Manually extracting data from scanned documents is an expensive, time consuming and error-prone process. We use Machine Learning to extract the data automatically and convert it into a format that can then be used in your downstream workflow. The scanned data to be extracted can be typed or handwritten, structured within tables and forms or a complex combination.

A human “sign-off” or oversight over the automatically extracted data is required in several use cases (for instance to ensure accuracy because the data is sensitive.) This has to be done in a way such that the efficiency gains due to the automation are not negatively impacted - bringing into question the gains due to automatic data extraction itself. We have designed innovative mechanisms such that the reviews (either by an individual or a team) are as efficient as possible.

The automatically extracted data enables further capabilities to be built on top. For example, in applications with unstructured data, there is a need not only to digitize data from scanned documents, but also to go a step further and make sense of the data. NLP is used, based on the application requirements, to gain insights from the extracted data. For example, with medical records, it is required to digitize and validate (inline with several medical standards) the extracted data. Further analysis on the data to, for example, identify trends, inconsistencies etc. help the medical team to ensure optimal healthcare. 

In real world applications there is wide variation in the document format for a given application. For example, consider an application to extract information from all invoices received by a business. While some information is essentially similar across all invoices, there is likely to be variation in 1) how the invoice is formatted; and 2) additional information based on the industry/vertical. Our framework/REST APIs are built to take into account these variations so we can quickly insert this capability into your workflow (and help implement the necessary customizations) or provide a complete solution for your application end-to-end.