US 7,752,152 B2 · Using predictive user models for language modeling on a personal device with user behavior models based on statistical modeling

Abstract

A system and method for prediction of a user goal for command/control of a personal device (e.g., mobile phone) is provided. The system employs statistical model(s) that can predict a command based, at least in part, on past user behavior (e.g., probability distribution over a set of predicates, and, optionally arguments). Further, the system can be employed with a speech recognition component to facilitate language modeling for predicting the user goal. The system can include predictive user models (e.g., predicate model and argument model) that receive a user input (e.g., utterance) and employ statistical modeling to determine the likely command without regard to the actual content of the input (e.g., utterance). The system employs features for predicting the next user goal which can be stored in a user data store. Features can capture personal idiosyncrasies or systematic patterns of usage (e.g., device-related, time-related, predicate-related, contact-specific and/or periodic features).

Claims20

Dependent claims

A system for the prediction of a user goal for command/control of a personal device, comprising:

a user data store that stores information regarding a particular user's behavior with respect to the personal device;
a predictive user model that receives a user input and employs statistical modeling of the information stored in the user data store to predict a command/control, based at least in part on a past pattern of behavior by the user, to be automatically performed by the personal device;
and
an execution component that automatically performs the command/control determined according to the predictive user model,
wherein at least one of a parameter and a structure of the predictive model is updated in an online manner, and
wherein the online update is performed according to a learning rate (β), wherein the learning rate is:
β
=
ESS
N
+
ESS
,
wherein N is the total number of individual observations and ESS is an equivalent sample size of training data.

11 more elements

A method of controlling a personal communication device, comprising:

inputting a spoken command to the personal communication device;
providing a predictive user model to determine, from the spoken command, a likely contact to be called by the personal communication device;
providing, as a first component of the predictive user model, a predicate model having a plurality of predicate functions supported by an application of the personal communication device,
providing, as a second component of the predictive user model, an argument model, the argument model being a function of the predicate model and employing statistical modeling and information stored in a personalized user data store to determine a likely contact corresponding to a predicate function, wherein the argument model represents for a contact c:
p(BC=c=true|f(c)), wherein p(BC=c=true|f(c)) is normalized for all c via:
p(C=c(g)|F=f)=
p
⁡
(
B
C
-
c
⁡
(
g
)
=
true
❘
f
⁡
(
c
⁡
(
g
)
)
)
∑
c
′
⁢
⁢
p
⁡
(
B
C
-
c
′
=
true
❘
⁢
f
⁡
(
c
′
)
)
,
in which F is a set of variables that are observed when G is predicted, G is a random variable corresponding to a next goal, g is a goal, C is Aπ(g), Aπ is a random variable whose values are possible argument values for predicate π, BC=c is a binary variable corresponding to C=c, and p is a probability;
providing a speech recognition component;
modifying an output of the speech recognition component by an output of the predictive user model to form a modified output;
providing a language model that determines a contact based on the modified output;
and
calling the contact,
wherein the information stored in the personalized user data store comprises an identification of:
what applications are currently running on the personal communication device and a number of calls, emails and text messages that have been received from a particular contact, wherein a contact is associated with a set of contact-specific features which are used to predict binary variables, wherein the determination of the likely contact is based at least partially on a binary variable being defined as true.

60 more elements

One or more processor-accessible storage media comprising processor-executable instructions of controlling a personal communication device that include:

inputting a spoken command to the personal communication device;
providing a predictive user model to determine, from the spoken command, a likely contact to be called by the personal communication device;
providing, as a first component of the predictive user model, a predicate model having a plurality of predicate functions supported by an application of the personal communication device,
providing, as a second component of the predictive user model, an argument model, the argument model being a function of the predicate model and employing statistical modeling and information stored in a personalized user data store to determine a likely contact corresponding to a predicate function, wherein the argument model represents for a contact c:
p(BC=c=true|f(c)), wherein p(BC=c=true|f(c)) is normalized for all c via:
p(C=c(g)|F=f)=
p
⁡
(
B
C
=
c
⁡
(
g
)
=
true
❘
f
⁡
(
c
⁡
(
g
)
)
)
∑
c
′
⁢
⁢
p
⁡
(
B
C
=
c
′
=
true
❘
f
⁡
(
c
′
)
)
,
in which F is a set of variables that are observed when G is predicted, G is a random variable corresponding to a next goal, g is a goal, C is Aπ(g), Aπ is a random variable whose values are possible argument values for predicate π, BC=c is a binary variable corresponding to C=c, and p is a probability;
providing a speech recognition component;
modifying an output of the speech recognition component by an output of the predictive user model to form a modified output;
providing a language model that determines a contact based on the modified output;
and
calling the contact,
wherein the information stored in the personalized user data store comprises an identification of:
what applications are currently running on the personal communication device and a number of calls, emails and text messages that have been received from a particular contact, wherein a contact is associated with a set of contact-specific features which are used to predict binary variables, wherein the determination of the likely contact is based at least partially on a binary variable being defined as true.

59 more elements

Dependencies are inferred from the claim text.

Using predictive user models for language modeling on a personal device with user behavior models based on statistical modeling

Life of this patent

Using predictive user models for language modeling on a personal device with user behavior models based on statistical modeling

Life of this patent

Abstract

Claims20