You can deploy this model and start earning money today!

vamshi0310/
finetuned-disfluency-correction

Name: finetuned-disfluency-correction
Availability: InStock

safetensors

Views

No views yet

Fine-tuned Disfluency Correction (T5-small)

This is a fine-tuned T5-small model for correcting disfluencies in English sentences.
It converts sentences like "I I um want to to go to the the market" into fluent sentences: "I want to go to the market".

Dataset

The model was trained on a custom dataset of disfluent → fluent sentence pairs.

Evaluation Metrics (Test Set)

TER: 1.38
ChrF++: 99.26

Try the Model

You can try it live here: Disfluency Correction Demo

Usage

python

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

# Load the model and tokenizer from Hugging Face Hub
tokenizer = AutoTokenizer.from_pretrained("vamshi0310/finetuned-disfluency-correction")
model = AutoModelForSeq2SeqLM.from_pretrained("vamshi0310/finetuned-disfluency-correction")

# Example disfluent sentence
sentence = "I I um want to to go to the the market"

# Tokenize and generate corrected sentence
inputs = tokenizer(f"correct disfluency: {sentence}", return_tensors="pt")
outputs = model.generate(**inputs, max_length=256)
corrected = tokenizer.decode(outputs[0], skip_special_tokens=True)

print("Original:", sentence)
print("Corrected:", corrected)