Update README.md
Browse files
README.md
CHANGED
|
@@ -91,6 +91,26 @@ These substitutions represent phonetic similarities and common mistakes made by
|
|
| 91 |
|
| 92 |
To use the model for spell correction:
|
| 93 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 94 |
|
| 95 |
## License
|
| 96 |
|
|
|
|
| 91 |
|
| 92 |
To use the model for spell correction:
|
| 93 |
|
| 94 |
+
```python
|
| 95 |
+
import torch
|
| 96 |
+
from transformers import MT5ForConditionalGeneration, MT5Tokenizer
|
| 97 |
+
|
| 98 |
+
# Load the tokenizer and model
|
| 99 |
+
tokenizer = MT5Tokenizer.from_pretrained('LocalDoc/azerbaijani_spell_corrector')
|
| 100 |
+
model = MT5ForConditionalGeneration.from_pretrained('LocalDoc/azerbaijani_spell_corrector')
|
| 101 |
+
|
| 102 |
+
# Function to correct sentences
|
| 103 |
+
def correct_sentence(sentence):
|
| 104 |
+
input_text = "correct: " + sentence
|
| 105 |
+
input_ids = tokenizer.encode(input_text, return_tensors='pt', max_length=128, truncation=True)
|
| 106 |
+
outputs = model.generate(input_ids=input_ids, max_length=128, num_beams=5, early_stopping=True)
|
| 107 |
+
corrected_sentence = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
| 108 |
+
return corrected_sentence
|
| 109 |
+
|
| 110 |
+
# Example usage
|
| 111 |
+
incorrect_sentence = "Pul dogru adamlarda deyil"
|
| 112 |
+
print(correct_sentence(incorrect_sentence))
|
| 113 |
+
```
|
| 114 |
|
| 115 |
## License
|
| 116 |
|