Cartography on The Bhagavad Gita

A semantic and sentiment-driven exploration of the Bhagavad Gita using NLP techniques. This project extracts knowledge graphs, uncovers deep semantic patterns, and maps the relationships between characters and concepts using word embeddings, TF-IDF, BM25, and LSTM models—without relying on any large language models.

Project Objectives

• Scrape and preprocess the Bhagavad Gita verses.
• Perform semantic analysis using TF-IDF, BM25, and Word2Vec.
• Use LSTM for sentiment classification.
• Extract relationships between entities.
• Generate a cartographic knowledge graph of verses and concepts.

---

Repository Structure

File	Description
TBG_1_WEBSCRAPPING.ipynb	Web scraping Bhagavad Gita verses from online sources.
TBG_2_preprocess_semantic.ipynb	Text preprocessing including cleaning, stopword removal, and lemmatization.
TBG_3_tfidf_and_bm25_semantic.ipynb	Computes TF-IDF and BM25 similarity scores between verses.
TBG_4_wrd_embed_semantic.ipynb	Generates semantic similarity using Word2Vec vectors.
TBG_5_w2v_util_semantic.ipynb	Utility functions for word embedding handling and similarity computation.
TBG_6_Relation_extraction.ipynb	Extracts relational triples and co-occurrences among key entities.
TBG_7_evaluation_metrics.ipynb	Evaluation of semantic and sentiment-based models.
Cartography_KG_on_Bhagvat_Gita_using_LSTM.ipynb	Builds the knowledge graph and runs sentiment classification using LSTM.
English.csv	Preprocessed Bhagavad Gita verses in English.
w2v.bin	Pretrained binary Word2Vec model.
bhagavad_gita_word2vec.model	Word2Vec model trained specifically on Bhagavad Gita text.

---

Methodology

1. Web Scraping

• Scrapes verses in English.
• Saves data to a structured .csv format.

2. Preprocessing

• Tokenization
• Lemmatization
• Stopword removal
• POS filtering

3. Semantic Analysis

• TF-IDF and BM25: Score verses for lexical similarity.
• Word2Vec: Trained on Bhagavad Gita to compute deep semantic closeness.

4. Relationship Extraction

• Entity co-occurrence and dependency parsing.
• Generate relationship triples for graph construction.

5. Sentiment Classification

• LSTM-based model to classify each verse into positive/negative/neutral sentiments.

6. Cartography (Knowledge Graph)

• Constructs a Knowledge Graph using extracted entities and sentiments.
• Maps key figures (e.g., Krishna, Arjuna) and themes (e.g., Dharma, Karma).

---

Setup Instructions

1. Clone the repository:

git clone https://github.com/Anshul-ydv/Cartography_on_The_Bhagavad_Gita.git
cd BhagavadGita-Cartography

2. Install dependencies:

pip install -r requirements.txt

3. Run the notebooks in order for full pipeline execution:

• TBG_1_WEBSCRAPPING.ipynb
• TBG_2_preprocess_semantic.ipynb
• ... up to ...
• Cartography_KG_on_Bhagvat_Gita_using_LSTM.ipynb

---

Requirements

Include the following in requirements.txt:

numpy
pandas
scikit-learn
gensim
nltk
matplotlib
seaborn
tensorflow
keras
networkx
beautifulsoup4
requests

---

Sample Output

• Similarity matrices (TF-IDF, BM25, Word2Vec)
• Sentiment plot of verses over chapters
• Knowledge Graph visualization of key themes

---

Acknowledgements

• Bhagavad Gita translation sourced from bhagavad-gita.org.
• Inspired by semantic AI and traditional Indian scriptures.

---

License

This project is licensed under the MIT License. See the LICENSE file for details.

---