Dev Conference by Neo4j
Session Track: AI
Session Time:
Session description
In this session, you will learn how to take images and translate their content into a graph representation leveraging the Tesseract OCR Engine. Using the location of words identified by Tesseract, you will learn how to create a hierarchy of document chunk nodes--level, block, paragraph, and line. By having a hierarchy of chunks, you will be able to easily traverse different chunk sizes that relate to the same information. This can prove beneficial for RAG, where smaller chunks tend to be better for vector similarity and larger chunks tend to serve as better context for question and answer.
Manager, Pfizer
Kim Adler is currently a Data Translator Manager within the Digital Manufacturing Operations & Insights (O&I) Group at Pfizer. In this role, she serves as product owner and data scientist for an internal search engine for operations data that lives within the O&I Knowledge Graph database. Prior to joining Pfizer full-time, Kim completed a community detection capstone project with the O&I group as part of the Masters in Business Analytics program at MIT Sloan School of Management. When not in deep thought about graphs, Kim can be found going on long walks with her dog, Sophie.
