A computer system processes speech. Speech of a user is received from an audio sensor. An interruption in the speech of the user is determined and the speech of the user is analyzed to determine a context of the speech. One or more language data sources are analyzed to generate a suggestion corresponding to the interruption and based on the context of the speech. The suggestion is presented to the user. Embodiments of the present invention further include a method and program product for prompting a user in substantially the same manner described above.