💬 Sentiment Analysis & NLP Pipeline

Applied to 15,000 Disneyland Reviews

Overview

Comprehensive NLP pipeline on 15,000 Disneyland park reviews covering sentiment classification, text summarization, named entity recognition, and n-gram analysis using BERT and traditional ML.

Dataset

• Source: Disneyland Reviews (Kaggle)
• Size: 15,000 reviews
• Task: Sentiment classification (1-5 star ratings)

Results

• BERT: 89% sentiment classification accuracy
• 4-7% accuracy improvement via T5 summarization preprocessing pipeline
• NER identifies locations, organizations, entities
• Word cloud and n-gram analysis of key themes

NLP Tasks

1. Text Preprocessing — contractions, tokenization, stopwords, lemmatization (NLTK + spaCy)
2. Exploratory Analysis — rating distribution, word clouds, n-gram analysis, NER
3. Sentiment Classification — BERT, Logistic Regression, SVM, Naive Bayes, TextBlob (TF-IDF)
4. Text Summarization — extractive (Summa) and abstractive (T5 transformer), ROUGE evaluation

Tech Stack

Python | BERT | T5 | PyTorch | Hugging Face | NLTK | spaCy | Scikit-learn | Pandas | WordCloud

How to Run

git clone https://github.com/Phoenixking-04/Sentiment-Analysis-NLP.git
pip install pandas scikit-learn nltk spacy transformers torch wordcloud textblob sumy rouge-score matplotlib seaborn
python -m spacy download en_core_web_sm
jupyter notebook NLP.ipynb

---

🔗 Developer: Kalyankumar Sandireddy