A streamlined set of parameters for use for Adm and Reco situations to configure speech services.

Base class for Conversational Understanding Sessions

A client using speech recognition with data (for example from a file or audio source). The data is broken up into buffers and each buffer is sent to the Speech Recognition Service. No modification is done to the buffers, so the user can apply their Silence Detection. Returns only text recognition results. The audio must be PCM, mono, 16-bit sample, with sample rate of 8000 Hz or 16000 Hz.

A client using speech recognition with data (for example from a file or audio source). The data is broken up into buffers and each buffer is sent to the Speech Recognition Service. No modification is done to the buffers, so the user can apply their Silence Detection. Return both text recognition results as well as structured intent results in JSON form from the LUIS (see https://LUIS.ai) service.

A generic Intent structure

Event data for when the microphone has been started and stopped.

A client using speech recognition from the microphone. The microphone is turned on and data from the microphone is sent to the Speech Recognition Service. A built in Silence Detector is applied to the microphone data before it is sent to the recognition service. Returns only text recognition results.

A client using speech recognition from the microphone. The microphone is turned on and data from the microphone is sent to the Speech Recognition Service. A built in Silence Detector is applied to the microphone data before it is sent to the recognition service. Return both text recognition results as well as structured intent results in JSON form from the LUIS (see https://LUIS.ai) service.

SpeechRecognitionService partial response event data

A streamlined set of parameters for use for Adm and Reco situations to configure speech services.

In ShortPhrase mode the client gets one final multiple n-best choice result and in LongDictation mode the client will receive multiple final results (each with just a single choice), based on where the server thinks sentence pauses are.

A Confidence value for the recognized text and the recognized text with a set of processing types applied to it. This is one element of an n-best set.

The SpeechAudioFormat contains information about how the audio data was recorded and stored including the type of compression used, number of channels, sample rate, bits per sample and other attributes.

SpeechRecognitionService error event data

Event data for when a Speech Recognition has finished, the recognized text has been parsed with LUIS for intent and entities, and the structured JSON result is available.

Use Microsoft's CognitiveServices to work with the Speech Recognition Service. This factory can be used to create an object with which to make a recognition request to the Speech Recognition Service. There are four types of objects that this factory can create.

(1) A DataRecognitionClient -- for speech recognition with data (for example from a file or audio source). The data is broken up into buffers and each buffer is sent to the Speech Recognition Service. No modification is done to the buffers, so the user can apply their own Silence Detection. Returns only text recognition results. The audio must be PCM, mono, 16-bit sample, with sample rate of 8000 Hz or 16000 Hz.

(2) A DataRecognitionClientWithIntent -- for speech recognition with data (for example from a file or audio source). The data is broken up into buffers and each buffer is sent to the Speech Recognition Service. No modification is done to the buffers, so the user can apply their own Silence Detection. Returns both text recognition results as well as structured intent results in JSON form from the LUIS (see https://LUIS.ai) service. The audio must be PCM, mono, 16-bit sample, with sample rate of 8000 Hz or 16000 Hz.

(3) A MicrophoneRecognitionClient -- for speech recognition from the microphone. The microphone is turned on and data from the microphone is sent to the Speech Recognition Service. A built in Silence Detector is applied to the microphone data before it is sent to the recognition service. Returns only text recognition results.

(4) A MicrophoneRecognitionClientWithIntent -- for speech recognition from the microphone. The microphone is turned on and data from the microphone is sent to the Speech Recognition Service. A built in Silence Detector is applied to the microphone data before it is sent to the recognition service. Returns both text recognition results as well as structured intent results in JSON form from the LUIS (see https://LUIS.ai) service.

SpeechRecognitionService response event data

Indicates the type of compression applied to the audio data

Indicates the Confidence level ascribed by the Speech Recognition Server to each n-best result. This is different than the EngineConfidence. The EngineConfidence is a float from 0.0 to 1.0 from representing the confidence the base recognition algorithm on the correctness of the recognized text. A value of 1 would mean 100% confidence. The Confidence level is based on a quantization of the EngineConfidence and also takes into account processing after the base engine that affects the confidence, like ITN.

Status on how the recognition was produced. For example, was it produced as a result of successful reco, or as a result of canceling the connection, etc...

Error codes that can be returned from the speech client that aren't recognition based errors.

Determines what type of request to make to the Speech Recognition Server.

In ShortPhrase mode, an utterance may only up to 15 sec long, as data is sent to the server, the client will receive multiple partial results and one final multiple n-best choice result.

In LongDictation mode, an utterance may be only up to 2 minutes long, as data is sent to the server, the client will receive multiple partial results and multiple final results, based on where the server thinks sentence pauses are.