通过生成式人工智能和知识图谱提升文化档案访问.pdf

1、ENHANCING ACCESS TO CULTURAL ARCHIVES WITH GENERATIVE AI&KNOWLEDGE GRAPHSRAJESH KUMAR GNANASEKARANBackground on MSAs LoSWhy focus on Maryland State Archives Legacy of Slavery(LoS)?The Maryland State Archives(MSA)undertook a multi-year effort to digitize historical records from the LoS project.Projec

2、t to discover stories of unknown heroes of slave flight and resistance.This effort resulted in a large,digital database containing over 400,000 records,including Certificates of Freedom(CoF),Domestic Traffic Ads(DTA),and Manumissions.Goal of this project:to enhance access to these digitized collecti

3、ons of the LoS datasets in an ethical and scalable manner.3Figure:Sample images of the DTA scanned ads3Domestic Traffic Ads Dataset sampleFigure:Sample images of the DTA scanned adsManumission Record SampleFigure:Sample image of the Manumission record issued for Polly and her children to be free on

4、December 1,1789Certificates of Freedom Dataset sampleFigure:Scanned CoF document of Lot Bell issued in Caroline County in 1816Leveraging Generative AI SolutionsChatLoS v1 using OpenAI GPT&RAG(Retrieval-Augmented Generation)With ChatGPT introduced in late 2022,we experimented with a conversational ch

5、at interface(“ChatLoS”)that pulls context from the Domestic Traffic Ads(DTA)dataset and provides context-aware responsesDemonstrates interactive natural language conversational capabilitiesGreat for targeted semantic search but limited in contextIssues arise around context-limitation,unable to perfo

6、rm data aggregation analysis or cross-collection analysisUnderstanding Retrieval-Augmented GenerationHow RAG Enhances ChatLoS for LoS accessWhat is RAG?Retrieval-Augmented Generation(RAG)is a hybrid AI approach that retrieves relevant information from external sources before generating responses.Thi