Table 1.
Literature review on Urdu NER.
Fig 1.
Methodology of the proposed Urdu NER system.
Table 2.
Summary of the used datasets.
Table 3.
Summary of hyperparameter tuning.
Table 4.
Results with different deep learning configurations.
Fig 2.
Training and Validation loss for best-performing models (a) BiLSTM-GRU with Floret embedding on IJCNLP, (b) BiLSTM-GRU with Floret embedding on Jahangir, (c) BiLSTM-GRU with FastText embedding on MKPUCIT, and (d) BiLSTM-GRU with Floret embedding on UNER.
Table 5.
Performance comparison of the proposed approach with state-of-the-art on all benchmark datasets.
Fig 3.
Confusion Matrix of best-performing models (a) BiLSTM-GRU with Floret embedding on IJCNLP, (b) BiLSTM-GRU with Floret embedding on Jahangir, (c) BiLSTM-GRU with FastText embedding on MKPUCIT, and (d) BiLSTM-GRU with Floret embedding on UNER.