Image Analysis for Historical Knowledge Discovery and Preservation

Digitized archival collections offer rich sources for cultural heritage research. Current systems, though, fail to effectively integrate heterogeneous data forms like images, textual descriptions, and metadata. This study offers a sophisticated framework that blends image recognition, natural language processing (NLP), and knowledge graph methods to facilitate access, enable discovery, and reveal concealed historical patterns. The system comprises several modules such as a data input module, preprocessing pipeline, image recognition, NLP analysis, knowledge graph generation, semantic search engine, and an interactive user interface. Testing the system on a large historical archive dataset proved high accuracy, efficiency, and usability for all modules. Major takeaways are 93.4% accuracy in object detection for images, 94.8% accuracy for named entity recognition, and 95.5% search accuracy in semantic search feature.