Granted Patent B2

US 12,647,532 · App. 18/441,548 · Granted Jun 2, 2026

Inventors: Davide Giovanardi (San Jose, CA); Bilung Lee (Irvine, CA); Felix Schneider (Baden-Wurttemberg, DE); Marco Turchi (Trento, IT); Alexander Waibel (Sammamish, WA); Yun Zhang (Pittsburgh, PA)

Assignee: Zoom Communications, Inc.

H04N7/155H04N7/152

Abstract

Claims (79)

1 . A method performed by one or more computing devices, the method comprising:

establishing a video conference for a plurality of participants;

while the video conference is in progress,

receiving a first portion of a transcript of the video conference;

generating a first meeting summary based on the first portion of the transcript comprising:

determining different sets of chunks based on the first portion of the transcript, each set of chunks comprising a different subset of one or more chunks of the first portion of the transcript,

generating a plurality of candidate meeting summaries comprising applying a meeting summarization model to the different sets of chunks, and

selecting a candidate meeting summary from the plurality of candidate meeting summaries that has a highest quality score as the first meeting summary,

causing the first meeting summary to be presented in a user interface accessible by a client computing device associated with at least one of the plurality of participants;

receiving a second portion of the transcript of the video conference;

generating a second meeting summary based on the first portion and the second portion of the transcript; and

causing the second meeting summary to be presented in the user interface.

2 . The method of claim 1 , wherein a quality score comprises one or more of a confidence score output by the meeting summarization model or a similarity metric calculated based on embeddings of a corresponding candidate meeting summary and the corresponding set of chunks.

3 . The method of claim 1 , wherein generating the plurality of candidate meeting summaries comprises:

generating a first candidate meeting summary by providing a first chunk into the meeting summarization model;

generating a second candidate meeting summary by providing the first chunk and a second chunk into the meeting summary model; and

generating a third candidate meeting summary by providing the one or more chunks into the meeting summary model.

4 . The method of claim 1 , wherein generating the second meeting summary based on the first portion and the second portion of the transcript comprises:

generating a second plurality of candidate meeting summaries by applying the meeting summarization model to subsets of chunks in the second portion of the transcript and one or more of the chunks in the first portion of the transcript that are not used in generating the selected candidate meeting summary.

5 . The method of claim 1 , wherein the second meeting summary comprises the first meeting summary and an additional meeting summary, and wherein:

generating the first meeting summary comprises applying the meeting summarization model to the first portion of the transcript;

generating the second meeting summary based on the first portion and the second portion of the transcript comprises applying the meeting summary model to a first concatenation of the first portion of the transcript and the second portion of the transcript by forcing the meeting summary model to output the first meeting summary before the additional meeting summary in the second meeting summary is output; and

presenting the second meeting summary in the user interface comprises presenting the additional meeting summary.

6 . The method of claim 5 , further comprising:

receiving a third portion of the transcript of the video conference;

generating a third meeting summary based on the second portion and the third portion of the transcript by applying the meeting summarization model to a concatenation of the second portion of the transcript and the third portion of the transcript by forcing the meeting summary model to output the second meeting summary before other meeting summary in the third meeting summary is output; and

presenting the other meeting summary in the third meeting summary in the user interface.

7 . The method of claim 5 , further comprising:

receiving an edit to the first meeting summary, wherein generating the second meeting summary based on the first portion and the second portion of the transcript comprises applying the meeting summarization model to a second concatenation of the first portion of the transcript and the second portion of the transcript by forcing the meeting summary model to output the edited first meeting summary before the additional meeting summary in the second meeting summary is output.

8 . A computing device, comprising:

a non-transitory computer-readable medium; and

a processor communicatively coupled to the non-transitory computer-readable medium, the processor configured to execute processor-executable instructions stored in the non-transitory computer-readable medium to:

establish a video conference for a plurality of participants;

while the video conference is in progress,

receive a first portion of a transcript of the video conference;

generate a first meeting summary based on the first portion of the transcript, comprising executing processor-executable instructions stored in the non-transitory computer-readable medium to:

determine different sets of chunks based on the first portion of the transcript, each set of chunks comprising a different subset of one or more chunks of the first portion of the transcript,

apply a meeting summarization model to the different sets of chunks to generate a plurality of candidate meeting summaries, and

select a candidate meeting summary from the plurality of candidate meeting summaries that has a highest quality score as the first meeting summary;

cause the first meeting summary to be presented in a user interface accessible by a client computing device associated with at least one of the plurality of participants;

receive a second portion of the transcript of the video conference;

generate a second meeting summary based on the first portion and the second portion of the transcript; and

cause the second meeting summary to be presented in the user interface.

