In this section, the performance-ratio of the proposed system strategies are visualized with respect to the accuracy ratios attained over the tables mentioned over the previous section. In this perception, the dataset outcome ratio of SANAD is clearly estimated via many attributes such as: Category, Training data, Testing Data, Total estimation and per category basis. These all will be estimated under three different dataset categories of SANAD such as: Arabiya, Khaleej and Akhbarona. The following figure, Fig. 4 illustrates the SANAD Classification dataset outcome ratio of different datasets. The accuracy ratio and the outcome estimation of Arabiya dataset is deviated under several attributes with different values such as the attribute Category is bounded with 5, training data is bounded up with 16,650 data, testing data is bounded up with 1,850 data, Total estimations are bounded with 18,500 and per category is bounded with 3,700. The accuracy ratio and the outcome estimation of Khaleej dataset is deviated under several attributes with different values such as the attribute Category is bounded with 7, training data is bounded up with 40,950 data, testing data is bounded up with 4,550 data, Total estimations are bounded with 45,500 and per category is bounded with 6,500. The accuracy ratio and the outcome estimation of Akhbarona dataset is deviated under several attributes with different values such as the attribute Category is bounded with 7, training data is bounded up with 42,210 data, testing data is bounded up with 4,690 data, Total estimations are bounded with 46,900 and per category is bounded with 6,700.
|
Single-label Dataset
|
|
AR-5
|
KH-7
|
AB-7
|
RT-40
|
BIGRU
|
97.41
|
96.46
|
92.23
|
57.25
|
BILSTM
|
96.43
|
95.05
|
90.14
|
57.70
|
CGRU
|
97.19
|
96.86
|
94.00
|
62.99
|
CLSTM
|
96.97
|
96.59
|
92.66
|
61.46
|
CNN
|
95.62
|
96.33
|
92.72
|
64.24
|
GRU
|
96.76
|
96.04
|
89.56
|
52.51
|
HANGRU
|
96.00
|
96.66
|
92.95
|
59.61
|
HANLSTM
|
96.38
|
96.55
|
92.21
|
59.44
|
LSTM
|
96.54
|
94.09
|
90.29
|
56.67
|
Average
|
96.59
|
94.07
|
91.86
|
59.10
|
Datasets Name
|
DL Model
|
Arabiya
|
Khaleej
|
Akhbarona
|
Arabiya
|
BIGRU
|
97.41
|
80.31
|
83.11
|
BILSTM
|
96.43
|
76.77
|
86.25
|
CGRU
|
97.19
|
83.05
|
85.11
|
CLSTM
|
96.97
|
80.03
|
85.53
|
CNN
|
95.62
|
71.85
|
83.32
|
GRU
|
96.76
|
77.63
|
83.68
|
HANGRU
|
96.00
|
76.58
|
84.51
|
HANLSTM
|
96.38
|
75.48
|
84.39
|
LSTM
|
96.54
|
80.00
|
85.23
|
Khaleej
|
BIGRU
|
83.78
|
96.46
|
70.52
|
BILSTM
|
81.57
|
95.05
|
69.02
|
CGRU
|
76.86
|
96.86
|
65.71
|
CLSTM
|
77.14
|
96.59
|
69.53
|
CNN
|
76.65
|
96.33
|
68.30
|
GRU
|
79.89
|
96.04
|
69.62
|
HANGRU
|
77.62
|
96.66
|
66.61
|
HANLSTM
|
80.43
|
96.55
|
70.47
|
LSTM
|
78.92
|
94.09
|
68.10
|
Akhbarona
|
BIGRU
|
83.08
|
75.93
|
92.23
|
BILSTM
|
82.11
|
74.79
|
90.14
|
CGRU
|
86.65
|
77.76
|
94.00
|
CLSTM
|
87.24
|
75.60
|
92.66
|
CNN
|
82.97
|
77.78
|
92.72
|
GRU
|
74.54
|
70.35
|
89.56
|
HANGRU
|
86.05
|
77.41
|
92.95
|
HANLSTM
|
79.62
|
76.18
|
92.21
|
LSTM
|
82.00
|
73.96
|
90.29
|
The following figure, Fig. 5 shows the N-gram accuracy ranges of Arabiya dataset, in which the ranges of N-grams are illustrated from 4 to 14 and N-gram accuracies are measured properly with proper dataset ratio. The classification accuracy levels of Arabiya dataset are deviated under several N-gram ratios such as: 0.8789 is obtained on 4-gram, 0.9178 is obtained on 5gram, 0.927 is obtained on 6gram, 0.9389 is obtained on 7gram, 0.9541 is obtained on 8gram, 0.9541 is obtained on 9gram, 0.9605 is obtained on 10gram, 0.9638 is obtained on 11gram, 0.9665 is obtained on 12gram, 0.9649 is obtained on 13gram and 0.9551 is obtained on 14gram.
