NokCut

Thai Word Segmentation using TCC + Bidirectional RNNs (PyTorch)

Credit code from A Beginner's Guide to Deep NLP with PyTorch - Dr. Prachya Boonkwan

Colab Notebook : https://colab.research.google.com/drive/1WS08VsjlZGAmCGsoI7AlRm-Do3zo-b-g

Install

pip install nokcut

Train by BEST I Corpus Training set. (90% training , 10% test)

ep 6
loss: 0.017879242024514966
f1 : 98.47012481095481

F1 From BEST I Corpus Test set

F-measure: 96.94929
Recall: 122271.00000/125850.00000 = 97.15614

Precision: 122271.00000/126387.00000 = 96.74333

Number of incorrect : 3579.00000 words

Mr. Wannaphong Phatthiyaphaibun wannaphong@kkumail.com