Sentiment Analysis Using LSTM

Inspiration

The increasing reliance on social media for customer feedback inspired me to analyze the "Twitter US Airline Sentiment" dataset. Understanding customer sentiment can significantly improve airline services and brand reputation.

What it does

The project analyzes 14,640 tweets related to U.S. airlines to determine the sentiment expressed by users. Using LSTM models, it identifies positive, negative, or neutral sentiments, along with the reasons behind them.

How I built it

I built the project using Python, focusing on data preprocessing, feature extraction, and sentiment analysis. LSTM models, both base and stacked, were implemented to capture the temporal dependencies in tweet data for accurate sentiment prediction.

Challenges I ran into

Key challenges included handling noisy data, selecting the right model architecture for optimal performance, and ensuring the models captured the nuanced expressions in tweets without overfitting.

Accomplishments that I'm proud of

I'm proud of successfully implementing LSTM models that accurately predict tweet sentiment, achieving a high confidence level. The project also provides valuable insights into common issues faced by airline customers.

What I learned

I learned advanced techniques in natural language processing (NLP), including the importance of data preprocessing and the effectiveness of LSTM models in handling sequential data for sentiment analysis.

What's next for Sentiment Analysis Using LSTM

Next, I plan to expand the analysis to include real-time sentiment tracking, improve the model's accuracy with more advanced architectures, and explore transfer learning to apply the model to other industries.

Built With

• keras
• lstm
• machine-learning
• python
• tweetfeel-twitter-sentiment