A Brazilian Portuguese Real-Time Voice Recognition to deal with sensitive data
- 0 Collaborators
Voice recognition is generally performed by cloud providers centrally, and it is not suitable for financial institutions, because sensitive user data cannot be provided openly to third parties. This work proposes a simple voice-bot for Brazilian Portuguese real-time voice recognition. ...learn more
Project status: Under Development
Intel Technologies
AI DevCloud / Xeon,
Intel Python,
MKL,
Intel CPU,
Ngraph
Overview / Usage
This work proposes a simple voice-bot for Brazilian Portuguese real-time voice recognition, definite purpose, specific vocabulary and in a secure way without exposure of sensitive data for use on mobile devices.
Methodology / Approach
Deep learning, based on Keras, Tensorflow, and Neon.
Benchmark with DeepSpeech (https://github.com/NervanaSystems/deepspeech).
Technologies Used
Python, Keras, Tensorflow, Nervana-Neon, Nervana-Aeon.