Error Correction

For tasks that require the creation of correct input strings by speech (such as dictation), the system must provide support for error correction. The support need not be entirely embebbed in the speech interface, but can make use of facilities provided by the underlying application (for example, a word processor).

• Provide an alternative input mode that overrides speech input.
• Provide positive methods for identifying the erroneous portions of an input string.
• For high-accuracy systems, provide a facility for choosing among alternative decodings.
• In speech-only systems, provide a dialog-based recovery procedure.