Fictoroberta

🚀 FictoRoBERTa: A Model for Writing Style Prediction

This project fine-tunes the RoBERTa base model on the Brown corpus to predict whether a text is written in a non - fiction straightforward style or a fictional descriptive style. Note that here, "fiction" and "non - fiction" only refer to the writing style, not the factual correctness of the text content. For more details, please refer to this blog and the referenced paper: https://bekushal.medium.com/fictometer-a-simple-and-explainable-algorithm-for-sentiment-analysis-31186d2a8c7e

🚀 Quick Start

Prerequisites

Make sure you have the transformers library installed. If not, you can install it using pip install transformers.

Usage Example

from transformers import AutoModelForSequenceClassification, AutoTokenizer

# Load the model
model_name = "bekushal/FictoRoBERTa"

model = AutoModelForSequenceClassification.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)

# Example input text
input_text = "It is a sunny day with a nice wind blowing and I am feeling very happy."

# Preprocess the input
inputs = tokenizer(input_text, return_tensors="pt", truncation=True, padding=True)

# Perform inference
outputs = model(**inputs)

# Get predicted class probabilities
predicted_probabilities = outputs.logits.softmax(dim=-1)

# Get predicted class label
predicted_label = predicted_probabilities.argmax().item()

# Convert predicted label to human-readable format
predicted_class = "fiction" if predicted_label == 1 else "non-fiction"

# Display results
print("Predicted class:", predicted_class)
print("Predicted class probabilities [non-fiction, fiction]:", predicted_probabilities)

📄 License

This project is licensed under the Apache - 2.0 license.