July 11th, 2025

Using AI to Extract Structured Insights from Seerah Texts

Using AI to Extract Structured Insights from Seerah Texts At Greentech Apps Foundation (GTAF), we are exploring how artificial intelligence (AI) can be used to unlock deeper insights from the Seerah (biography of Prophet Muhammad ﷺ). In our latest R&D project, we applied keyword extraction and topic classification techniques to the classical text Ar-Raheeq Al-Makhtum (The Sealed Nectar). By leveraging tools like natural language processing (NLP) and clustering algorithms, we identified key themes, structured events, and relationships across chapters. This allows us to make the Seerah more searchable, interconnected, and useful for both learners and developers. These insights will help us build more intelligent Islamic apps where users can discover relevant events, filter content by topics (e.g., battles, hijrah, treaties), and link Seerah to Qur’an and Hadith more effectively. This marks an important step in using AI for Islamic education and digital tarbiyah.

Published: 20 September 2024

Author: Mohammad Galib Shams, Nabil Mosharraf Hossain

At Greentech Apps Foundation (GTAF), one of our research initiatives involves applying cutting-edge AI models to classical Islamic texts, including the Seerah (biography of the Prophet Muhammad ﷺ). Our recent project focused on automatically extracting structured insights—such as people, places, events, and dates—from raw Seerah paragraphs.

The Challenge

The Seerah is rich with historical events, characters, and locations. However, much of this information is buried in long, narrative paragraphs, making it difficult to analyze or search programmatically. Our goal was to convert unstructured text into structured data to enable better study, visualisation, and linking across related events.

Our Approach

We used the LLaMA 3 70B model from Groq through the LlamaIndex framework to generate high-confidence keywords from Seerah passages. Here’s a high-level overview of how it works:

Prompt Engineering: We crafted detailed prompts that guide the model to extract keywords by category: Person, Location, Event, Group, and Date. For each event, the model also links the keyword back to the exact part of the paragraph it was derived from, and assigns a confidence score.
JSON Output: The model is instructed to return a structured JSON output, which allows us to save and query the extracted data easily.
Keyword Mapping: We plan to use this structured data to map events chronologically and thematically—helping build a knowledge graph for the Seerah.
Automation: The project includes code that runs inference on paragraphs, stores results in a database, and periodically checks for weekly summaries to send to stakeholders.

Example

From a passage describing the Quraysh’s reaction to the Prophet’s migration, the model successfully extracted:

Persons: ‘Ali, Abu Bakr, Asma’
Locations: Al-Ka’bah, Makkah
Events: Beating of ‘Ali, interrogation of ‘Ali, assault on Asma’

Each item is tagged with confidence scores (e.g., 0.98) and anchored back to the paragraph for verification.

Potential Applications

Enhanced Seerah Visualisation: Timeline builders, maps, and character networks.
Cross-Linking with Hadith: Match extracted people or events with related hadith entries.
Educational Use: Build interactive learning experiences with tagged Seerah content.

What’s Next?

This project is still in its early stages. We plan to fine-tune prompts, evaluate accuracy more rigorously, and eventually publish a dataset and research paper based on our findings. The long-term vision is to support a new way of engaging with Islamic history through intelligent, structured technologies.

Share the Knowledge