Monitoring text to speech output to effect control of barge-in

1. Speech apparatus comprising:

a speech input channel including a speech recogniser;

a speech output channel including a text-to-speech converter;

a barge-in prediction arrangement arranged to assess an overall quality of speech-form passages being produced by the text-to-speech converter from input text-form utterances whereby to produce barge-in prediction indications in correspondence to speech-form passages assessed as inadequate; and

a barge-in control for setting barge-in behaviour parameters for determining how the apparatus handles barge-in by a user during speech output by the apparatus, the barge-in control being arranged to be responsive to said barge-in prediction indications to adjust at least one said barge-in behaviour parameter for the corresponding speech-form passages assessed as inadequate,

the barge-in control being arranged to be responsive to the said indications to adjust at least one said barge-in behaviour parameter for periods corresponding to the most likely barge-in points.

2. Apparatus according to claim 1 , wherein the text-to-speech converter is arranged to generate, in the course of converting a text-form utterance into a speech-form passage, values of predetermined features that are indicative of the overall quality of the speech-form passage, the barge-in prediction arrangement comprising:

a classifier arranged to be responsive to the feature values generated by the text-to-speech converter to provide a confidence measure of the speech-form passage concerned; and

a comparator for comparing confidence measures produced by the classifier against one or more stored threshold values, in order to determine when a said barge-in prediction indication is to be produced.

3. Apparatus according to claim 1 , wherein the text-to-speech converter includes a concatenative speech generator which, in generating a speech-form passsage, produces an accumulated unit selection cost in respect of the speech units used to make up the speech-form passage; the barge-in prediction arrangement comprising a comparator for comparing the selection cost produced by the speech generator against one or more stored threshold values, in order to determine when a said barge-in prediction indication is to be produced.

4. Apparatus according to claim 1 , wherein the text-to-speech converter is arranged to be responsive to the generation of a said barge-in prediction indication to insert a pause in the corresponding speech-form passage.

5. Apparatus according to claim 1 , wherein the barge-in control is arranged to respond to said barge-in prediction indication to set a said behaviour parameter such that any detected noise in the speech input channel is treated as barge-in during the corresponding speech-form passage, the barge-in control being arranged subsequently to reset this parameter such that noise in the speech input channel is only treated as a barge-in after it has been recognized as speech.

6. A method of controlling barge-in in speech apparatus that has both a speech input channel including a speech recogniser and a speech output channel including a text-to-speech converter; the method comprising the steps of:

(a) automatically assessing overall quality of speech-form passages being produced by the text-to-speech converter from input text-form utterances whereby to produce barge-in prediction indications in correspondence to speech-form passages assessed as inadequate; and

(b) adjusting, for periods corresponding to the speech-form passages assessed as inadequate, as indicated by said barge-in prediction indications, at least one barge-in behaviour parameter that determines how the apparatus handles barge-in by a user during speech output by the apparatus.

7. A method according to claim 6 , wherein the text-to-speech converter generates, in the course of converting a text-form utterance into a speech-form passage, values of predetermined features that are indicative of the overall quality of the speech-form passage, step (a) including:

using a classifier responsive to the feature values generated by the text-to-speech converter to provide a confidence measure of the speech-form passage concerned; and

comparing said confidence measures against one or more stored threshold values, in order to determine when a said barge-in prediction indication is to be produced.

8. A method according to claim 6 , wherein the text-to-speech converter includes a concatenative speech generator which, in generating a speech-form passage, produces an accumulated unit selection cost in respect of the speech units used to make up the speech-form passage; step (a) including comparing the selection cost produced by the speech generator against one or more stored threshold values, in order to determine when a said barge-in prediction indication is to be produced.

9. A method according to claim 6 , comprising the further step of inserting, in response to the generation of a said barge-in prediction indication, a pause in the corresponding speech-form passage.

10. A method according to claim 6 , wherein in step (b) a said behaviour parameter is adjusted such that any detected noise in the speech input channel is treated as a barge-in during the corresponding speech-form passage, this parameter being subsequently re-adjusted such that noise in the speech input channel is only treated as barge-in after it has been recognized as speech.