Supervised machine learning apparatus, supervised machine learning method, and supervised machine learning program

摘要

When the boundary between correct answers of a plurality of facts provided adjacent to each other is different from the boundary of a morpheme as a result of morpheme analysis, the correct answers are effectively used for machine learning. A supervised machine learning device that extracts facts for correct answers attached to training data from test data, based on the feature values of each morpheme constituting the training data and the first new training data. A learning unit that generates a model for extracting a fact from the first new training data and extracting a first new fact from the first new training data, and a relationship between the model and the facts identified by a predetermined label And extracting a first new fact identified by the first new label and the first new label from the training data, and a first new fact including the extracted first new fact And a generation unit for generating various training data.