Recognizer Methods
The recognizer resource is the half of an MRCP session responsible for turning inbound audio into a result—loading grammars, running recognition, managing the input timers that decide when a caller has finished speaking, and returning the interpreted outcome. The methods below are the control messages a client sends to drive that resource through its lifecycle.
These methods are defined in IETF RFC 4463 (MRCPv1) and IETF RFC 6787 (MRCPv2). The protocol evolved between the two versions, so a method available in one is not guaranteed to exist—or to carry the same name—in the other. The table records, for each method, whether it is part of the MRCPv1 specification, the MRCPv2 specification, and whether it is supported on Capacity Private Cloud. Select any method name for its full reference, including the headers it accepts and the responses it returns.
SET-PARAMS and GET-PARAMS are generic methods shared by every MRCP resource rather than recognizer-specific ones; they are listed here for completeness.
Method support by MRCP version
Method | MRCP V1 | MRCP V2 | Platform Support |
✓ | ✓ | ✓ | |
✓ | ✓ | ✓ | |
✓ | ✓ | ✓ | |
✓ | ✓ | ✓ | |
* | ✓ | ✓ | |
✓ | ✓ | ✓ | |
✓ | — | ✓ | |
— | ✓ | ✓ | |
✓ | ✓ | ✓ |
✓ = defined or supported | — = not part of that specification | * = see note below
Version notes
INTERPRET. The INTERPRET method—which runs semantic interpretation against supplied text rather than live audio—was introduced with MRCPv2 and is not part of the MRCPv1 specification. Capacity Private Cloud extends its MRCP implementation to support INTERPRET under both MRCPv1 and MRCPv2, which is useful for clients standardized on the older protocol that still need interpretation-only requests.
Start-timer rename. The method named RECOGNITION-START-TIMERS in MRCPv1 was renamed START-INPUT-TIMERS in MRCPv2. The two perform the same role—arming the no-input and recognition timers after the prompt has played—so choose the name that matches the protocol version you are running.