distilbert-sentimentx-twitter

SentimentX-DistilBERT

This model is a fine-tuned version of distilbert-base-uncased trained on the Sentiment140 dataset. It is designed to classify the sentiment of short-form text (specifically tweets) into two categories: Positive and Negative.

Model Description

• Developed by: [Kalyan Sai]
• Model type: Transformer-based Text Classification
• Language(s) (NLP): English
• License: Apache 2.0
• Finetuned from model: distilbert-base-uncased

Intended Uses & Limitations

Intended Use

This model is intended for real-time sentiment monitoring of social media feeds, specifically Twitter (X). It is optimized for high-throughput inference environments.

Limitations

• The model was trained on a binary dataset (Positive/Negative). It may not perform accurately on neutral text.
• Performance may decrease on text with heavy use of modern slang or emojis not present in the 2009 Sentiment140 dataset.

Training and Evaluation Data

The model was trained on a sample of 100,000 tweets from the Sentiment140 dataset, which contains 1.6 million tweets automatically labeled (0 = negative, 4 = positive) based on emoticons.

Training Procedure

• Hardware: Trained on Kaggle NVIDIA Tesla T4 GPU.
• Optimizer: AdamW
• Learning Rate: Default (5e-5)
• Epochs: 2
• Mixed Precision: FP16 enabled for acceleration.

Evaluation Results

The model achieved the following results on the evaluation set:

• Accuracy: 82.44%
• F1-Score: 82.50%
• Precision: 83.01%
• Recall: 81.99%

How to Get Started with the Model

Use the code below to get started with the model locally:

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

tokenizer = AutoTokenizer.from_pretrained("mr-checker/sentimentX-distilbert")
model = AutoModelForSequenceClassification.from_pretrained("mr-checker/sentimentX-distilbert")

text = "SentimentX is working perfectly!"
inputs = tokenizer(text, return_tensors="pt")

with torch.no_grad():
    logits = model(**inputs).logits

predicted_class_id = logits.argmax().item()
# 0: Negative, 1: Positive