TRAINING DATA GENERATION SYSTEM, TRAINING DATA GENERATION METHOD, AND PROGRAM

Fecha de publicación: 22/12/2022
Fuente: WIPO (eseential oils OR extracts)
The purpose of the present disclosure is to facilitate the suppression of an increase in training data. According to the present invention, an acquisition unit (11) acquires first and second reference data (A1, A2). A generation unit (12) generates a plurality of pieces of first candidate data and a plurality of pieces of second candidate data. The plurality of pieces of first candidate data include the first reference data (A1) and first extension data. The plurality of pieces of second candidate data include the second reference data (A2) and second extension data. An extraction unit (13) respectively extracts a plurality of first feature amounts from the plurality of pieces of first candidate data and a plurality of second feature amounts from the plurality of pieces of second candidate data. A calculation unit (14) calculates a plurality of degrees of mutual similarity between the plurality of first feature amounts and the plurality of second feature amounts. A determination unit (15) determines, as two or more pieces of similar data, two or more pieces of candidate data that correspond to the degree of similarity, which is at least a threshold, from the plurality of pieces of first candidate data and the plurality of pieces of second candidate data. An exclusion unit (16) excludes, from a candidate of training data, data satisfying a prescribed condition among the two or more pieces of similar data.