GENERATING IMPROVED DIGITAL TRANSCRIPTS UTILIZING DIGITAL TRANSCRIPTION MODELS THAT ANALYZE DYNAMIC MEETING CONTEXTS
The present disclosure relates to systems, non-transitory computer-readable media, and methods for improving digital transcripts of a meeting based on user information. For example, a digital transcription system creates a digital transcription model to automatically transcribe audio from a meeting based on documents associated with meeting participants, event details, user features, and other meeting context data. In one or more embodiments, the digital transcription model creates a digital lexicon based on the user information, which the digital transcription system uses to generate the digital transcript. In some embodiments, the digital transcription model trains and utilizes a digital transcription neural network to generate the digital transcript.
1 . A computer-implemented method comprising:
receiving audio data captured by a recording device during a meeting;
identifying, from a collection of digital content maintained on a content management system, digital documents stored or accessed by a user account of a participant in the meeting;
determining, by analyzing a subject associated with the meeting and subjects associated with the digital documents, meeting context data for the meeting, the meeting context data including at least a subset of the digital documents corresponding to the subject associated with the meeting; and
generating, utilizing a digital transcription neural network to process the audio data and the meeting context data, a digital transcript of the meeting, the digital transcript including words identified within the meeting context data and the audio data.
2 . The computer-implemented method of claim 1 , wherein determining the meeting context data comprises selecting the subset of the digital documents by comparing the subject associated with the meeting and the subjects associated with the digital documents.
3 . The computer-implemented method of claim 1 , wherein determining the meeting context data comprises:
determining, by comparing content of the digital documents with event details of the meeting, relevance levels for respective documents of the digital documents, wherein the event details include the subject for the meeting; and
selecting the subset of the digital documents by comparing the relevance levels to a threshold relevance level.
4 . The computer-implemented method of claim 1 , further comprising:
identifying an additional user account within an organization associated with the meeting, wherein the additional user account is not associated with any identified participant of the meeting;
in response to identifying the additional user account, identifying, from the collection of digital content, additional digital documents stored or accessed by the additional user account; and
including at least a subset of the additional digital documents in the meeting context data for the meeting.
5 . The computer-implemented method of claim 1 , further comprising:
identifying meeting notes provided by one or more participants of the meeting; and
including the meeting notes in the meeting context data for the meeting.
6 . The computer-implemented method of claim 5 , wherein the meeting notes include timestamp information indicating when one or more note entries were recorded relative to the audio data captured by the recording device during the meeting.
7 . The computer-implemented method of claim 1 , further comprising:
identifying one or more additional user accounts of one or more additional participants of the meeting;
identifying, from the collection of digital content, additional digital documents stored or accessed by the one or more additional user accounts; and
including at least a subset of the additional digital documents in the meeting context data for the meeting.
8 . The computer-implemented method of claim 7 , further comprising:
receiving additional audio data captured during the meeting by one or more recording devices associated with the one or more additional participants of the meeting.
9 . A system comprising:
at least one processor; and
a non-transitory computer memory comprising instructions that, when executed by the at least one processor, cause the system to:
receive audio data captured by a recording device of a user account participating in a meeting;
identify, from a collection of digital content maintained on a content management system, digital documents stored or accessed by the user account participating in the meeting; and
generate, utilizing a digital transcription neural network to process the audio data and the digital documents, a digital transcript of the meeting, the digital transcript including words identified within the digital documents and the audio data.
10 . The system of claim 9 , wherein the collection of digital content belongs to and is maintained on the content management system for an organization associated with the meeting.
11 . The system of claim 10 , further comprising instructions that, when executed by the at least one processor, cause the system to identify the user account by:
identifying, based on event details of the meeting, a member of the organization as a participant in the meeting; and
determining the user account belongs to the participant.
12 . The system of claim 9 , wherein identifying the digital documents is further based on comparing content from the collection of digital content with a subject associated with the meeting.
13 . The system of claim 12 , further comprising instructions that, when executed by the at least one processor, cause the system to identify the subject for the meeting from one or more of a meeting agenda, a calendar invitation, an attachment to the calendar invitation, or a collaboration graph indicating relationships between users, projects, and documents of an organization associated with the meeting.
14 . The system of claim 9 , further comprising instructions that, when executed by the at least one processor, cause the system to:
identify meeting notes provided by one or more participants of the meeting; and
include the meeting notes in the digital documents for the meeting.
15 . The system of claim 9 , further comprising instructions that, when executed by the at least one processor, cause the system to:
identify one or more additional user accounts of one or more additional participants of the meeting;
identify, from the collection of digital content, additional digital documents stored or accessed by the one or more additional user accounts; and
include the additional digital documents in the digital documents for the meeting.
16 . A non-transitory computer-readable storage medium comprising instructions that, when executed by at least one processor, cause a computer system to:
receive audio data captured by a recording device of a user account participating in a meeting;
identify, from a collection of digital content maintained on a content management system, digital documents stored or accessed by the user account participating in the meeting; and
generate, utilizing a digital transcription neural network to process the audio data and the digital documents, a digital transcript of the meeting, the digital transcript including words identified within the digital documents and the audio data.
17 . The non-transitory computer-readable storage medium of claim 16 , further comprising instructions that, when executed by the at least one processor, cause the computer system to filter the digital documents for processing by the digital transcription neural network by:
comparing content of the digital documents with event details of the meeting to determine relevance levels for respective documents of the digital documents; and
selecting at least a subset of the digital documents by comparing the relevance levels to a threshold relevance level.
18 . The non-transitory computer-readable storage medium of claim 16 , further comprising instructions that, when executed by the at least one processor, cause the computer system to process the audio data and the digital documents by:
utilizing input layers of the digital transcription neural network to encode learned feature vectors from the audio data and the digital documents;
utilizing hidden layers of the digital transcription neural network to modify the learned feature vectors according to learned parameters of the hidden layers; and
providing the learned feature vectors to output layers of the digital transcription neural network to convert words spoken in the meeting from the audio data to text by translating the learned feature vectors according to learned parameters of the output layers.
19 . The non-transitory computer-readable storage medium of claim 16 , further comprising instructions that, when executed by the at least one processor, cause the computer system to train the digital transcription neural network by:
generating synthetic audio data from a plurality of digital training documents corresponding to a meeting subject utilizing a text-to-speech model;
providing the synthetic audio data to the digital transcription neural network to learn encoded features of the synthetic audio data; and
training the digital transcription neural network utilizing the plurality of digital training documents as a ground-truth to the synthetic audio data.
20 . The non-transitory computer-readable storage medium of claim 16 , wherein the digital transcription neural network comprises a digital lexicon neural network configured to generate a weighted frequency distribution of words or word patterns within the digital documents based on respective relevance levels between the words or word patterns and a subject associated with the meeting.