LANGUAGE MODEL CREATION DEVICE, SPEECH RECOGNITION DEVICE, LANGUAGE MODEL CREATION METHOD AND PROGRAM STORAGE MEDIUM

摘要

The recognition accuracy in the process of recognizing the speech of another speaker may be improved even if a word that is not included in a speech of one speaker of mutually conversing speakers is included in the speech of another speaker responding to that speech. A language model creation device (18) for creating a language model that is used for a speech recognition process includes an extraction unit (100) and a model creation unit (101). The extraction unit (100) extracts second data from a data group (20) on the basis of a word or a word string included in the speech of a first speaker who is having a conversation with another speaker, which is to be recognized. The data group (20) is constituted by a collection of pairs of associated words which associate first data that is a word or a word string included in a first speech with the second data that is a word associated with the first data in the words or the word strings included in a second speech responding to the first speech. The model creation unit (101) creates a language model in accordance with the speech of another speaker by making use of the second data extracted by the extraction unit (100).