# US 2024/0202455 A1 · Identifying multimedia asset similarity using blended semantic and latent feature analysis

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

- Number: US 2024/0202455 A1
- Document ID: US20240202455A1
- Status: Published application
- Kind: A1
- Application number: 18/592,140
- Filed: 2024-02-29
- Published: 2024-06-20
- Earliest claimed: 2012-02-17
- Assignees: TiVo Solutions Inc.
- Applicants: TiVo Solutions Inc.
- Inventors: David Arthur; Doug Mittendorf
- Web page: https://patentgrep.com/patent/US20240202455A1

## CPC

- G06F 40/30: Semantic analysis
- G06F 16/41: Indexing; Data structures therefor; Storage structures
- G06F 16/43: Querying
- G06F 16/489: {using time information}
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Methods and system for determining a similarity relationship between a plurality of digital assets and a target digital asset comprises creating a normalized semantic feature vector associated with a search query, discovering the target asset based on the normalized semantic feature vector, generating a normalized latent feature vector associated with the target asset, comparing the normalized semantic feature vector with semantic feature vectors for each of the digital assets to generate a semantic comparison value, comparing the normalized target latent feature vector with latent feature vectors for each of the digital assets to generate a latent comparison value, blending the semantic comparison vector value with the latent feature comparison vector value to create a target comparison value for each of the digital assets, and reporting the digital assets having the highest target comparison values to the user or group of users.

## Claims

### Claim at position 1

1-20. (canceled)

### Claim 21

A method comprising:
receiving a query for content, the query associated with a user profile;
identifying a plurality of content items with respective metadata corresponding to the query;
for each of the plurality of content items, computing a semantic score by comparing the respective metadata and the query;
associating, in a matrix, the plurality of content items with a plurality of user profiles based on user viewing activity data for the plurality of user profiles;
factorizing the matrix to generate two matrices, wherein a first of the two matrices represents latent features for the plurality of user profiles and a second of the two matrices represents latent features for the plurality of content items;
calculating a target latent feature vector by performing a dot product of the two matrices;
computing a latent score by performing a similarity comparison of the target latent feature vector and the second of the two matrices;
determining a first factor proportional to an amount of user viewing activity data for the user profile;
for the plurality of content items, computing respective combined scores by combining the latent score weighted by the first factor and a respective semantic score weighted by a second factor that is a function of the first factor, wherein as one of the first factor or the second factor increases, the other of the first factor or the second factor decreases; and
causing to be displayed the plurality of content items ordered based on the respective combined scores.

### Claim 22 (depends on 21)

The method of claim 21, further comprising:
in response to determining that the first factor meets or exceeds a pre-determined threshold, determining a similarity confidence level.

### Claim 23 (depends on 21)

The method of claim 21, wherein the user viewing activity data for the plurality of user profiles comprises one or more of user ratings, explicit user feedback, implicit user feedback, user recommendations, user interactions, or user reviews.

### Claim 24 (depends on 21)

The method of claim 21, wherein comparing the respective metadata and the query comprises:
identifying one or more keywords of the query; and
comparing each of the one or more keywords to one or more terms of the respective metadata.

### Claim 25 (depends on 24)

The method of claim 24, further comprising generating at least one of a searchable index or a searchable inverted index for the one or more keywords of the query.

### Claim 26 (depends on 21)

The method of claim 21, wherein the target latent feature vector comprises target features corresponding to one or more of title, creation date, director, producer, writer, production studio, actors, characters, dialog, subject matter, genre, objects, settings, locations, themes, or legal clearance to third party copyrighted material associated with the query.

### Claim 27 (depends on 21)

The method of claim 21, wherein computing the latent score comprises computing a cosine similarity of the target latent feature vector and the second of the two matrices.

### Claim 28 (depends on 21)

The method of claim 21, wherein causing to be displayed the plurality of content items ordered based on the respective combined scores comprises causing to be displayed content items of the plurality of content items having highest combined scores of the respective combined scores.

### Claim 29 (depends on 21)

The method of claim 21, wherein factorizing the matrix to generate the two matrices comprises factorizing the matrix using a collaborative filtering model.

### Claim 30 (depends on 29)

The method of claim 29, wherein the collaborative filtering model comprises one or more of a Bayesian network model, a clustering model, a latent semantic model, a Probabilistic Latent Semantic Analysis model, a Latent Dirichlet Allocation model, or a Markov Decision Process model.

