DeepgramNova2<T>

Models & Types · AiDotNet.SpeechRecognition.ProprietaryAPI

A Deepgram Nova-2 stand-in: a Transformer encoder with a CTC layer, since Deepgram has disclosed only that Nova-2 is a Transformer-based end-to-end model.

Inheritance: NeuralNetworkBase<T> → AudioNeuralNetworkBase<T> → DeepgramNova2<T>

Implements: ISpeechRecognizer<T>

For Beginners

This model turns 16 kHz audio into an 80-band log-mel spectrogram, shortens it 4x with two convolutions and encodes it with Transformer layers. A linear layer gives every frame a probability for each token; collapsing repeats and blanks gives the transcript.

How It Works

This is a stand-in, not Deepgram’s production model. Deepgram describes Nova-2 only as a proprietary Transformer-based end-to-end recognizer trained in a two-stage curriculum; it has not said whether it is an encoder-decoder, CTC or transducer model, nor given its size. The stand-in is the plainest end-to-end Transformer recognizer: 4x convolution subsampling, sinusoidal positions added to the scaled input, 12 pre-norm Transformer encoder layers of width 512 with 8 heads and 2,048 ReLU feed-forward units (the base Transformer of Vaswani et al. 2017), and a CTC layer (Graves et al. 2006) decoded greedily. SpecAugment masks the features in training.

Example

var architecture = new NeuralNetworkArchitecture<float>(
    inputType: InputType.OneDimensional,
    taskType: NeuralNetworkTaskType.SpeechRecognition,
    inputSize: 16000, outputSize: 34);
var model = new DeepgramNova2<float>(architecture);
var text = model.Transcribe(waveform).Text;

Constructors

Constructor Summary
DeepgramNova2(NeuralNetworkArchitecture<T> architecture, DeepgramNova2Options options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer)
DeepgramNova2(NeuralNetworkArchitecture<T> architecture, string modelPath, DeepgramNova2Options options)

Properties

Property Summary
IReadOnlyList<string> SupportedLanguages { get; }
bool SupportsStreaming { get; }
bool SupportsWordTimestamps { get; }

Methods

Method Summary
IReadOnlyDictionary<string, T> DetectLanguageProbabilities(Tensor<T> audio)
IStreamingTranscriptionSession<T> StartStreamingSession(string language)
ModelMetadata<T> GetModelMetadata()
Task<TranscriptionResult<T>> TranscribeAsync(Tensor<T> audio, string language, bool? includeTimestamps, CancellationToken cancellationToken)
Tensor<T> ForwardForTraining(Tensor<T> input)
TranscriptionResult<T> Transcribe(Tensor<T> audio, string language, bool? includeTimestamps) Transcribes audio: the encoder and the CTC layer with greedy decoding.
string DetectLanguage(Tensor<T> audio)