Granted Patent B1

US 12,567,412 · App. 18/306,628 · Granted Mar 3, 2026

Inventor: Aleksandra Swerdlow (Santa Clara, CA)

Assignee: Zoom Communications, Inc.

G10L15/22G06F40/134G10L15/08G10L15/30G10L2015/088

Abstract

Claims (66)

1 . A method comprising:

detecting, by a server comprising one or more processors configured to execute non-transitory computer-readable instructions, speech from a conference participant within an audio stream of a conference, wherein the conference participant is associated with an organization;

converting, by the server, the speech to text data;

detecting, by the server, a name in the text data, wherein the name lacks a match in a contact list associated with persons in the organization;

determining, by the server, from the contact list, a list of person candidates associated with the name based on scores determined for individual persons in the contact list relative to the name;

determining, by the server, a connection strength score between the conference participant and a person in the list of person candidates;

determining, by the server, based on the connection strength score, that the person is an intended person referred to by the conference participant; and

generating, by the server, a transcript of the conference based on the text data, wherein the transcript includes a name of the intended person;

wherein the connection strength score is computed by a machine learning model trained on historical conference data to associate speech patterns with conference participants and the transcript includes an embedded hyperlink to a dynamically generated profile card for the intended person, the profile card being retrieved from a remote server in response to selection of the hyperlink.

2 . The method of claim 1 , further comprising:

determining whether the connection strength score meets a threshold;

adding the person to a second list of person candidates when the connection strength score meets the threshold;

selecting a candidate name with a highest connection strength score from the second list of person candidates; and

generating the transcript, wherein the transcript includes the candidate name.

3 . The method of claim 1 , further comprising:

transmitting a confirmation request to a user device of the conference participant, wherein the confirmation request is a pop-up notification requesting a confirmation of the name.

4 . The method of claim 1 , further comprising:

transmitting a confirmation request to a user device of the conference participant, wherein the confirmation request is a pop-up notification requesting a confirmation of the name; and

receiving a response that indicates that the name is confirmed.

5 . The method of claim 1 , further comprising:

embedding a hyperlink to the name in the transcript, wherein the hyperlink is associated with a profile card of the person associated with the name.

6 . The method of claim 1 , wherein the connection strength score is based on a number of interactions between the person and the conference participant.

7 . The method of claim 1 , wherein the connection strength score is based on a recency score of one or more interactions between the person and the conference participant.

8 . The method of claim 1 , wherein the machine learning model is trained for contextual awareness.

9 . A server comprising:

one or more processors configured execute non-transitory computer-readable instructions to:

detect speech from a conference participant within an audio stream of a conference, wherein the conference participant is associated with an organization;

convert the speech to text data;

detect a name in the text data, wherein the name lacks a match in a contact list associated with persons in the organization;

determine, from the contact list, a list of person candidates associated with the name based on scores determined for individual persons in the contact list relative to the name;

determine a connection strength score between the conference participant and a person in the list of person candidates;

determine, based on the connection strength score, that the person is an intended person referred to by the conference participant; and

generate a transcript of the conference based on the text data, wherein the transcript includes a name of the intended person;

wherein the connection strength score is computed by a machine learning model trained on historical conference data to associate speech patterns with conference participants and the transcript includes an embedded hyperlink to a dynamically generated profile card for the intended person, the profile card being retrieved from a remote server in response to selection of the hyperlink.

10 . The server of claim 9 , wherein the one or more processors are further configured to:

determine a respective connection strength score for each person of the list of person candidates;

determine a context of the text data; and

determine the intended person based on the connection strength scores for each person and the context.

11 . The server of claim 9 , wherein the one or more processors are further configured to:

transmit a confirmation request to a user device of the conference participant, wherein the confirmation request is a chat room notification requesting a confirmation of the name.

12 . The server of claim 9 , wherein the one or more processors are further configured to:

transmit a confirmation request to a user device of the conference participant, wherein the confirmation request is a chat room notification requesting a confirmation of the name; and

receiving a response that indicates that the name is confirmed.

13 . The server of claim 9 , wherein the connection strength score is based on a number of common interactions of the person and the conference participant, wherein a common interaction is based on activity of the person associated with the name and activity of the conference participant with a common application.

14 . The server of claim 9 , wherein the contact list includes conference participants not associated with the organization.

15 . A non-transitory computer-readable medium comprising instructions stored on a memory, that when executed by a processor, cause the processor to perform operations comprising:

detecting speech from a conference participant within an audio stream of a conference, wherein the conference participant is associated with an organization;

converting the speech to text data;

detecting a name in the text data, wherein the name lacks a match in a contact list associated with persons in the organization;

determining, from the contact list, a list of person candidates associated with the name based on scores determined for individual persons in the contact list relative to the name;

determining a connection strength score between the conference participant and a person in the list of person candidates;

determining, based on the connection strength score, that the person is an intended person referred to by the conference participant; and

generating a transcript of the conference based on the text data, wherein the transcript includes a name of the intended person;

wherein the connection strength score is computed by a machine learning model trained on historical conference data to associate speech patterns with conference participants and the transcript includes an embedded hyperlink to a dynamically generated profile card for the intended person, the profile card being retrieved from a remote server in response to selection of the hyperlink.

16 . The non-transitory computer-readable medium of claim 15 , wherein the instructions, when executed by the processor, cause the processor to perform operations comprising:

detecting the name by a keyword that references one or more subjects.

17 . The non-transitory computer-readable medium of claim 15 , wherein the instructions, when executed by the processor, cause the processor to perform operations comprising:

detecting the name based on a determination of keywords within a neighboring word range of a detected phrase.

18 . The non-transitory computer-readable medium of claim 15 , wherein the instructions, when executed by the processor, cause the processor to perform operations comprising:

transmitting a confirmation request to a user device of the conference participant, wherein the confirmation request is an email notification requesting a confirmation of the name.

19 . The non-transitory computer-readable medium of claim 15 , wherein the instructions, when executed by the processor, cause the processor to perform operations comprising:

transmitting a confirmation request to a user device of the conference participant, wherein the confirmation request is an email notification requesting a confirmation of the name; and

receiving a response that indicates that the name is confirmed.

20 . The non-transitory computer-readable medium of claim 15 ,

wherein the

machine learning model is trained for contextual awareness.

Assignments (2)

References Cited (10)