# US 2017/0251258 A1 · Techniques for context aware video recommendation

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2017/0251258 A1
- Document ID: US20170251258A1
- Status: Published application
- Kind: A1
- Application number: 15/052,964
- Filed: 2016-02-25
- Published: 2017-08-31
- Earliest claimed: 2016-02-25
- Assignees: Adobe Systems Incorporated
- Applicants: Adobe Systems Incorporated
- Inventors: Viswanathan Swaminathan; Saayan Mitra; Gang Wu
- Web page: https://patentgrep.com/patent/US20170251258A1

## CPC

- H04N 21/4668: {for recommending content, e.g. movies}
- G06F 16/73: Querying
- H04N 21/252: {Processing of multiple end-users' preferences to derive collaborative data}
- H04N 21/25866: {Management of end-user data (customer care in data networks H04L41/5077)}
- H04N 21/2668: Creating a channel for a dedicated end-user group, e.g. insertion of targeted commercials based on end-user profiles {(information retrieval from the Internet by querying with filtering and personalisation G06F16/9535; arrangements for replacing or switching information during the broadcast H04H20/10; push services over packet-switching network H04L12/1859; adaptation of message content in packet-switching networks H04L51/063)}
- H04N 21/44218: {Detecting physical presence or behaviour of the user, e.g. using sensors to detect if the user is leaving the room or changes his face expression during a TV programme (methods or arrangements for recognising human body or animal bodies or body parts G06V40/10; methods or arrangements for acquiring or recognising human faces, facial parts, facial sketches, facial expressions G06V40/16; methods or arrangements for recognising movements or behaviour G06V40/20; arrangements for identifying users in broadcast systems H04H60/45)}
- H04N 21/44226: {on social networks}
- H04N 21/4532: {involving end-user characteristics, e.g. viewer profile, preferences (monitoring of user activities for profile generation for accessing a video database G06F16/739; user profiles in network data switching protocols H04L67/306; processing of user preferences or user profiles in wireless networks H04W8/18)}
- H04N 21/4826: {using recommendation lists, e.g. of programmes or channels sorted out according to their score}
- H04N 21/6125: {involving transmission via Internet (transmission by internet of broadcast information H04H60/82)}
- H04N 21/44204: {Monitoring of content usage, e.g. the number of times a movie has been viewed, copied or the amount which has been watched (monitoring of user activities for profile generation for accessing a video database G06F16/739; protecting generic digital content where the protection is independent of the precise nature of the content G06F21/10; arrangements for monitoring the use made of the broadcast services in broadcast systems H04H60/31)}
- H04N 21/4667: {Processing of monitored end-user data, e.g. trend analysis based on the log file of viewer selections}
- H04N: PICTORIAL COMMUNICATION, e.g. TELEVISION
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Systems and methods for identifying, in a network environment in which users watch videos that are downloaded or streamed over a network, a video in which a user is likely to be interested based on session context. For example, a server or other computing system identifies prior session contexts in which prior users watched videos and session progress data for prior sessions in which these prior users watched the videos. The server or other computing system determines a session context of a user for whom a video is to be recommended. For this user, the server or other computing system generates a recommendation identifying one or more videos in which the user is likely to be interested, where the user has not previously watched the recommended videos. The recommendation is generated based on the prior session contexts, the session progress data, and the session context of the user.

## Claims

### Claim 1

A computer-implemented method for recommending video content in a network environment in which users watch videos that are downloaded or streamed over a network, the method comprising:
identifying, by a processing device, prior session contexts in which prior users watched videos and session progress data for prior sessions in which the prior users watched the videos;
building, by the processing device, an interest-estimation model from the prior session contexts and the session progress data, wherein building the interest-estimation model comprises:
identifying a feature set comprising (i) user features comprising characteristics specific to the prior users, (ii) video features comprising characteristics specific to the videos, and (iii) session features comprising characteristics specific to the prior sessions,
determining that a first feature in the feature set is dependent on a second feature in the feature set,
reducing the feature set for use by a factorization machine model, wherein reducing the feature set comprises selecting a feature subset comprising one or more of the user features, one or more of the video features, and one or more of the session features, wherein the first feature is omitted from the feature subset based on the first feature being dependent on the second feature, and
estimating, based on a weighted combination of the feature subset provided by the factorization machine model, a set of user interests for videos associated with a set of contexts;
determining, by the processing device, a session context of a current user;
generating, by the processing device, a recommendation identifying a video in which the current user is likely to be interested and that has not been previously watched by the current user, wherein the processing device generates the recommendation based on the session context matching one of the set of contexts; and
causing, by the processing device and based on the recommendation, a content server to transmit the video to a user device associated with the current user.

### Claim 2

(canceled)

### Claim 3 (depends on 1)

The method of claim 1, wherein generating the recommendation comprises recommending a subset of videos from a set of videos to the current user, and further comprising providing, responsive to a selection input from the current user, the video from the subset of videos recommended.

### Claim 4 (depends on 1)

The method of claim 1, wherein the recommendation is also generated based on user ratings included in information describing the prior sessions.

