System

A system comprising:

• circuitry configured to:
• acquire category data and rank data, extract feature vectors from historical performance records stored in a storage device, and compute structured target values comprising a target metric, a target change rate, and a time frame using a deep neural network comprising a text-encoding subnetwork applying transformer-based self-attention and a dense prediction subnetwork applying nonlinear activation functions;
• construct a first prompt sentence embedding the structured target values and input the first prompt sentence to a generative neural network model to produce a natural-language goal sentence, and store the category data, the rank data, and the natural-language goal sentence in association in the storage device;
• receive achievement content, construct a second prompt sentence embedding the achievement content and the natural-language goal sentence, input the second prompt sentence to the generative neural network model, and produce a primary evaluation comprising a degree-of-achievement score and qualitative assessment labels;
• acquire audio information and image information from a terminal device via a communication interface coupled to a packet-switched network, apply an emotion analysis model to extract emotion category labels and intensity scores, and complement the primary evaluation using a weighting function applied to the degree-of-achievement score and the intensity scores to produce an achievement evaluation result;
• generate visualization data comprising time-series information, achievement-rate information, and rank-comparison information and transmit the visualization data and the natural-language goal sentence to the terminal device for integrated display;
• and
• accumulate the achievement evaluation result and the emotion data as updated historical records and perform parameter adjustment of the deep neural network and the generative neural network model using mini-batch weight updates.

A method comprising:

• acquiring category data and rank data, obtaining historical performance records from a storage device, extracting feature vectors from the historical performance records, and computing structured target values using a machine learning model applied to the feature vectors based on the category data and the rank data;
• generating a first prompt sentence encoding the structured target values, the category data, and the rank data and inputting the first prompt sentence to a generative neural network model to produce a natural-language goal sentence;
• receiving achievement content via a communication interface coupled to a packet-switched network, generating a second prompt sentence based on the achievement content, inputting the second prompt sentence to the generative neural network model, and performing a primary evaluation of a degree of achievement based on model output;
• acquiring audio information and image information from a terminal device, applying an emotion analysis model to the audio information and the image information to extract emotion data, and complementing the primary evaluation using the emotion data to produce an achievement evaluation result;
• and
• accumulating the achievement evaluation result and the emotion data as updated historical records and performing parameter adjustment of the machine learning model using the updated historical records.