9 . The computing device of claim 8 , wherein a quality score comprises one or more of a confidence score output by the meeting summarization model or a similarity metric calculated based on embeddings of a corresponding candidate meeting summary and the corresponding set of chunks.

10 . The computing device of claim 8 , wherein generating the plurality of candidate meeting summaries comprises:

generating a first candidate meeting summary by providing a first chunk into the meeting summarization model;

generating a second candidate meeting summary by providing the first chunk and a second chunk into the meeting summary model; and

generating a third candidate meeting summary by providing the one or more chunks into the meeting summary model.

11 . The computing device of claim 8 , wherein generating the second meeting summary based on the first portion and the second portion of the transcript comprises:

generating a second plurality of candidate meeting summaries by applying the meeting summarization model to subsets of chunks in the second portion of the transcript and one or more of the chunks in the first portion of the transcript that are not used in generating the selected candidate meeting summary.

12 . The computing device of claim 8 , wherein the second meeting summary comprises the first meeting summary and an additional meeting summary, and wherein:

generating the first meeting summary comprises applying the meeting summarization model to the first portion of the transcript;

generating the second meeting summary based on the first portion and the second portion of the transcript comprises applying the meeting summary model to a concatenation of the first portion of the transcript and the second portion of the transcript by forcing the meeting summary model to output the first meeting summary before the additional meeting summary in the second meeting summary is output; and

presenting the second meeting summary in the user interface comprises presenting the additional meeting summary.

13 . The computing device of claim 12 , wherein the processor is configured to execute the processor-executable instructions stored in the non-transitory computer-readable medium to further:

receive a third portion of the transcript of the video conference;

generate a third meeting summary based on the second portion and the third portion of the transcript by applying the meeting summarization model to a concatenation of the second portion of the transcript and the third portion of the transcript by forcing the meeting summary model to output the second meeting summary before other meeting summary in the third meeting summary is output; and

present the other meeting summary in the third meeting summary in the user interface.

14 . A non-transitory computer-readable medium comprising processor-executable instructions configured to cause one or more processors to:

establish a video conference for a plurality of participants;

while the video conference is in progress,

receive a first portion of a transcript of the video conference;

generate a first meeting summary based on the first portion of the transcript, comprising processor-executable instructions stored in the non-transitory computer-readable medium to cause the one or more processors to:

determine different sets of chunks based on the first portion of the transcript, each set of chunks comprising a different subset of one or more chunks of the first portion of the transcript,

apply a meeting summarization model to the different sets of chunks to generate a plurality of candidate meeting summaries, and

select a candidate meeting summary from the plurality of candidate meeting summaries that has a highest quality score as the first meeting summary;

cause the first meeting summary to be presented in a user interface accessible by a client computing device associated with at least one of the plurality of participants;

receive a second portion of the transcript of the video conference;

generate a second meeting summary based on the first portion and the second portion of the transcript; and

cause the second meeting summary to be presented in the user interface.

15 . The non-transitory computer-readable medium of claim 14 , wherein a quality score comprises one or more of a confidence score output by the meeting summarization model or a similarity metric calculated based on embeddings of a corresponding candidate meeting summary and the corresponding set of chunks.

16 . The non-transitory computer-readable medium of claim 14 , wherein generating the plurality of candidate meeting summaries comprises:

generating a first candidate meeting summary by providing a first chunk into the meeting summarization model;

generating a second candidate meeting summary by providing the first chunk and a second chunk into the meeting summary model; and

generating a third candidate meeting summary by providing the one or more chunks into the meeting summary model.

17 . The non-transitory computer-readable medium of claim 14 , wherein the second meeting summary comprises the first meeting summary and an additional meeting summary, and wherein:

generating the first meeting summary comprises applying the meeting summarization model to the first portion of the transcript;

generating the second meeting summary based on the first portion and the second portion of the transcript comprises applying the meeting summary model to a concatenation of the first portion of the transcript and the second portion of the transcript by forcing the meeting summary model to output the first meeting summary before the additional meeting summary in the second meeting summary is output; and

presenting the second meeting summary in the user interface comprises presenting the additional meeting summary.

Assignments (2)

Continuity (1)

References Cited (42)