Table 5 Results of different models on quora and custom-built dataset.
From: Paraphrase detection for Urdu language text using fine-tune BiLSTM framework
Model | Accuracy-quora (%) | Exe-time (s) | Accuracy-custom (%) | Exe-time (s) |
|---|---|---|---|---|
CNN | 87.03 | 23.45 | 83.43 | 21.04 |
LSTM | 90.19 | 17.89 | 88.09 | 15.31 |
CNN-LSTM | 91.48 | 19.21 | 90.67 | 18.58 |
BERT | 89.72 | 22.16 | 87.35 | 23.42 |
Proposed BiLSTM | 95.34 | 19.11 | 94.14 | 16.87 |
