Constructor (SYNC path).
Users call: new OfflineRecognizer(config)
Async factory calls this with an internal descriptor.
Create a new OfflineStream bound to this recognizer.
The optional hotwords argument enables contextual biasing for this stream only. Hotwords are supported only by transducer models decoded with decodingMethod 'modified_beam_search'. Separate multiple phrases with '/', and optionally append a per-phrase boosting score, e.g. 'PHOEBE :2.0/DON QUIXOTE'. When modelConfig.modelingUnit and modelConfig.bpeVocab are set, phrases are given as normal words; otherwise each phrase must be a sequence of tokens from tokens.txt separated by spaces. See also https://k2-fsa.github.io/sherpa/onnx/hotwords/index.html
Optionalhotwords: string
Optional hotwords for this stream.
Decode an offline stream asynchronously (non-blocking).
Get recognition result for a stream.
StaticcreateCreate an OfflineRecognizer asynchronously (non-blocking).
OfflineRecognizer wraps the native offline recognizer.