Adventures in Understanding Documents
summary
A Clojure pipeline extracts text from PDFs using PDF‑to‑text and tesseract OCR, shells out to small binaries, applies regex and fuzzy-clj matching (threshold ~68–75%) and core.match rules, and stores weighted facts in a graph (forj/neo4j) so each processed document increases contextual associations. Endpoints are exposed as REST resources with Liberator and bidi, messages flow over a fast bus, and the system incrementally improves underwriting decisions — it runs in production and reduced manual tagging.
AI-written summary · about
chapters (4)
related talks
39mEnterprise-ready document processing for AI
31mConquering PDFs: document understanding beyond plain text
18mFree Texts to Knowledge Graphs - Holy Grail of Semantic Computing on an Industrial Scale
37mA novel way to manage and represent unstructured clinical data
33mClojure Dispatches From a National Research Lab
41mAccelerating privacy-enhancing data processing
29mBetter Understand Your Documents with AI
43mImplementing AI Document Processing at Scale: Syngenta