ASR and NVIDIA

 NVIDIA's hardware and software solutions for automated speech recognition are great. Typically, #ASR projects involve real-world sequence learning tasks, predicting sequences of labels from noisy, input data. I started ASR a long time ago with #HMM and modeling silence, noise signals, and acoustic signals transcribed into words or sub-word units. More recently, I worked with the #NeMo conversational toolkit based on the #PyTorch environment. The experimentational capabilities and modularity are excellent. Today, the pretrained models are game-changing resources. Also, the acoustic models based on #Jasper and #QuarzNet are good choices for training or using pre-trained acoustic models.  

#naturallanguageprocessing, #NLP, #ConversationalAI, #Speech, #speechrecognition, #speechtotext #AI #nvidia, #chatbots #deeplearning #machinelearning #Convolutional #convolutionalNeuralAcousticModel #convolutionalneuralnetworks #artificialintelligence

 

https://developer.nvidia.com/blog/develop-smaller-speech-recognition-models-with-nvidias-nemo-framework