System

A system comprising:

• circuitry configured to:
• acquire, via an acoustic transducer coupled to a communication interface, acoustic signal data and convert the acoustic signal data into a time-series digital signal;
• execute a speech recognition algorithm on the time-series digital signal to generate character string data, and store the character string data in association with time information in a transcript data structure;
• apply a natural language processing algorithm to the character string data to perform morphological analysis, part-of-speech analysis, and entity recognition to extract expression data, and generate index information mapping each extracted expression to occurrence position records within the character string data;
• calculate importance index values for the extracted expressions, compute mutual similarity values between expression pairs using vector representations, and integrate expressions into concept unit data by clustering expressions exceeding a similarity threshold;
• construct a parameterized instruction sequence based on the character string data, the index information, and the concept unit data, and provide the parameterized instruction sequence to a generative neural network model having a transformer-based architecture including a plurality of self-attention layers to generate structured output data;
• and
• transmit the index information and the structured output data to a terminal apparatus via a packet-switched network for visual presentation.

A method performed by circuitry, the method comprising:

• acquiring acoustic signal data via an acoustic transducer and converting the acoustic signal data into a time-series digital signal based on a predetermined sampling condition;
• executing a speech recognition algorithm on the time-series digital signal to sequentially generate character string data corresponding to the acoustic signal data;
• applying a natural language processing algorithm to the character string data to perform morphological analysis and part-of-speech analysis to extract expression data from the character string data;
• generating index information indicating a correspondence between each extracted expression and the character string data including the extracted expression, and storing the index information in a search data structure;
• and
• constructing a parameterized instruction sequence based on at least the character string data and the index information, and providing the parameterized instruction sequence to a generative neural network model to cause the generative neural network model to generate structured output data.