Semantic Locations Prediction

• US 20160321551A1
• Filed: 04/28/2016
• Published: 11/03/2016
• Est. Priority Date: 04/29/2015
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Priness, Ido, Dotan-Cohen, Dikla, Somech, Haim, Cohn, Ido, Weinberg, Shira
• Granted Patent: US 10,909,464 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06N 5/047 Pattern matching networks; Rete networks
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: G06Q 10/00 Administration; Management
• CPC Class Codes: G06Q 30/0251 Targeted advertisements
• CPC Class Codes: G06Q 30/0261 based on user location
• CPC Class Codes: G06Q 30/0269 based on user profile or attribute

Abstract

Aspects of the technology described herein provide a personalized computing experience for a user based on a predicted future semantic location of the user. In particular, a likely future location (or sequences of future locations) for a user may be determined, including contextual information about the future location. Using information from the current context of the user'"'"'s current location with historical observations about the user and expected user events, out-of-routine events, or other lasting or ephemeral information, a prediction of one or more future semantic locations and corresponding confidences may be determined and used for providing personalized computing services to the user. The prediction may be provided to an application or service such as a personal assistant service associated with the user, or may be provided as an API to facilitate consumption of the prediction information by an application or service.

Claims

1. A computerized system comprising:

one or more sensors configured to provide sensor data comprising at least location information;

one or more processors; and

computer storage memory having computer-executable instructions stored thereon which, when executed by the processor, implement a method of providing a personalized computing experience to a user based on a predicted future location of the user, the method comprising;

determining, using the one or more sensors, a current context including at least a current location of the user;

determining a set of historic visits by the user to the current location;

based on the current context and set of historic visits, determining a pattern-based prediction of a future location of the user;

determining an explicit signal for the user indicating information related to the user'"'"'s location at a future time corresponding to the pattern-based prediction;

performing conflation of the explicit signal and pattern-based prediction to determine a coherent prediction of the user'"'"'s future location; and

initiating a personalized computing experience to the user based on the coherent prediction of the user'"'"'s future location.

2. The computerized system of claim 1, wherein the personalized computing experience comprises a scheduling application, route-planning application, intelligent notification application, personalized content computing service, or automation application.

3. The computerized system of claim 1, wherein the coherent prediction of the user'"'"'s future location comprises a next semantic location for the user including at least a likely arrival time, likely length of stay, likely departure time from the current location, user activity likely to be performed at the next semantic location, a venue of the next semantic location, or another person likely to be present at the next semantic location.

4. The computerized system of claim 1, wherein the pattern-based prediction is determined based on a next location visited after the current location having the highest observation count in a subset of the set of historic visits, the subset having at least location and another context in common with the current context.

5. The computerized system of claim 4, wherein the each historic visit in the subset of historic visits includes at least a behavior feature pattern or a periodic feature pattern in common with the current visit.

6. The computerized system of claim 1, wherein determining the current context comprises determining one or more features of the current visit, and wherein determining the pattern-based prediction comprises:

determining a set of candidate pattern-based predictions, each candidate pattern-based prediction determined using a subset of the set of historic visits, each subset of the set of historic visits having at least one feature in common with the current visit;

determining a corresponding prediction confidence with each candidate pattern-based prediction;

and selecting the candidate prediction having the highest prediction confidence as the determined pattern-based prediction.

7. The computerized system of claim 6, wherein the prediction confidence is determined as a product of a prediction probability and a prediction significance corresponding to the candidate prediction

8. The computerized system of claim 6, wherein determining the set of candidate pattern-based predictions comprises, for each candidate prediction:

performing visit filtering to determine the subset of historic visits;

determining a similarity score for each historic visit in the subset with respect to the current visit;

determining from the subset of historic visits, an example set of historic visits based on a comparison of the similarity score to a similarity threshold, the example set comprising those historic visits having a similarity score that satisfies the similarity threshold; and

determining the candidate pattern-based prediction as the next location that occurs the most often, after the current location, in the example set.

9. The computerized system of claim 8, wherein the similarity score for each historic visit is based on the number of features in common with the current visit, and wherein the similarity threshold is predetermined or based on the number of historic visits in the subset.

10. The computerized system of claim 8, wherein determining a similarity score for each historic visit comprises determining a visitation sequence similarity between the historic visit and the current visit using a Levenschtein distance.

11. The computerized system of claim 1, wherein the explicit signal comprises information associated with a future location of the user, and wherein the explicit signal includes information indicating a flight, scheduled event, out-of-routine event, or ephemeral information.

