This study addresses the challenge of sentiment and emotion classification in Twitter posts by leveraging distant supervision and deep learning. The authors collected a large-scale dataset of 17.5 million tweets, automatically labeled with Ekman's six basic emotions using emojis. They compared various conventional machine learning models and deep learning models, including transformer-based models, to establish baseline results. The experimental results and ablation analysis showed that a BiLSTM model with FastText pre-trained word embeddings and an attention mechanism outperformed other models, achieving an F1-score of 70.92% for sentiment classification and 54.85% for emotion detection. The study also explored the impact of dataset size, class imbalance, pre-trained word embeddings, and attention mechanisms on model performance. The findings highlight the effectiveness of distant supervision and deep learning in handling large-scale, real-world datasets for sentiment and emotion analysis.This study addresses the challenge of sentiment and emotion classification in Twitter posts by leveraging distant supervision and deep learning. The authors collected a large-scale dataset of 17.5 million tweets, automatically labeled with Ekman's six basic emotions using emojis. They compared various conventional machine learning models and deep learning models, including transformer-based models, to establish baseline results. The experimental results and ablation analysis showed that a BiLSTM model with FastText pre-trained word embeddings and an attention mechanism outperformed other models, achieving an F1-score of 70.92% for sentiment classification and 54.85% for emotion detection. The study also explored the impact of dataset size, class imbalance, pre-trained word embeddings, and attention mechanisms on model performance. The findings highlight the effectiveness of distant supervision and deep learning in handling large-scale, real-world datasets for sentiment and emotion analysis.