### Claim 31

A system comprising:
one or more communication paths configured to receive a query; and
control circuitry configured to:
receive, via the one or more communication paths, a query for content, the query associated with a user profile;
identify a plurality of content items with respective metadata corresponding to the query;
for each of the plurality of content items, compute a semantic score by comparing the respective metadata and the query;
associate, in a matrix, the plurality of content items with a plurality of user profiles based on user viewing activity data for the plurality of user profiles;
factorize the matrix to generate two matrices, wherein a first of the two matrices represents latent features for the plurality of user profiles and a second of the two matrices represents latent features for the plurality of content items;
calculate a target latent feature vector by performing a dot product of the two matrices;
compute a latent score by performing a similarity comparison of the target latent feature vector and the second of the two matrices;
determine a first factor proportional to an amount of user viewing activity data for the user profile;
for the plurality of content items, compute respective combined scores by combining the latent score weighted by the first factor and a respective semantic score weighted by a second factor that is a function of the first factor, wherein as one of the first factor or the second factor increases, the other of the first factor or the second factor decreases; and
cause to be displayed the plurality of content items ordered based on the respective combined scores.

### Claim 32 (depends on 31)

The system of claim 31, wherein the control circuitry is further configured to:
in response to determining that the first factor meets or exceeds a pre-determined threshold, determine a similarity confidence level.

### Claim 33 (depends on 31)

The system of claim 31, wherein the user viewing activity data for the plurality of user profiles comprises one or more of user ratings, explicit user feedback, implicit user feedback, user recommendations, user interactions, or user reviews.

### Claim 34 (depends on 31)

The system of claim 31, wherein the control circuitry, when comparing the respective metadata and the query, is configured to:
identify one or more keywords of the query; and
compare each of the one or more keywords to one or more terms of the respective metadata.

### Claim 35 (depends on 34)

The system of claim 34, wherein the control circuitry is further configured to generate at least one of a searchable index or a searchable inverted index for the one or more keywords of the query.

### Claim 36 (depends on 31)

The system of claim 31, wherein the target latent feature vector comprises target features corresponding to one or more of title, creation date, director, producer, writer, production studio, actors, characters, dialog, subject matter, genre, objects, settings, locations, themes, or legal clearance to third party copyrighted material associated with the query.

### Claim 37 (depends on 31)

The system of claim 31, wherein the control circuitry, when computing the latent score, is configured to compute a cosine similarity of the target latent feature vector and the second of the two matrices.

### Claim 38 (depends on 31)

The system of claim 31, wherein the control circuitry, when causing to be displayed the plurality of content items ordered based on the respective combined scores, is configured to cause to be displayed content items of the plurality of content items having highest combined scores of the respective combined scores.

### Claim 39 (depends on 31)

The system of claim 31, wherein the control circuitry, when factorizing the matrix to generate the two matrices, is configured to factorize the matrix using a collaborative filtering model.

### Claim 40 (depends on 39)

The system of claim 39, wherein the collaborative filtering model comprises one or more of a Bayesian network model, a clustering model, a latent semantic model, a Probabilistic Latent Semantic Analysis model, a Latent Dirichlet Allocation model, or a Markov Decision Process model.

### Claim 41

A non-transitory computer-readable medium comprising instructions thereon that, when executed, perform a method comprising:
receiving a query for content, the query associated with a user profile;
identifying a plurality of content items with respective metadata corresponding to the query;
for each of the plurality of content items, computing a semantic score by comparing the respective metadata and the query;
associating, in a matrix, the plurality of content items with a plurality of user profiles based on user viewing activity data for the plurality of user profiles;
factorizing the matrix to generate two matrices, wherein a first of the two matrices represents latent features for the plurality of user profiles and a second of the two matrices represents latent features for the plurality of content items;
calculating a target latent feature vector by performing a dot product of the two matrices;
computing a latent score by performing a similarity comparison of the target latent feature vector and the second of the two matrices;
determining a first factor proportional to an amount of user viewing activity data for the user profile;
for the plurality of content items, computing respective combined scores by combining the latent score weighted by the first factor and a respective semantic score weighted by a second factor that is a function of the first factor, wherein as one of the first factor or the second factor increases, the other of the first factor or the second factor decreases; and
causing to be displayed the plurality of content items ordered based on the respective combined scores.
