Emotions Dataset – a curated collection of English Twitter messages annotated with six primary emotions: anger, fear, joy, love, sadness, and surprise. This dataset serves as a valuable resource for analyzing emotional expressions in short-form text on social media platforms.

Each entry in this dataset includes:

• text: A string feature representing the content of a Twitter message.
• label: A classification label indicating the predominant emotion, with values ranging from 0 to 5 corresponding to sadness, joy, love, anger, fear, and surprise respectively.

In this project, we employed various machine learning models to classify emotions from Twitter messages. Below are the models utilized and their descriptions:

• LSTM with PyTorch

  • Description: Long Short-Term Memory (LSTM) networks implemented using PyTorch to capture long-range dependencies in text sequences.
  • Usage: LSTM models were trained on the Emotions dataset to understand sequential patterns and classify tweets into one of six emotion categories.

• DeBERTa v3 with Hugging Face Transformers

  • Description: State-of-the-art transformer-based model from Hugging Face Transformers library, specifically DeBERTa v3, fine-tuned on the emotions classification task.
  • Integration: Leveraged pre-trained DeBERTa v3 model and fine-tuned it on the Emotions dataset to utilize bidirectional contextual embeddings for accurate emotion prediction.

• XLNet with Hugging Face Transformers

  • Description: XLNet is another transformer-based model that overcomes limitations of traditional transformers by leveraging permutation language modeling.
  • Advantages: XLNet captures bidirectional dependencies more effectively than traditional transformers, offering enhanced understanding of context in text sequences.
  • Implementation: Integrated XLNet from Hugging Face Transformers library, fine-tuned on the Emotions dataset to explore its effectiveness in emotion classification tasks.

Each model was evaluated using standard metrics such as accuracy, precision, recall, and F1-score to assess its performance in predicting emotions from tweets.