# Twitter Sentiment Analysis using NLP

This repository contains a Jupyter notebook (TwitterSentimentAnalysis.ipynb) that demonstrates a simple pipeline for sentiment analysis on Twitter data using classic NLP preprocessing and a Multinomial Naive Bayes classifier.

The notebook covers:
- Loading and inspecting the dataset (Twitter_Data.csv)
- Basic data cleaning and label mapping
- Text preprocessing: lowercasing, removing non-alphanumeric characters, stopword removal, and Porter stemming
- Converting tokenized tweets into features using CountVectorizer
- Training a Multinomial Naive Bayes classifier
- Evaluating model accuracy on a test split

Repository files
- TwitterSentimentAnalysis.ipynb — main notebook demonstrating the pipeline
- Twitter_Data.csv — dataset used by the notebook (expected to be in the repo root)
- README.md — this file

Getting started

1. Clone the repository
   git clone https://github.com/hasnatsakil/Sentiment-Analysis-using-NLP.git
   cd Sentiment-Analysis-using-NLP

2. Install dependencies
   The notebook uses Python 3 and the following packages:
   - numpy
   - pandas
   - scikit-learn
   - nltk
   - jupyter

   Install with pip:
   pip install numpy pandas scikit-learn nltk jupyter

   Or with pip and a requirements file (create one if desired):
   pip install -r requirements.txt

3. Download NLTK data
   The notebook uses the English stopword list. Run:
   python -c "import nltk; nltk.download('stopwords')"

4. Run the notebook
   jupyter notebook
   Then open `TwitterSentimentAnalysis.ipynb` and run the cells in order.

Dataset
- The notebook expects `Twitter_Data.csv` in the repository root. The dataset contains two columns used by the notebook:
  - clean_text: the pre-cleaned tweet text
  - category: sentiment as -1.0, 0.0, or 1.0 (negative, neutral, positive)

Model & Preprocessing summary
- Tokenize by splitting preprocessed text (the notebook applies a custom `tweet_to_words` function which:
  - converts text to lowercase
  - removes non-alphanumeric characters
  - splits into tokens
  - removes English stopwords (NLTK)
  - applies Porter stemming)
- Features: sklearn.feature_extraction.text.CountVectorizer with a fixed vocabulary size (5,000)
- Classifier: sklearn.naive_bayes.MultinomialNB
- Evaluation: train/test split (80/20) and accuracy score (the notebook shows ~0.744 reported accuracy on the test split)

Notes & suggestions
- The notebook performs stemming and stopword removal; depending on your goals you may prefer lemmas (WordNetLemmatizer) or to avoid stemming.
- Try TF-IDF features (TfidfVectorizer), n-grams, or more expressive models (LogisticRegression, SVM, or deep learning) for better performance.
- Verify label distribution and consider stratified splits if classes are imbalanced.
- If the dataset is large, use incremental/mini-batch training or more efficient vectorization.

License & Contact
- Add a LICENSE file as appropriate for your project.
- For questions contact the repository owner.

Enjoy exploring the notebook and improving the sentiment analysis pipeline!