AI for document understanding

OCR and NLP as key to automated travel expenses or intelligent archiving

  • Sponsored by: Capgemini
  • Scientific Lead: M.Sc. Matthias Wissel
  • Project Lead: Dr. Ricardo Acevedo Cabra
  • Term: Winter semester 2018/2019

Documents in all forms and formats are the way businesses communicate with each other and to a big extend also still interact with their customers and employees. The information within these documents is highly valuable to the businesses in various situations. One possible use of this data would be to automate the processing of travel expense reimbursement. Another is archiving and auditing which often is necessary due to regulatory reasons. These tasks can be tedious if performed manually and repeatedly. To help in these processes and further leverage this information we can extract and process them, from raw documents or even scans or pictures of the documents.

The steps and building blocks to make this possible will be developed and combined in this project. To achieve this the projects spanning over the next two semesters will include the use of techniques from computer vision and optical character recognition (OCR) as well as natural language processing (NLP) and search to build a prototype application for one of the mentioned use cases. Depending on the available skills in the team the building blocks for the first term will be chosen. From the technical side we will be using python and available open source libraries and tools.

Results: The results of this project are explained in detail in the final documentation and presentation.