Recognition Modes

Different scenarios of speech recognition can require subtle differences in behavior from a speech recognizer. Although starting the recognition process is standard in all cases, the means of stopping the process and the return of results may differ.

The following recognition modes are available.

Recognition Mode	Description
Automatic	Recognizes in telephony scenarios. Default if no mode is specified.
Multiple	Recognizes in dictation scenarios.
Single	Recognizes in multimodal scenarios.

Remarks

The mode attribute on the listen element distinguishes the following three modes of recognition: automatic, single and multiple. These are distinguished by how and when the speech recognizer returns results. When results are returned, an onreco event occurs.

For example, using tap-and-talk in a multimodal application, users control the period of time they wish to speak to the device by tapping and holding a form field. The browser uses a GUI event, for example, pen up, to control when recognition should stop and return results. However, in voice-only scenarios, such as telephony or hands-free, users have no direct control over the browser, and the recognition platform determines when to stop recognition and return results (typically once a complete path through the grammar has been recognized and the endsilence value has been exceeded. Further, dictation and other scenarios where intermediate results may need to be returned before recognition is stopped (here called "open microphone") not only require an explicit stop but also need to return multiple recognition results to the application before the recognition process is stopped.

Generally, automatic mode is more useful in telephony profiles, multiple mode in all kinds of dictation and possibly hotword scenarios, and single mode in multimodal profiles.

See Also

Recognition Mode: Automatic |  Recognition Mode: Multiple | Recognition Mode: Single