Sentiment-Analysis

Data Preprocessing:

The data undergoes preprocessing to enhance its quality:

• Removal of special characters.
• Elimination of single characters and spaces.
• Conversion of all words to lowercase.

LSTM Neural Network Model:

The project utilizes an LSTM neural network model with pre-trained word embeddings based on Glove embeddings (glove.6B.50d) from a large Twitter US Airline database.

Steps for Result Improvement and Analysis:

1. Integration of Dropout layers to prevent overfitting.
2. Adjustment of the learning rate from 0.1 to 0.001 for fine-tuning.
3. Set the batch size to 200 for efficient training.
4. Incorporation of validation data for performance assessment.

These measures collectively optimize the model's performance and facilitate comprehensive result analysis.