### Claim 5 (depends on 1)

The method of claim 1, wherein the session context is a type of device or operating system that the current user is currently using.

### Claim 6 (depends on 1)

The method of claim 1, wherein the session context comprises data describing a context of the current user.

### Claim 7 (depends on 6)

The method of claim 6, wherein the data describing the context of the current user comprises at least one of:
a demographic for the current user;
a current time of day for the current user; or
a current geographic location of the current user.

### Claim 8 (depends on 1)

The method of claim 1, wherein the session context comprises data describing a video context.

### Claim 9 (depends on 1)

The method of claim 1, wherein the session context comprises data describing attributes specific to a session.

### Claim 10 (depends on 1)

The method of claim 1, wherein the session progress data is generated by the processing device based on consumption data identifying how much of the videos the prior users watched.

### Claim 11

A system comprising:
a processing device; and
a non-transitory computer readable medium communicatively coupled to the processing device,
wherein the processing device is configured for executing instructions stored on the non-transitory computer-readable medium and thereby performing operations comprising:
identifying prior session contexts in which prior users watched videos and session progress data for prior sessions in which the prior users watched the videos;
building an interest-estimation model from the prior session contexts and the session progress data, wherein building the interest-estimation model comprises:
identifying a feature set comprising (i) user features comprising characteristics specific to the prior users, (ii) video features comprising characteristics specific to the videos, and (iii) session features comprising characteristics specific to the prior sessions,
determining that a first feature in the feature set is dependent on a second feature in the feature set,
reducing the feature set for use by a factorization machine model, wherein reducing the feature set comprises selecting a feature subset comprising one or more of the user features, one or more of the video features, and one or more of the session features, wherein the first feature is omitted from the feature subset based on the first feature being dependent on the second feature, and
estimating, based on a weighted combination of the feature subset provided by the factorization machine model, a set of user interests for videos associated with a set of contexts;
determining a session context of a current user;
generating a recommendation identifying a video in which the current user is likely to be interested and that has not been previously watched by the current user, wherein the processing device generates the recommendation based on the session context matching one of the set of contexts; and
causing a content server to transmit the video to a user device associated with the current user.

### Claim 12

(canceled)

### Claim 13 (depends on 11)

The system of claim 11, wherein generating the recommendation comprises recommending a subset of videos from a set of videos to the current user, and further comprising providing, responsive to a selection input from the current user, the video from the subset of videos recommended.

### Claim 14 (depends on 12)

The system of claim 12, wherein the processing device is further configured for generating the recommendation based on user ratings included in information describing the prior sessions.

### Claim 15 (depends on 11)

The system of claim 11, wherein the session context is a type of device or operating system that the current user is currently using.

### Claim 16 (depends on 11)

The system of claim 11, wherein the session context comprises data describing at least one of:
a demographic for the current user;
a current time of day for the current user; or
a current geographic location of the current user.

### Claim 17 (depends on 11)

The system of claim 11, wherein the session context comprises data describing a video context.

### Claim 18 (depends on 11)

The system of claim 11, wherein the session context comprises data describing attributes specific to a session.

### Claim 19

A non-transitory computer readable medium storing program code executable by a processing device, wherein the program code comprises:
program code for identifying prior session contexts in which prior users watched videos and session progress data for prior sessions in which the prior users watched the videos;
program code for building an interest-estimation model from the prior session contexts and the session progress data, wherein building the interest-estimation model comprises:
identifying a feature set comprising (i) user features comprising characteristics specific to the prior users, (ii) video features comprising characteristics specific to the videos, and (iii) session features comprising characteristics specific to the prior sessions,
determining that a first feature in the feature set is dependent on a second feature in the feature set,
reducing the feature set for use by a factorization machine model, wherein reducing the feature set comprises selecting a feature subset comprising one or more of the user features, one or more of the video features, and one or more of the session features, wherein the first feature is omitted from the feature subset based on the first feature being dependent on the second feature, and
estimating, based on a weighted combination of the feature subset provided by the factorization machine model, a set of user interests for videos associated with a set of contexts;
determining a session context of a current user;
program code for generating a recommendation identifying a video in which the current user is likely to be interested and that has not been previously watched by the current user, wherein the processing device generates the recommendation based on the session context matching one of the set of contexts; and
program code for causing, based on the recommendation, a content server to transmit the video to a user device associated with the current user.

### Claim 20

(canceled)

### Claim 21 (depends on 1)

The method of claim 1, wherein the processing device selects the feature subset by applying, to the feature subset, one or more of a variance analysis process, a forwards/backwards selection process, and a lasso-based regression analysis process.

### Claim 22 (depends on 21)

The method of claim 21, wherein the one or more of the variance analysis process, the forwards/backwards selection process, and the lasso-based regression analysis process computes a correlation between first data values for the first feature and second data values for the second feature,
wherein processing device determines that the first feature is dependent on the second feature based on the computed correlation exceeding a threshold correlation.

### Claim 23 (depends on 1)

The method of claim 1, wherein the session context comprises data describing a combination of a demographic for the current user, a current time of day for the current user, and a current geographic location of the current user.
