the processor is further operative to compute at least one of a task initiative index and a dialogue initiative index for the current turn of the dialogue, using (i) the at least one cue from the current turn, (ii) the probability assignment associated with the at least one cue, and (iii) a current value of at least one of the task initiative index and the dialogue initiative index, and further wherein the processor predicts at least one of the task initiative holder and the dialogue initiative holder for the next turn of the dialogue, using at least one of the task initiative index and the dialogue initiative index computed for the current turn.

Claim 35

The apparatus of claim 29 wherein the current value of at least one of the task initiative index and the dialogue initiative index includes initial values of the task initiative index and the dialogue initiative index assigned prior to performing the compute and predict operations for a first turn of the dialogue.

Claim 36

The apparatus of claim 28 wherein the probability assignment assigns a value in the range (0,1) to each of a plurality of possible initiative holders, such that the values for all the possible initiative holders sum to 1, and the value assigned to a given initiative holder indicates the amount of support a corresponding cue provides for the given initiative holder actually holding the initiative in the dialogue.

Claim 37

A processing system for tracking shifts in initiative during a dialogue using a set of cues, the dialogue including a sequence of turns, the system comprising: a speech recognition unit for converting an input speech signal representing at least a portion of a dialogue to a digital format; a processor having an input coupled to an output of the speech recognition unit for receiving the converted speech signal therefrom, wherein the processor is operative to predict in conjunction with a current turn of the dialogue at least one of a task initiative holder and a dialogue initiative holder for a next turn of the dialogue, using at least one cue observed during the current turn, and a probability assignment associated with the at least one cue; a memory coupled to the processor for storing the probability assignment; and a speech synthesizer having an input coupled to an output of the processor, the speech synthesizer receiving output signals from the processor corresponding to portions of the dialogue, such that the processing system acts as a participant in the dialogue.

Claim 38

A processing system for tracking shifts in initiative during a dialogue using a set of cues, the dialogue including a sequence of turns, the system comprising: a speech recognition unit for converting an input speech signal representing at least a portion of a dialogue to a digital format; a processor having an input coupled to an output of the speech recognition unit for receiving the converted speech signal therefrom, wherein the processor is operative to predict in conjunction with a current turn of the dialogue at least one of a task initiative holder and a dialogue initiative holder for a next turn of the dialogue, using at least one cue observed during the current turn, and a probability assignment associated with the at least one cue; a memory coupled to the processor for storing the probability assignment; and a speech synthesizer having an input coupled to an output of the processor, the speech synthesizer receiving output signals from the processor corresponding to portions of the dialogue, such that the processing system acts as a participant in the dialogue.

Claim 41

An article of manufacture for use in a processing system for tracking shifts in initiative during a dialogue, the dialogue including a sequence of turns, comprising a machine readable medium containing one or more programs which when executed implement the steps of: predicting in conjunction with a current turn of the dialogue, in response to receipt of an input signal representing one of speech and text associated with a portion of the dialogue, at least one of a task initiative holder and a dialogue initiative holder for a next turn of the dialogue, using at least one cue observed during the current turn, and a probability assignment associated with the observed cue; and adjusting the probability assignment associated with the observed cue if at least one of the predicted task initiative holder and predicted dialogue initiative holder do not agree with corresponding respective actual task and dialogue initiative holders for the next turn of the dialogue.

Claim 42

An article of manufacture for use in a processing system for tracking shifts in initiative during a dialogue, the dialogue including a sequence of turns, comprising a machine readable medium containing one or more programs which when executed implement the steps of: predicting in conjunction with a current turn of the dialogue, in response to receipt of an input signal representing one of speech and text associated with a portion of the dialogue, at least one of a task initiative holder and a dialogue initiative holder for a next turn of the dialogue, using at least one cue observed during the current turn, and a probability assignment associated with the observed cue; and adjusting the probability assignment associated with the observed cue if at least one of the predicted task initiative holder and predicted dialogue initiative holder do not agree with corresponding respective actual task and dialogue initiative holders for the next turn of the dialogue.

Claim 45

An article of manufacture for use in a processing system for tracking shifts in initiative during a dialogue, the dialogue including a sequence of turns, comprising a machine readable medium containing one or more programs which when executed implement the steps of: identifying the presence of at least one cue in a current turn of the dialogue, in response to receipt of an input signal representing one of speech and text associated with a portion of the dialogue; and predicting in conjunction with the current turn of the dialogue at least one of a task initiative holder and a dialogue initiative holder for a next turn of the dialogue, using the at least one cue and a probability assignment associated with that cue.

Claim 46

An article of manufacture for use in a processing system for tracking shifts in initiative during a dialogue, the dialogue including a sequence of turns, comprising a machine readable medium containing one or more programs which when executed implement the steps of: identifying the presence of at least one cue in a current turn of the dialogue, in response to receipt of an input signal representing one of speech and text associated with a portion of the dialogue; and predicting in conjunction with the current turn of the dialogue at least one of a task initiative holder and a dialogue initiative holder for a next turn of the dialogue, using the at least one cue and a probability assignment associated with that cue.