LANGUAGE MODEL CREATION DEVICE, LANGUAGE MODEL CREATION METHOD AND PROGRAM

摘要

PROBLEM TO BE SOLVED: To provide a language model creation device capable of learning a language model including many unknown word chains without greatly changing origination probability distribution of a language phenomenon obtained from a writing text.SOLUTION: A language model creation device includes: an unknown model generation unit within a task for generating, with a writing text as a learning text, probability distribution (hereinafter called as an unknown model within a task) that gives a score of an unknown likeliness within a task with respect to an arbitrary text; an unknown text selection unit within a task for selecting, with the unknown model within a task and the writing text, an unknown text within a task from a prepared language resource (hereafter called as another language resource) different from the writing text; and a language model extension unit for creating a language model from the writing text and the unknown text within a task.