|
---
|
|
license: apache-2.0
|
|
---
|
|
<h2>Re-Punctuate:</h2>
|
|
|
|
Re-Punctuate is a T5 model that attempts to correct Capitalization and Punctuations in the sentences.
|
|
|
|
<h3>DataSet:</h3>
|
|
|
|
DialogSum dataset (115056 Records) was used to fine-tune the model for Punctuation and Capitalization correction.
|
|
|
|
<h3>Usage:</h3>
|
|
|
|
<pre>
|
|
|
|
from transformers import T5Tokenizer, TFT5ForConditionalGeneration
|
|
tokenizer = T5Tokenizer.from_pretrained('SJ-Ray/Re-Punctuate/')
|
|
model = TFT5ForConditionalGeneration.from_pretrained('SJ-Ray/Re-Punctuate/')
|
|
input_text = 'the story of this brave brilliant athlete whose very being was questioned so publicly is one that still captures the imagination'
|
|
inputs = tokenizer.encode("punctuate: " + input_text, return_tensors="tf")
|
|
result = model.generate(inputs)
|
|
decoded_output = tokenizer.decode(result[0], skip_special_tokens=True)
|
|
print(decoded_output)
|
|
|
|
</pre>
|
|
<h4> Example: </h4>
|
|
<b>Input:</b> the story of this brave brilliant athlete whose very being was questioned so publicly is one that still captures the imagination <br>
|
|
<b>Output:</b> The story of this brave, brilliant athlete, whose very being was questioned so publicly, is one that still captures the imagination.
|
|
|
