ARIAS U.S. RAG Chatbot

This AI chatbot is tailored to ARIAS U.S. content: it indexes the full archive of Quarterly articles dating back to the mid-1990s, enabling users to search by topic, context, or author. Built using LangChain and Pinecone vector embeddings over document chunks, the chatbot returns precise quotes, article references, and summaries drawn from the ARIAS archives.

Key challenges included ingesting decades of PDF/HTML archival formats, segmenting articles for semantic embedding, ensuring author-metadata linking, and balancing recall vs precision in context search. Also caching common queries and managing embedding updates as new Quarterly issues publish.

The ARIAS U.S.–centered chatbot enabled fast, accurate retrieval from the Quarterly archive, reducing manual search effort, increasing member engagement, and supporting research and arbitrator reference tasks.