Module keenasr-react-native

KeenASR React Native Plugin

Offline speech recognition SDK for React Native, wrapping the native KeenASR SDK for iOS and Android.

Quick Start

import KeenASR, { onFinalResponse, onPartialResult } from 'keenasr-react-native';

// 1. Initialize the SDK
await KeenASR.initialize('keenA1m-nnet3chain-en-us');

// 2. Create a decoding graph
await KeenASR.createDecodingGraphFromPhrases('myGraph', [
  'hello world', 'goodbye world',
]);

// 3. Prepare for listening
await KeenASR.prepareForListening('myGraph');

// 4. Subscribe to results
const unsub = onFinalResponse((response) => {
  console.log(response.asrResult.text);
  response.release();
});

// 5. Start listening
await KeenASR.startListening();
Copy

Key Concepts

• ASR Bundle: Language model and acoustic model files required for recognition.
• Decoding Graph: A compiled recognition grammar built from phrases. Must be created before recognition can start.
• Recognizer Lifecycle: Initialize → Create Graph → Prepare → Listen → Process. See RecognizerState for the state machine.
• Response Lifecycle: Responses hold native memory and must be released via release() when no longer needed.
• VAD (Voice Activity Detection): Controls how the recognizer detects the start and end of speech. See VadParameter.
• GoP (Goodness of Pronunciation): Phone-level pronunciation scoring. Enable via computeGoP: true in prepareForListening.