# US 11,909,783 B2 · Providing trust and safety functionality during virtual meetings

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

- Number: US 11,909,783 B2
- Document ID: US11909783B2
- Status: Granted
- Kind: B2
- Application number: 17/733,957
- Filed: 2022-04-29
- Published: 2023-11-02
- Granted: 2024-02-20
- Earliest claimed: 2022-04-29
- Assignees: Zoom Video Communications, Inc.
- Applicants: Zoom Video Communications, Inc.
- Inventors: Vanessa Badillo DiPaulo; Josh Patrick Parecki; Dun-Yi Wang
- Web page: https://patentgrep.com/patent/US11909783B2

## CPC

- H04L 65/1089: by adding media; by removing media
- G06V 20/40: in video content (extracting overlay text G06V20/62; video retrieval G06F16/70; processing of video elementary streams in video servers H04N21/234; processing of video elementary streams in video clients H04N21/44)
- G10L 15/08: Speech classification or search
- H04L 65/403: Arrangements for multi-party communication, e.g. for conferences (data switching systems for conference H04L12/18; arrangements for connecting several subscribers to a common circuit, i.e. affording conference facilities H04M3/56; television conferencing systems H04N7/15)
- H04L 65/765: {intermediate}
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING

## Abstract

Techniques for providing trust and safety functionality during virtual meetings are provided. In an example, a method includes establishing a video conference between client devices. The method includes receiving, from a first client device of the client devices, a stream corresponding to the video conference and including one or more items of content. The method further involves applying, to the stream, a machine learning model that is trained to identify inappropriate or harmful content in the stream. The method further includes determining, via the machine learning model, that an item of content in the stream is inappropriate or harmful. The method further includes editing the stream to remove the item of content. The method further includes transmitting the edited stream to the client devices.

## Claims

### Claim 1

A system for real-time and near-real time content moderation, the system comprising:
a non-transitory computer-readable medium;
a communications interface; and
a processor communicatively coupled to the non-transitory computer-readable medium and the communications interface, the processor configured to execute processor-executable instructions stored in the non-transitory computer-readable medium to:
establish a video conference between a plurality of client devices;
receive, from a first client device of the plurality of client devices, a stream corresponding to the video conference and comprising one or more items of content;
configure content moderation settings in response to a previous activity of a user associated with the first client device;
provide the content moderation settings to a machine learning model that is trained to identify inappropriate or harmful content in the stream;
apply, to the stream, the machine learning model;
determine, via the machine learning model, that an item of content in the stream is inappropriate or harmful;
edit the stream to remove the item of content; and
transmit, to the plurality of client devices, the edited stream.

### Claim 2 (depends on 1)

The system of claim 1, wherein the processor is configured to execute further processor-executable instructions stored in the non-transitory computer-readable medium to:
transmit, via the communications interface and to a moderator client device, the item of content; and
receive, from the moderator client device, a request to edit the item of content in the stream,
wherein the editing is responsive to receiving the request to edit the item of content.

### Claim 3 (depends on 1)

The system of claim 1, wherein the processor is configured to execute further processor-executable instructions stored in the non-transitory computer-readable medium to:
receive, from a moderator client device, a set of additional content moderation settings,
wherein applying the machine learning model to the stream further comprises applying the additional content moderation settings to the machine learning model and wherein the machine learning model identifies the item of content at least in part based on the additional content moderation settings.

### Claim 4 (depends on 1)

The system of claim 1, wherein the one or more items of content include audio and wherein the processor is configured to execute further processor-executable instructions stored in the non-transitory computer-readable medium to:
perform speech recognition on the audio to identify one or more recognized words; and
identify, based on the one or more recognized words, one or more keywords corresponding to inappropriate or harmful content.

### Claim 5 (depends on 1)

The system of claim 1, wherein the one or more items of content include text and wherein the machine learning model is trained to:
determine that one or more keywords identified as inappropriate or harmful are present in the text; and
in response to the determining, identify the content as inappropriate or harmful.

### Claim 6 (depends on 1)

The system of claim 1, wherein the processor is configured to execute further processor-executable instructions stored in the non-transitory computer-readable medium to:
in response to the determination of the item of content in the stream being inappropriate, perform one or more of (i) terminating services to a user associated with the first client device, (ii) muting the user, and (iii) turning off video of the user.

### Claim 7 (depends on 1)

The system of claim 1, wherein the processor is configured to execute further processor-executable instructions stored in the non-transitory computer-readable medium to:
calculate a level of risk of the user based on the previous activity, wherein
the content moderation settings are based on the calculated level of risk.

