Thai Word Segmentation with a Brain-Inspired Sparse Distributed Representations Learning Memory

Table 5

Word segmentation performance with noise.

Method	Train data	Test data	Precision (%)	Recall (%)	F1-score (%)	IoU	Time (s/word)
Deepcut	Best	Noise 1%	92.99	93.50	93.25	0.87	0.002349547
THSDR	Best	Noise 1%	97.89	98.00	97.95	0.96	0.407567715
Deepcut	Best	Noise 3%	88.77	89.22	88.99	0.80	0.002298372
THSDR	Best	Noise 3%	93.44	93.70	93.57	0.88	0.408186889
Deepcut	Best	Noise 5%	84.62	85.00	84.81	0.74	0.002310482
THSDR	Best	Noise 5%	90.50	91.02	90.76	0.83	0.408712781
Deepcut	Best	Noise 10%	74.96	75.03	74.99	0.60	0.002285629
THSDR	Best	Noise 10%	79.29	82.00	80.62	0.68	0.426231586