Parameterized statistical interaction policies
View Patent ↗A method and apparatus are disclosed for selecting interaction policies. Values may be provided for a group of parameters for user models. Interaction policies within a specific tolerance of an optimal interaction policy for the user models may be learned. Up to a predetermined number of the learned interaction policies, within a specific tolerance of an optimal policy for the user models, may be selected for a wireless communication device. The wireless communication device, including the selected interaction policies, may determine whether any of a group of parameters, representing a user preference or contextual information with respect to use of the wireless communication device, is updated. When any of the group of parameters has been updated, the wireless communication device may select one of the selected interaction policies, such that the selected one of the selected interaction policies may determine a better interaction behavior for the wireless communication device.
1. A machine-implemented method for selecting up to a predetermined number of a plurality of interaction policies for use on a wireless communication device, the method comprising:
providing sets of values for a plurality of parameters of a plurality of user models, the plurality of user models corresponding to expected users of the wireless communication device, each of the plurality of parameters representing at least one of a user preference or contextual information with respect to use of the processing device;
learning which ones of the plurality of interaction policies are within a specific tolerance of an optimal interaction policy for the plurality of user models, wherein the learning of which of the ones of the plurality of interaction policies are within the specific tolerance of the optimal interaction policy for the plurality of user models further comprises:
calculating an average score of each of the plurality of interaction policies with respect to each of the plurality of user models; and
determining which ones of the plurality of interaction policies have a respective average score within the specific tolerance of an optimal interaction policy score with respect to each of the plurality of user models;
selecting, for the wireless communication device, up to the predetermined number of the learned ones of the plurality of interaction policies that are within the specific tolerance of the optimal interaction policy for the plurality of user models; and
simulating, at least a predetermined number of times, user actions for each of the plurality of user models and each of the plurality of interaction policies, wherein the calculating of the score of each of the plurality of interaction policies with respect to each of the plurality of user models is based on the simulating.
2. The machine-implemented method of claim 1 further comprising:
selecting, for the wireless communication device, up to the predetermined number of the learned ones of the plurality of interaction policies that are within the specific tolerance of the optimal interaction policy for the plurality of user models and that are expected to reach a goal of the user in a minimum number of dialog turns.
3. The machine-implemented method of claim 1 wherein the sets of values of the plurality of parameters for the plurality of the plurality of user models are based, at least partly, on market studies.
4. The machine-implemented method of claim 1 wherein:
the plurality of user models is implemented via an interaction model; and
each of the set of values for the plurality of parameters define a user model of the plurality of user models when used with the interaction model.