Executive Update

Utilizing AI to Extract Structure from PDFs

Posted February 16, 2021 | Technology |
PDF data
In this Executive Update, the authors describe a system using AI technologies to automate data extraction to any one of many structured formats. The system performs minimal manual annotations to capture the semantics of specific sections for any particular document template. The authors highlight the business drivers behind such a system, describe the architecture of the system, show how the system performs compared to human-assisted analysis, and showcase examples of documents processed. In addition, they cover some difficulties around this process and share details about the neural network architecture they use to achieve high accuracy.
About The Author
Shahane Eksuzyan
Shahane Eksuzyan is a research scientist at DataFoundry.ai Labs. Dr. Eksuzyan’s research interests include ML and AI. She is an expert in using Python for prototyping and developing models. Prior to getting involved in computer science, Dr. Eksuzyan earned a PhD in chemical physics from the National Academy of Sciences of the Republic of Armenia. She can be reached at seksuzyan@datafoundry.ai.
Sedrak Vardanyan
Sedrak Vardanyan is a team leader and senior research scientist at DataFoundry.ai Labs. He is an experienced researcher focused on applied science. For nearly 20 years, Dr. Vardanyan has been dealing with mathematical modeling problems through different fields of applied science: mechanical engineering, economics, university education, and business. He can be reached at svardanyan@datafoundry.ai.
Raj Ramesh
Raj Ramesh is a transformation expert who helps individuals adapt and organizations transform in this new world where AI is becoming commonplace. He is Chief AI Officer at DataFoundry and has worked with senior-level business and IT leaders in Fortune 500 companies as an advisor and architect to foster deeper business-IT collaboration through consulting, advising, and building customized frameworks. Dr. Ramesh has written extensively and spoken… Read More
Don’t have a login? Make one! It’s free and gives you access to all Cutter research.