The following figure, Fig. 6 shows the N-gram accuracy ranges of Khaleej dataset, in which the ranges of N-grams are illustrated from 4 to 14 and N-gram accuracies are measured properly with proper dataset ratio. The classification accuracy levels of Khaleej dataset are deviated under several N-gram ratios such as: 0.95 is obtained on 4gram, 0.9391 is obtained on 5gram, 0.9423 is obtained on 6gram, 0.9457 is obtained on 7gram, 0.9589 is obtained on 8gram, 0.9386 is obtained on 9gram, no values obtained to 10gram, no values obtained to 11gram, 0.9694 is obtained on 12gram, 0.9549 is obtained on 13gram and 0.9523 is obtained on 14gram.
The following figure, Fig. 7 shows the N-gram accuracy ranges of Akhbarona dataset, in which the ranges of N-grams are illustrated from 4 to 14 and N-gram accuracies are measured properly with proper dataset ratio. The classification accuracy levels of Akhbarona Dataset are deviated under several N-gram ratios such as: 0.778 is obtained on 4gram, 0.813 is obtained on 5gram, 0.8267 is obtained on 6gram, no values are obtained on 7gram, no values are obtained on 8gram, no values are obtained on 9gram, 0.8554 is obtained on 10gram, 0.8563 is obtained on 11gram, 0.8512 is obtained on 12gram, 0.8409 is obtained on 13gram and no values are obtained on 14gram.
The following Fig. 8 shows the N-gram accuracy ranges of RTA dataset, in which the ranges of N-grams are illustrated from 4 to 14 and the N- gram accuracies are measured properly with dataset ration. The classification accuracy levels of RTA Dataset are deviated under several N-grams ratio such as 0.613 is obtained on 4gram, 0.650 is obtained on 5gram, 0.6648 is obtained on 6gram, 0.6646 is obtained on 7gram, 0.6609 is obtained on 8gram, 0.6638 is obtained on 9gram, 0.6619 is obtained on 10gram, 0.6597 is contained on 11gram, 0.6578 is obtained on 12gram, 0.6464 is obtained on 13gram, 0.6373 is obtained on 14gram.
The following Fig. 9 shows the N-gram accuracy ranges of ANT dataset, in which the ranges of N-grams are illustrated from 4 to 14 and the N- gram accuracies are measured properly with dataset ration. The classification accuracy levels of ANT Dataset are deviated under several N-grams ratio such as 0.6113 is obtained on 4gram, 0.7418 is obtained on 5gram, 0.8140 is obtained on 6gram, 0.8526 is obtained on 7gram, 0.8734 is obtained on 8gram, 0.8723 is obtained on 9gram, 0.8932 is obtained on 10gram, 0.9001 is contained on 11gram, 0.9248 is obtained on 12gram, 0.9080 is obtained on 13gram, 0.9189 is obtained on 14gram.
The following figure, Fig. 10 illustrates the dataset outcome ratio of different Arabic languages.
The following figure, Fig. 11 shows the classification accuracy ranges of different Arabic languages such as: Palestinian and Syrian with respect to Language Detection principles.
The following figure, Fig. 12 shows the classification accuracy ranges of different Arabic languages such as: Jordanian and Lebanese with respect to Language Detection principles.
The following figure, Fig. 13 (a) and (b) shows the classification accuracy ranges of three different Arabic languages such as: Palestinian, Jordanian and Syrian as well as Jordanian, Syrian and Lebanese with respect to Language Detection principles.
The following figure, Fig. 14 shows that the classification accuracy ranges of four different Arabic languages such as: Palestinian, Syrian, Jordanian and Lebanese with respect to Language Detection principles.
The following figure, Fig. 15 illustrates the PADIC dataset outcome ratio of two different Arabic languages.
The following figure, Fig. 16 shows that the classification accuracy ranges of 2 different Arabic languages such as: Palestinian and Syrian with respect to PADIC Dataset using Language Detection principles.
The following figure, Fig. 17 illustrates the MADAR dataset outcome ratio of four different Arabic languages.
The following figure, Fig. 18 (a) and (b) shows that the classification accuracy ranges of different Arabic languages such as: Classification Results of Palestinian and Syrian and Jordanian and Lebanese with respect to MADAR Dataset using Language Detection principles.
The following figure, Fig. 19 (a) and (b) shows that the classification accuracy ranges of 4 different Arabic languages such as: Palestinian-Jordanian-Syrian and Jordanian-Syrian-Lebanese using Language Detection principles.
Classification Accuracy Ratio (a) Palestinian, Jordanian and Syrian (b) Jordanian, Syrian and Lebanese
The following figure, Fig. 20 shows that the classification accuracy ranges of 4 different Arabic languages such as: Palestinian, Syrian, Jordanian and Lebanese using Language Detection principles.
The figure below represents the time elapse on training the nine model of the Arabiya dataset to achieve best accuracy.