Dataset
The Matthew gospel from King James Version (hard to segment and penetrate)
For the dataset composed of ancient english the training is not so successful, generating unpredictable result for too opaque language.
Shifting from King James Version to New International Version (NIV)
NIV's language is more straight forward.
Result
The loss function from TensorBoard, go converge quickly after several epochs
The TensorGraph, contains of two segmentation network and one RNN.
Live Demo!
Open the terminal
Frontend Framework Plan
React.js?
Comments