Annual reports give a comprehensive overview of companies their activities and (financial) performance. Reviewing these annual reports, however, is currently a manual task. Besides that this is laborious, it is also biased.

As PwC's role as a trusted advisor we aimed to improve the review process its quality and objectiveness using machine learning. Therefore, we developed an application that quantitatively analyses the content of annual reports and provide the auditors with several metrics such as readability and sentiment score.

During this talk we will provide insight into our process by taking you along the journey an annual report will follow. From the uploading of the PDF report to our application on Azure, to the extraction of text via OCR with pdftotree and ending at the analysis of the text through NLP with spaCy.