This repository contains the code and resources for SentimentBERT-AIWriting, a fine-tuned version of bert-base-uncased for sentiment classification, tailored for AI-assisted argumentative writing. It classifies text into three categories: positive, negative, and neutral.
SentimentBERT-AIWriting extends the original BERT (Bidirectional Encoder Representations from Transformers) capabilities to the task of sentiment classification. It was trained on a diverse dataset of statements collected from various domains to ensure robustness and accuracy across different contexts.
The SentimentBERT-AIWriting model is designed to assist in understanding the sentiment of texts. This can be particularly useful for platforms requiring an understanding of user sentiment, such as customer feedback analysis, social media monitoring, and enhancing AI writing tools.
You can use this model with the Hugging Face transformers library. Below is an example code snippet:
from transformers import BertTokenizer, BertForSequenceClassification
tokenizer = BertTokenizer.from_pretrained('MidhunKanadan/SentimentBERT-AIWriting')
model = BertForSequenceClassification.from_pretrained('MidhunKanadan/SentimentBERT-AIWriting')
text = "Your text goes here"
inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True, max_length=128)
outputs = model(**inputs)
logits = outputs.logits
predictions = logits.argmax(-1)
labels = ['negative', 'neutral', 'positive']
predicted_label = labels[predictions.item()]
print(f"Text: {text}
predicted_label: {predicted_label}
")Here are some example statements and their corresponding sentiment predictions by the SentimentBERT-AIWriting model:
Positive
- Statement: "Despite initial skepticism, the new employee's contributions have been remarkable!"
- Predicted Label:
positive
Negative
- Statement: "Nuclear energy can be a very efficient power source, but at the same time, it poses significant risks."
- Predicted Label:
negative
Neutral
- Statement: "The documentary provides an overview of the event."
- Predicted Label:
neutral
These examples demonstrate how SentimentBERT-AIWriting can effectively classify the sentiment of various statements.
While SentimentBERT-AIWriting is trained on a diverse dataset, no model is immune from bias. The model's predictions might still be influenced by inherent biases in the training data. It's important to consider this when interpreting the model's output, especially for sensitive applications.
The code used for fine-tuning the SentimentBERT-AIWriting model can be found in the finetuning_script.py file.
I welcome contributions to this model! You can suggest improvements or report issues by opening an issue on this repository.
If you find this model useful for your projects or research, feel free to cite it and provide feedback on its performance.