Back to results
US 2026/0289069 A1Published application
System
AssigneeSoftBank Group Corp.
InventorKei ONO
Life of this patent
Estimating the term…
Abstract
A system includes a processor that is configured to acquire meeting recording data and convert audio data included in the meeting recording data into text data, generate a prompt for instructing a generative artificial intelligence model to extract meeting topics, decisions, and action items from the text data, input the generated prompt into the generative artificial intelligence model, receive extracted information from the generative artificial intelligence model, summarize the extracted information, and identify links to related materials based on the extracted information.
Claims20
A system comprising:
- circuitry configured to:
- acquire, based on session identification information, recording data from an external storage device via a communication interface coupled to a packet-switched network, extract an audio signal from the recording data, and transmit the audio signal to a speech recognition processing apparatus to obtain text data;
- generate a first prompt sentence based on the text data and input the first prompt sentence together with the text data to a generative neural network model to extract topic data, decision data, and action item data, and store the extracted data as structured records;
- generate a second prompt sentence based on at least one of the text data and the structured records and input the second prompt sentence to the generative neural network model to generate summary information, and store the summary information in association with link information indicating location information corresponding to documents shared during the session;
- associate time information of the recording data with utterance units of the text data in a time-indexed transcript, enabling identification of a recording data time range corresponding to a text segment matching a search term;
- and
- generate notification information based on the summary information, the structured records, and the link information, and transmit the notification information to participant terminal devices via the communication interface.
4 more elements
A system comprising:
- circuitry configured to:
- acquire, based on session identification information, recording data from an external storage device via a communication interface coupled to a packet-switched network, demultiplex audio streams from the recording data, and transmit the audio streams to a speech recognition processing apparatus to obtain timestamped utterance units comprising speaker identifiers, start time offsets, end time offsets, and recognized text tokens;
- construct a time-indexed transcript by indexing the timestamped utterance units by start time offset, generate a first prompt sentence with an extraction instruction section and a text data content section, and input the first prompt sentence to a generative neural network model to extract topic records, decision records, and action item records, each linked by a session identifier and a time offset range;
- apply extractive pre-selection by scoring utterance units with term frequency-inverse document frequency weighting, generate a second prompt sentence with a token budget constraint from selected utterance units and key structured records, and input the second prompt sentence to the generative neural network model to produce summary information stored with associated link information;
- resolve action item records to responsible participant identifiers by cross-referencing speaker identifier fields with a participant registry, and personalize notification information per participant by filtering action item records by responsible participant identifier;
- and
- transmit personalized notification information comprising summary sections, filtered action item list sections, and reference sections with formatted link information to participant terminal devices via the communication interface.
4 more elements
A method comprising:
- acquiring, based on session identification information, recording data from an external storage device via a communication interface coupled to a packet-switched network, extracting an audio signal, and transmitting the audio signal to a speech recognition processing apparatus to obtain text data;
- generating a first prompt sentence and inputting the first prompt sentence together with the text data to a generative neural network model to extract topic data, decision data, and action item data as structured records;
- generating a second prompt sentence and inputting the second prompt sentence to the generative neural network model to generate summary information stored in association with link information;
- associating time information of the recording data with utterance units of the text data in a time-indexed transcript;
- and
- generating notification information based on the summary information, the structured records, and the link information and transmitting the notification information to participant terminal devices via the communication interface.
3 more elements
Dependencies are inferred from the claim text.