📰 BBC News Topic Modeling

Discover hidden topics and themes from BBC News articles using Unsupervised NLP techniques: LDA and NMF. Explore insights, visualize topics, and compare model performance! 🚀

---

📂 Dataset

We use the BBC News Dataset (Kaggle).
The dataset contains news articles categorized into topics such as sports, politics, business, technology, and entertainment.

---

🛠️ Tools & Libraries

• Python 🐍
• Gensim
• Scikit-learn
• NLTK / spaCy
• pyLDAvis
• WordCloud

---

📝 Preprocessing

• Tokenization
• Lowercasing
• Stopword removal

---

📊 Topic Modeling

We applied two unsupervised NLP techniques:

1. Latent Dirichlet Allocation (LDA)
2. Non-negative Matrix Factorization (NMF)

Both models extract dominant topics and display the most significant words per topic.

---

🖼️ Results / Visualizations

LDA Visualization

The interactive LDA visualization is saved as: lda_visualization.html

Open it in a browser to explore topic-term relationships. 🌐

WordClouds

LDA Topics:

NMF Topics:

Each image visually represents the most significant words for the topic. 🌟

---

📈 Model Performance

• LDA Coherence Score: 0.3530
• NMF Reconstruction Error: 45.8986

Compare model performance to understand which approach captures topics more clearly. 🔍

---

🚀 Usage

1. Clone the repo:

git clone https://github.com/your-username/bbc-news-topic-modeling.git
cd bbc-news-topic-modeling

2. Install required packages

3. Run the notebook or Python script:

python topic_modeling.py

WordClouds and LDA visualizations are saved automatically. ✅

---

⚖️ License

This project is licensed under the MIT License. 📝

---

Made with ❤️ and ☕ by Adel