12. The computerized system of claim 1, wherein performing conflation of the explicit signal and pattern-based prediction comprises:

determining an explicit signal confidence associated with the explicit signal;

based at least on the explicit signal confidence, determining a level of conflict between the explicit signal and the pattern-based prediction; and

based on the level of conflict;

(1) overriding the pattern-based prediction with location information derived from the explicit signal to determine the coherent prediction;

(2) modifying the pattern-based prediction according to the explicit signal to determine the coherent prediction;

or

(3) determining that the explicit signal will not impact the pattern-based prediction, and providing the pattern-based prediction as the coherent prediction.

13. A computing device comprising a computer memory and a computer processor that is configured to allow a virtual assistant to determine and utilize a prediction of a next location for a user to provide a personalized experience, the computing device comprising:

the virtual assistant that is a computer program stored on the computer memory having computer instructions that when executed by the computer processor cause the virtual assistant to;

determine a current visit for the user to a current location;

determine a current context of the current visit;

determine a set of historic visits by the user to the current location, each historic visit including historic information about a subsequent location visited by the user after the current location;

based on the current context and set of historic visits, determine a history-based prediction of a future semantic location of the user;

determine an explicit signal for the user indicating an alternative future location of the user at a time corresponding to the history-based prediction of a future semantic location;

conflate the explicit signal and history-based prediction to determine the next location for the user; and

provide a personalized computing experience to the user based on the next location.

14. The computing device of claim 13, wherein the next location comprises a next semantic location for the user including at least an expected arrival time at the next location, an expected length of stay at the next location, an expected departure time from the current location, user activity likely to be performed at the next location, a venue of the next location, or another person likely to be present at the next location.

15. The computing device of claim 13, wherein determining the current context comprises determining a plurality of features of the current visit, and wherein the computer instructions, when executed by the computer processor, further cause the virtual assistant to determine a subset of historic visits having the plurality of features of the current visit, and wherein the history-based prediction is determined as the subsequent location visited having the highest observation count in the subset.

16. The computing device of claim 15, wherein the plurality of features includes a behavior pattern feature or a periodic feature.

17. The computing device of claim 13, wherein the personalized computing experience comprises a scheduling service, route-planning to the next location, providing a notification associated with the next location, providing content based on the next location, or providing an automation service.

18. A computerized method for providing a personalized computing experience to a user based on a predicted next semantic location of the user, the method comprising:

determining a current context for a user including a current location of the user, the current location determined using a location sensor;

determining a set of historic visits by the user to the current location, each historic visit including historic information about a subsequent location visited by the user after the current location;

based on the current context and set of historic visits, determining a pattern-based prediction of a future location of the user, the pattern based prediction including a future context related to the future location;

determining an explicit signal for the user indicating an alternative future location of the user at a time corresponding to the pattern-based prediction;

performing conflation of the explicit signal and pattern-based prediction to determine the next semantic location for the user; and

providing a personalized computing experience to the user based on the coherent prediction of the user'"'"'s future location.

initiating a personalized computing experience to the user based on next semantic location,

wherein the pattern-based prediction is determined based on a subset of the set of historic visits having at least location and a context feature in common with the current context, the pattern-based prediction determined as the subsequent location that occurs most often in the subset, and

wherein the future context comprises an expected arrival time at the next semantic location, an expected length of stay at the next semantic location, an expected departure time from the current location, user activity likely to be performed at the next semantic location, a venue of the next semantic location, or another person likely to be present at the next semantic location.

19. The computerized method of claim 18, wherein determining the current context comprises determining one or more context features of the current visit, and wherein determining the pattern-based prediction comprises:

determining a set of candidate pattern-based predictions, each candidate pattern-based prediction determined using a subset of the set of historic visits, each subset of the set of historic visits having at least one context feature in common with the current visit;

determining a corresponding prediction confidence with each candidate pattern-based prediction;

and selecting the candidate prediction having the highest prediction confidence as the determined pattern-based prediction.

20. The computerized method of claim 19, wherein determining the set of candidate pattern-based predictions comprises, for each candidate prediction:

performing visit filtering to determine the subset of historic visits;

determining a similarity score for each historic visit in the subset with respect to the current visit, based on a visitation sequence similarity between the historic visit and the current visit;

determining from the subset of historic visits, an example set of historic visits based on a comparison of the similarity score to a similarity threshold, the example set comprising those historic visits having a similarity score that satisfies the similarity threshold; and

determining the candidate pattern-based prediction as the next location that occurs the most often, after the current location, in the example set, wherein the similarity threshold is predetermined or based on the number of historic visits in the subset.