Case Study: Transforming Cultural Heritage Research with AWS-Powered AI for ÇEKÜL

Transforming Cultural Heritage Research with AWS-Powered AI for ÇEKÜL

In our project to modernize ÇEKÜL Foundation's cultural heritage archive access, we developed an interactive chatbot solution that transforms how researchers engage with historical documents, periodicals, and cultural records. We are proud to have contributed to making valuable cultural heritage more accessible through cutting-edge AI technology. The system enables researchers to dynamically query and retrieve information through natural language interactions, significantly enhancing the research experience.

Project Story

ÇEKÜL Foundation recognized the need to transform their traditional research platform to remain relevant in today's technology-driven world. While their existing website provided basic search functionality for cultural heritage archives, it represented an outdated approach to information access. ÇEKÜL needed to modernize their digital presence and research tools, as users increasingly expect AI-powered, conversational interfaces. This technological transformation was essential not only to improve archive accessibility but also to position ÇEKÜL as a forward-thinking cultural institution embracing contemporary digital solutions.

Technology Solution

We implemented a sophisticated AI-powered research platform leveraging AWS services and modern AI architectures. The solution combines large language models, advanced retrieval methods, and intelligent orchestration to provide an interactive research experience for ÇEKÜL's cultural heritage archives:

• Language Model Implementation: Primary text generation using Claude Haiku via Amazon Bedrock, with two auxiliary Claude Nova Lite agents for reference validation and image matching, orchestrated through LangGraph
• Retrieval-Augmented Generation (RAG): Multiple RAG approaches implemented in OpenSearch including specialized text splitting and BM25 for optimal information retrieval
• Reference and Image Integration: Comprehensive reference management system that maintains source metadata and validates contextual relevance, plus image integration that matches relevant archival images to textual content
• Infrastructure and DevOps: Amazon EC2 instances hosting LangServe backend with Uvicorn server, React frontend application, and Langfuse for metrics collection

The information flow begins when a user submits a research query, which is processed by Claude Haiku. The RAG system then retrieves relevant information across multiple indexes, while a Nova Lite agent validates reference relevance. After generating a comprehensive response, a second Nova Lite agent matches relevant images, and the frontend presents the consolidated response with text, validated references, and contextual images.

What We Discovered Along the Way

During this project, our team uncovered several fascinating insights that shaped our approach:

1. The Art of Document Processing: We experimented with various techniques to maintain contextual integrity while optimizing retrieval quality. The way you slice and dice documents makes a tremendous difference in how well an AI can understand and retrieve information from them.
2. Specialized Agents Are Game-Changers: Creating dedicated agents for specific tasks like reference filtering and image matching dramatically improved our results. These specialized agents were much better at determining which references were actually relevant and which images truly matched the context of a query.
3. Balance Is Everything: Perhaps our biggest takeaway was finding the sweet spot between sophisticated AI capabilities and practical efficiency. We learned that thoughtfully selecting text processing methods based on content type led to better performance without unnecessary complexity.

This project resulted in enhanced accessibility to ÇEKÜL's cultural heritage archives through a conversational interface, streamlined research processes with integrated reference management, and added value through automatic relevant image suggestions. ÇEKÜL now offers researchers a modern, AI-powered platform that makes cultural heritage exploration more intuitive and comprehensive.

We are delighted to have been part of this transformation project that makes valuable cultural knowledge more accessible while helping ÇEKÜL Foundation embrace innovative digital solutions. The insights gained continue to inform our approach to developing sophisticated yet practical AI systems for knowledge management and research.