Conflict NLP :- Topic Modeling, Sentiment Analysis, and Classification using LLMs

Extracts insights from 26K+ protest events using BERTopic, Top2Vec, and LLMs for real-world applications like crisis monitoring, policy research, and social unrest analysis.

Project Overview

This capstone project uses state-of-the-art NLP techniques to perform :-

• Topic Modeling using BERTopic, Top2Vec, and LLaMA2
• Sentiment Analysis to assess public sentiment across global conflicts
• Text Classification for conflict categorization

The goal is to transform raw conflict data into actionable intelligence for policy makers, researchers, and humanitarian aid groups.

Key Highlights

• 26,000+ conflict records from ACLED and Google Trends
• Built 4 different topic modeling pipelines (LDA, BERTopic, Top2Vec, LLaMA2)
• Boosted coherence score for BERT-based topics
• Visualized topic dominance, distributions & coherence
• Preprocessed multilingual noisy text: stopword removal, tokenization, vectorization

Techniques Used

Task	Methodology / Tools
Preprocessing	Python, NLTK, RegEx, Gensim
Topic Modeling	BERTopic, LDA, Top2Vec, LLaMA2
Dimensionality Reduction	UMAP, HDBSCAN
Sentiment Analysis	Hugging Face Transformers (BERT-based)
Classification	Logistic Regression, SVM, RandomForest
Visualization	matplotlib, seaborn, pyLDAvis, Plotly

Repository Structure

├── notebooks/
│   ├── BERTopic_Protest_Classification.ipynb
│   ├── LDA_Protest_Classification.ipynb
│   ├── LLaMA2_TopicModeling_protest_analysis.ipynb
│   └── Top2Vec_TopicModeling_Protest_Analysis.ipynb
│
├── presentations/
│   ├── WorldBank_Final.pptx
│   └── GWU_Capstone_Final.pptx
│
├── data/            # Not uploaded due to size/privacy
├── README.md

Use Cases

• Crisis Detection: Detect and visualize emerging unrest topics
• Policy Research: Extract protest drivers across countries
• Social Analytics: Map sentiment trends over time or region

How to Run

1. Clone the repo: git clone https://github.com/your-username/your-repo-name
2. Install dependencies from requirements.txt
3. Run the Jupyter notebooks inside notebooks/

Contact

Surya Vamsi Patiballa
Graduate Student, MS in Data Science — George Washington University (GWU)

• Email :- svamsi2002@gmail.com
• LinkedIn :- https://www.linkedin.com/in/surya-patiballa-b724851aa/
• Resume :- https://drive.google.com/file/d/178IYcArC6YYVdJiIwRmJYodzKZ-JXe-D/view?usp=sharing

"Transforming data into dialogue. Insights into action."

If you found this project insightful, feel free to star it!!!