CONTROLLER, SPOKEN DIALOG DEVICE, SPEECH RECOGNITION SERVER, AND PROGRAM

To provide a controller for a spoken dialog device with which it is possible to shorten the time needed till a response and lighten the unnaturalness of a dialog.SOLUTION: The controller includes an arithmetic unit that controls a spoken dialog device having a plurality of speech recognition models, the controller indicating to the spoken dialog device at the start of dialog with a subject that speech recognition and response generation be performed by one speech recognition model temporarily selected from the plurality of speech recognition models, determining a speech recognition model, on the basis of the result of speech recognition with respect to the utterance of the subject by a speech recognition server, that is estimated as being most suitable among the plurality of speech recognition models held by the spoken dialog device, and when the speech recognition model estimated as most suitable and the temporarily selected one speech recognition model are different, indicating to the spoken dialog device that speech recognition and response generation be performed after switching to the speech recognition model estimated as most suitable.SELECTED DRAWING: Figure 1 【課題】応答までに要する時間を短縮し、かつ、対話の不自然さを軽減することが可能になる、音声対話装置の制御装置を提供する。【解決手段】制御装置は、複数の音声認識モデルを有する音声対話装置を制御する演算部を備え、対象者との対話開始時に、音声対話装置に対し、まず、複数の音声認識モデルの中から暫定的に選択した１つの音声認識モデルにより音声認識及び応答生成を行うよう指示し、音声認識サーバにおける対象者の発話についての音声認識の結果に基づいて音声対話装置で保有する複数の音声認識モデルのうち最適と推定される音声認識モデルを決定し、当該最適と推定される音声認識モデルと暫定的に選択した１つの音声認識モデルとが相違する場合には、音声対話装置に対し、最適と推定される音声認識モデルに切替えして音声認識及び応答生成を行うよう指示する。【選択図】図１