| 摘要 |
Conventional server-client speech recognition devices have a problem in that the burden on users is large because the user must speak from the beginning when speech recognition results from either a server or a client do not come back. A speech recognition device according to the present invention: sends input speech to a server; receives first speech recognition results, which are results by the server of speech recognition of the input speech that has been sent; carries out speech recognition on the input speech and obtains second speech recognition results; refers to utterance rules that express the constitution of speech elements in the input speech and determines utterance rules matching the second speech recognition results; determines a speech recognition state that shows utterance elements for which the speech recognition results are not obtained on the basis of a correlative relationship between the presence or absence of the first speech recognition results and the presence or absence of the second speech recognition results, and the presence or absence of utterance elements constituting the utterance rules; generates a response phrase corresponding to the speech recognition state determined and querying utterance elements for which the speech recognition results are not obtained; and outputs the response phrase. |
