Predictive content retrieval using device movements

• US 9,959,506 B1
• Filed: 06/17/2014
• Issued: 05/01/2018
• Est. Priority Date: 06/17/2014
• Status: Expired due to Fees

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Karppanen, Jari Juhani
• Primary Examiner(s): Smith, Paulinho E
• Application Number:
  US14/306,776
• Time in Patent Office: 1,414 Days
• Field of Search: None
• CPC Class Codes: G06F 16/9574 of access to content, e.g. by caching
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

Features are disclosed for predicting or otherwise determining when a user will initiate an operation on a user computing device, such as requesting network-accessible content. Upon making the determination, the user computing device can proactively perform the determined operation or portions thereof. The user computing device may use a detection model or profile that associates user-initiated operations with data from sensors on the user computing device. The sensors may include movement sensors, environmental sensors, and the like. One benefit, among others, is that user-perceived performance can be improved because some or all of a user-initiated operation has been performed prior to user-initiation of the operation.

Claims

1. A system comprising:

a computing device comprising a computer readable memory and one or more hardware processors, wherein the computing device is programmed by executable instructions in the computer readable memory to at least;

transmit, to a server system that acts as an intermediary between user computing devices and content servers, sensor data indicative of one or more movements of the computing device and request data indicative of one or more content page requests initiated on the computing device;

receive, from the server system, a detection model that associates content page requests with computing device movements, the detection model based at least partly on the sensor data and the request data;

obtain further sensor data from one or more sensors of the computing device, the further sensor data indicative of one or more subsequent movements of the computing device;

determine, using the further sensor data and the detection model, a confidence score regarding a prediction that a user will use the computing device to request a content page during a period of time subsequent to the one or more subsequent movements;

compare the confidence score to one or more thresholds of a multi-tier confidence threshold hierarchy;

preemptively initiate, based at least partly on comparing the confidence score to the one or more thresholds, an operation associated with retrieval of the content page from a content server via the server system; and

subsequent to initiating the operation;

receive input indicating the user has requested the content page during the period of time; and

display the content page on a display of the computing device.

2. The system of claim 1, wherein the detection model further associates the request for the content page with data from one or more environmental sensors of the computing device.

3. The system of claim 1, wherein the detection model is further based at least partly on sensor data and request data provided to the server system by a plurality of separate computing devices.

4. The system of claim 1, wherein the computing device is further programmed to at least:

transmit the further sensor data and further request data indicative of the request for the content page to the server system; and

receive a new version of the detection model from the server system.

5. A computer-implemented method comprising:

as implemented by one or more processors configured to execute specific instructions,

receiving sensor data reflective of device movements of a user computing device;

receiving content requests generated by the user computing device for network-accessible content items;

determining, based at least partly on a detection model and further sensor data, a score indicative of how likely a request for a particular network-accessible content item is to occur, wherein the detection model associates a particular device movement pattern with the particular network-accessible content item;

analyzing the score with respect to one or more thresholds of a multi-tier threshold hierarchy; and

determining, based at least partly on results of analyzing the score with respect to the one or more thresholds, to proactively perform a content item request operation.

6. The computer-implemented method of claim 5, further comprising associating the particular device movement pattern with the particular network-accessible content item based at least partly on data regarding the device movements and the content requests.

7. The computer-implemented method of claim 5, further comprising initiating the content item request operation, wherein the content item request operation comprises prefetching the particular network-accessible content item.

8. The computer-implemented method of claim 5, further comprising associating the particular device movement pattern with the particular network-accessible content item based at least partly on accelerometer data generated by an accelerometer of the user computing device.

9. The computer-implemented method of claim 5, further comprising prefetching the particular network-accessible content item in response to detecting that the particular device movement pattern has occurred during a particular time of day.

10. The computer-implemented method of claim 5, further comprising associating the particular device movement pattern with the particular network-accessible content item based at least partly on environmental information.

11. The computer-implemented method of claim 10, further comprising obtaining the environmental information at least in part from at least one of:

a light sensor, a microphone, a camera, or a thermometer.

12. The computer-implemented method of claim 5, wherein the analyzing the score with respect to the one or more thresholds of the multi-tier threshold hierarchy comprises:

determining that the score satisfies a first threshold associated with prefetching the particular network-accessible content item to a server remote from the user computing device; and

determining that the score fails to satisfy a second threshold associated with prefetching the particular network-accessible content item to the user computing device;

wherein the determining to proactively perform the content request operation comprises determining to proactively prefetch the particular network-accessible content item to the server based at least partly on the score satisfying the first threshold.

13. The computer-implemented method of claim 5, wherein the analyzing the score with respect to the one or more thresholds of the multi-tier threshold hierarchy comprises determining that the score satisfies a threshold associated with pre-rendering the particular network-accessible content item, and wherein the determining to proactively perform the content request operation comprises determining to pre-render the particular network-accessible content item based at least partly on the score satisfying the threshold.

14. Non-transitory computer storage having stored thereon:

a detection component configured to cause one or more processors to execute a process comprising;

using a detection profile to determine a score indicative of how likely a user-initiated operation will follow one or more physical movements of a user computing device, wherein the detection profile associates user-initiated operations on the user computing device with data from one or more sensors of the user computing device;

analyzing the score with respect to one or more thresholds of a multi-tier threshold hierarchy; and

determining, based at least partly on results of analyzing the score with respect to the one or more thresholds, an operation to perform, wherein the operation to perform is one of a plurality of preemptive operations associated with the user-initiated operation; and

a detection profile generator configured to cause one or more processors to execute a process comprising using aggregated data from the one or more sensors and aggregated usage data regarding user-initiated operations on the user computing device to generate the detection profile.

15. The non-transitory computer storage of claim 14, wherein the data from one or more sensors comprises data from at least one of:

an accelerometer or a gyroscope.

16. The non-transitory computer storage of claim 14, wherein generating the detection profile comprises generating one of a classifier or a regression model.

17. The non-transitory computer storage of claim 14, wherein the detection component is configured to transmit, to a server system, data from the one or more sensors and data reflecting user-initiated operations on the user computing device.

18. The non-transitory computer storage of claim 14, wherein the detection profile generator is configured to generate the detection profile based at least partly on data from sensors of a plurality of user computing devices.

19. The non-transitory computer storage of claim 14, wherein the process further comprises using the detection profile to predict the user-initiated operation based at least partly on environmental information from one or more environmental sensors of the user computing device.

20. The non-transitory computer storage of claim 19, wherein the one or more environmental sensors comprise at least one of:

a light sensor, a microphone, a camera, or a thermometer.

21. The non-transitory computer storage of claim 14, the process executed by the detection component further comprising preemptively executing the operation in response to determining the operation to perform.

22. The non-transitory computer storage of claim 21, wherein preemptively executing the operation comprises at least one of:

prefetching a content page predicted to be requested by a user;

uncompressing at least a portion of a content cache comprising content associated with a content page predicted to be requested by the user;

causing a separate computing device to prefetch a content page predicted to be requested by the user;

launching an application predicted to be launched by the user;

or modifying a quality of service parameter associated with content predicted to be requested by the user.