Systems and methods for result arbitration in spoken dialog systems
A method for arbitrating spoken dialog results includes receiving a spoken utterance from a user within an environment; receiving first recognition results and a first confidence level associated with the spoken utterance from a first source; receiving second recognition results and a second confidence level associated with the spoken utterance from a second source; receiving human-machine-interface (HMI) information associated with the user; selecting between the first recognition results and the second recognition results based on at least one of the first confidence level, the second confidence level, and the HMI information.
1. A system for arbitrating spoken dialog results in a vehicle, the system comprising:
an event timer module, including a processor, configured to:
receive, from a first source, first recognition results associated with a spoken utterance;
receive, from a second source, second recognition results associated with the spoken utterance,
determine a first time associated with when the first recognition results were received; and
determine a second time associated with when the second recognition results were received;
a semantic interpretation module, including a processor, configured to receive the first recognition results and determine a first confidence level associated with the first recognition results, to receive the second recognition results and determine a second confidence level associated with the second recognition results;
a short cut module, including a processor, configured to determine whether the first confidence level is above a threshold and to provide the first recognition results to an output module if the first confidence level is above the threshold;
a result selection module, including a processor, configured to select and provide to the output module the first recognition results when the first confidence level is not above the threshold, and the first confidence level is greater than the second confidence level wherein the result selection module is further configured to select the first recognition results based in part on the first time and the second time.
2. The system of claim 1 , wherein the output module is configured to provide to the user the selected first recognition results and to provide a reset signal to the event timer module.
3. The system of claim 1 , wherein each of the first recognition results includes one or more linguistic slots associated with corresponding slot types.
4. The system of claim 3 , wherein the linguistic slot types include at least one of a proper name and an address.
5. The system of claim 1 , wherein selecting between the first recognition results and the second recognition results by the result selection module includes selecting based on a Hidden Markov Model.