### Claim 8 (depends on 1)

The system of claim 1, wherein the processor is configured to execute further processor-executable instructions stored in the non-transitory computer-readable medium to:
receive, from one or more of the plurality of client devices, an indication indicating additional inappropriate or harmful content in the stream; and
updating the training of the machine learning model with the indication and the additional content.

### Claim 9 (depends on 1)

The non-transitory computer-readable medium of claim 1, wherein the processor-executable instructions are further configured to cause the one or more processors to:
receive, from a moderator client device, a set of additional content moderation settings,
wherein applying the machine learning model to the stream further comprises applying the additional content moderation settings to the machine learning model and wherein the machine learning model identifies the item of content at least in part based on the additional content moderation settings.

### Claim 10

A method comprising:
establishing a video conference between a plurality of client devices;
receiving, from a first client device of the plurality of client devices, a stream corresponding to the video conference and comprising one or more items of content;
configuring content moderation settings in response to a previous activity of a user associated with the first client device;
providing the content moderation settings to a machine learning model that is trained to identify inappropriate or harmful content in the stream;
applying, to the stream, the machine learning model;
determining, via the machine learning model, that an item of content in the stream is inappropriate or harmful;
editing the stream to remove the item of content; and
transmitting, to the plurality of client devices, the edited stream.

### Claim 11 (depends on 10)

The method of claim 10, further comprising
transmitting, to a moderator client device, the item of content; and
receiving, from the moderator client device, a request to edit the item of content in the stream,
wherein the editing is responsive to receiving the request to edit the item of content.

### Claim 12 (depends on 10)

The method of claim 10, further comprising
receiving, from a moderator client device, a set of additional content moderation settings,
wherein applying the machine learning model to the stream further comprises applying the additional content moderation settings to the machine learning model and wherein the machine learning model identifies the item of content at least in part based on the additional content moderation settings.

### Claim 13 (depends on 10)

The method of claim 10, wherein the one or more items of content include audio, the method further comprising:
performing speech recognition on the audio to identify one or more recognized words; and
identifying, based on the one or more recognized words, one or more keywords corresponding to inappropriate or harmful content.

### Claim 14 (depends on 10)

The method of claim 10, wherein the one or more items of content include text and wherein the machine learning model is trained to:
determine that one or more keywords identified as inappropriate or harmful are present in the text; and
in response to the determining, identify the content as inappropriate or harmful.

### Claim 15 (depends on 10)

The method of claim 10, further comprising:
in response to the determination of the item of content in the stream being inappropriate, performing one or more of (i) terminating services to a user associated with the first client device, (ii) muting the user, and (iii) turning off video of the user.

### Claim 16 (depends on 10)

The method of claim 10, further comprising:
calculating a level of risk of the user based on the previous activity, wherein the
content moderation settings are based on the calculated level of risk.

### Claim 17

A non-transitory computer-readable medium comprising processor-executable instructions configured to cause one or more processors to:
establish a video conference between a plurality of client devices;
receive, from a first client device of the plurality of client devices, a stream corresponding to the video conference and comprising one or more items of content;
configure content moderation settings in response to a previous activity of a user associated with the first client device;
provide the content moderation settings to a machine learning model that is trained to identify inappropriate or harmful content in the stream;
apply, to the stream, the machine learning model;
determine, via the machine learning model, that an item of content in the stream is inappropriate or harmful;
edit the stream to remove the item of content; and
transmit, to the plurality of client devices, the edited stream.

### Claim 18 (depends on 17)

The non-transitory computer-readable medium of claim 17, wherein the one or more items of content include audio and wherein the processor-executable instructions are further configured to cause the one or more processors to:
perform speech recognition on the audio to identify one or more recognized words; and
identify, based on the one or more recognized words, one or more keywords corresponding to inappropriate or harmful content.

### Claim 19 (depends on 17)

The non-transitory computer-readable medium of claim 17, wherein the one or more items of content include text and wherein the processor-executable instructions are further configured to cause the one or more processors to:
determine that one or more keywords identified as inappropriate or harmful are present in the text; and
in response to the determining, identify the content as inappropriate or harmful.

### Claim 20 (depends on 17)

The non-transitory computer-readable medium of claim 17, wherein the processor-executable instructions are further configured to cause the one or more processors to:
receive, from one or more of the plurality of client devices, an indication indicating additional inappropriate or harmful content in the stream; and
updating the training of the machine learning model with the indication and the additional content.