System

A system comprising:

• circuitry configured to:
• receive, via a packet-switched network, profile attribute data and interaction history records from a terminal device and from an external information processing apparatus, and store the profile attribute data and the interaction history records in a storage device;
• perform preprocessing on the profile attribute data and the interaction history records comprising detecting and replacing missing values, normalizing numerical attribute values, and converting categorical attribute values into numerical vector representations to produce feature vectors, and compute aggregate statistical values from the interaction history records and append the aggregate statistical values to the feature vectors to produce enhanced feature vectors;
• train a prediction neural network using the enhanced feature vectors and labeled outcome data by executing iterative gradient-based optimization of a loss function to adjust weight matrices, and store trained parameters in the storage device;
• construct combined input tensors from a target feature vector of a user and candidate item feature vectors, execute forward inference through the prediction neural network to produce evaluation score values, and rank candidate items to identify a selected group comprising gift candidate items;
• construct a structured prompt for a generative neural network model by inserting profile information comprising at least an age category, a gender category, a preference field category, and an occupation category of the user, and item attribute data of each gift candidate item into a predefined template, and transmit the structured prompt to an information processing apparatus hosting the generative neural network model via the packet-switched network;
• receive a generated output token sequence, parse the generated output token sequence to extract recommendation reason information and explanation information for each gift candidate item;
• transmit the gift candidate items, the evaluation score values, and the recommendation reason information to the terminal device via the packet-switched network to cause rendering on a display of the terminal device;
• receive selection result data from the terminal device, store the selection result data in the storage device;
• and
• retrain the prediction neural network using the selection result data as updated labeled outcome data.

A method performed by circuitry, the method comprising:

• receiving, via a packet-switched network, profile attribute data and interaction history records from a terminal device, and storing the profile attribute data and the interaction history records in a storage device;
• performing preprocessing on the profile attribute data and the interaction history records to produce feature vectors;
• training a prediction neural network using the feature vectors and labeled outcome data by executing iterative optimization to adjust weight parameters of the prediction neural network, and storing the weight parameters in the storage device;
• executing forward inference through the prediction neural network using a feature vector associated with a user of the terminal device and feature vectors of candidate items to produce evaluation score values for the candidate items, and identifying a selected group of the candidate items based on the evaluation score values;
• constructing a structured prompt for a generative neural network model based on the profile attribute data and attribute data of each candidate item in the selected group, and transmitting the structured prompt to an information processing apparatus hosting the generative neural network model via the packet-switched network;
• receiving a generated output token sequence from the generative neural network model and extracting descriptive annotation data for each candidate item in the selected group;
• transmitting a response data structure comprising the selected group and the descriptive annotation data to the terminal device via the packet-switched network to cause rendering on a display of the terminal device;
• receiving selection result data from the terminal device and storing the selection result data in the storage device;
• and
• updating the prediction neural network using the selection result data as additional labeled outcome data.